J

Josherich’s Blog

Josherich的博客

https://www.josherich.me/
https://www.josherich.me/feed.xml (RSS订阅地址)

423 四十年目睹之重振少林往事

423 四十年目睹之重振少林往事 Just Pod 可以说是最近40年来对少林寺这个品牌，我觉得算是鞠躬至伟的，但是备受争议的一位人物——释永信方丈。当然现在他已经被应该消除了他的界谍，应该改成俗家名了——刘某某。 总之这么一个在中国的宗教界，或者过去的那种文化版图上非常亮眼的一位人物，就是现在轰然倒塌。而且关于他的传闻也有非常多，但是到目前为止，这些的真假我们都还很难去辨别。 但毫无疑问，少林寺这样的一个寺庙，以及它背后代表这个产业，它一定是经过了所谓的试用性时代。如果我们用这五个字的话，这个时代已经结束了。 所以我们今天要找到王凯老师好好聊一聊这个话题。为什么聊这个话题？是因为我在翻这些旧刊物的时候，我就发现三联生活周刊早在2010年就用一期封面报道来写过当时的少林寺和试用性。当时好像也是在挖一个关于他的一些背景的材料。 那会是因为一个什么样的原因你们会写它，而且是用直接封面的形式来写？我不知道大家有没有印象，就是一直在传说这个寺院要上市——少林寺上市。10年那会儿，其实关于他的传说一直非常之多，关于少林寺关于试用性的，每隔几年就一个，每隔几年就一个，这一次是终于到了一个休止符。 真的是这样。 这个寺庙当然我们过去在什么武侠小说，或者更久远的这些历史文献当中，少林寺它的名字也经常出现，而且确实从中国的佛教历史的角度，它的地位就是很高。 但不可否认的是，在20世纪，少林寺变成今天成为一个产业，达到这样的一个规模的话，那确实是伴随着释永信80年代进入这个寺庙，并且一手操盘，通过各种政商关系，包括他自己的一些聪明才智，把它营销到了今天的这样的一个规模。 所以这样的一个人轰然倒塌，深陷淋遇的这样的一个结局，挺符合改革开放时代的很多故事的。我们以前看激荡30年，或者什么大白局，这种故事里面有很多企业家好像都是这样子，激荡了快50年，也相对最精彩吧，我觉得非常精彩。 比如说一个衬衫的倒塌，一个房地产的倒塌，建立宝的倒塌，就是它可能太商业了。 是。 但少林寺它背后也有很多的文化层面的，文化、外交、宗教，我觉得它好像辐射特别广。 是。 你要说它光是个商品倒塌，但它不会有这么强烈的传播冲动，也不会有这么多故事。 就像这几天大家在猜测它倒塌的原因，到底是因为那些外交层面的： 你见了那些不能见的人 还是说更政治层面的，你背后的谁是谁，发生了什么变化？ 我们可以说吗？我觉得是踢嘴。 我们也不要去忘加采策。 对。 特别高兴见到陈一良。然后我必须解释一下，上次我被骂了，大家老说我声音忽大忽小，因为我上次特别重的感冒，一咳嗽，我就想躲在旁边，被大家分分辱骂。 然后另外一件事，没最准这麦克风。 然后另外一件是说我没有记者的善良，好像是因为讲了地震的一些什么狗在叫那些的。 其实我觉得，就是我们讨论一个问题特别多面。你可能你看到的现象和你在新闻报道里写的，包括你内心感触，一定是多重的。 就像我们今天在讨论这个寺院问题的时候，它也是一个特别多面的问题。 对。 我这两天看到的很多，就是在说这个所有的钱都被和尚拿去找小老婆了，都去生孩子了，那显然不是一个真实的面相。 对。 但是你要说这个寺院的经济是怎么回事，其实我们也是不知道的。因为它确实不是个企业。 好像是discovery去采访过少林寺的时候，他们不知道怎么翻译方丈，那时候就给释永信了一个翻译，然后就说怎么翻译呢，然后就把释永信翻译成了少林寺的CEO。我觉得其实恰恰是因为他的名声，给他带来了众多的问题。 尤其少林寺绝对不仅仅是一个登峰的寺庙，不是一个河南的寺院，他可能真的是一个全球的少林寺，他在公众的视野之中，就导致他的复杂性呈现出各种面相。 我觉得这个特别有意思。 我们10年去的时候，其实已经感受到那种很复杂，很多重或真或假的信息，然后或真或假的叙述方式。我当时是跟我们的，现在的三联主编李鸿鼓一起去的。 我觉得我们都还算老记者吧，但是都还是只能获得或真或假的信息，而并不是说我们写的报道就是一个真相。 很多报道都会说“我有真相，我有真相”，但最终大家都没有真相。 就是那次10年，就是为了当时少林寺上市的传闻。 对。 其实当时解释很清楚，而且我觉得他的解释是解释得通的，因为大家知道中国有一句俗话： “什么名山大川，大概都被名寺所占了。” 所以当这个寺院开始变成旅游景点的时候，一个最真实的问题就浮现在眼前了： 这些钱该谁收？ 是旅游部门收还是该寺院收？ 然后当时10年的时候，为什么会谣传少林寺要上市？其实大家如果翻一下当时的报道很清楚： 当时有个香港的港中旅，想跟登峰合作，就是把少林寺开发成整个景区，然后这个门票收入归他们所有，他们想在香港上市。 但是以释永信来说，凭什么这个门票不分给他一部分，觉得这应该是寺庙的收入，也确实就是说他至少应该分一部分。 所以当时他就作为全国人大代表，他其实就一直提出说： “对宗教场所，要不就你不收票，你彻底不收票，国庆寺那样；要不就是你收费，你跟我分成；要不就是你又收了费，大家都觉得是少林寺收了费，而我又没有分到钱。” 以他的性格和以他多年的经营，因为到10年的时候，少林寺其实经营的已经很有起色了。 我记得当时我们看到的，当时一年的收入会超过一个亿，就松山少林寺明胜风景区的门票是100块以上，一年收入至少超过一个亿。 规定是少林寺能拿到30%的收益。 我觉得当时是在这个收益上互相有了争执，所以会有一部分关于少林寺要上市的谣言，在海外先起来，然后在国内也起来。 所以这个也不仅是少林寺的问题，我觉得是中国无数寺院都有这个问题。 这不是跟刚说的改开以来的那些商业公司产权问题一样吗？以前都是国资的企业集体企业。 这两天看到很多大家都在说中国寺院赚了好多钱，国美山收了好多钱，然后包括我们去西安，比如大印塔，那个大森寺收了好多钱。 其实这里面很多钱不是寺院收的，是文旅部门收的。 这里面的区分你是不是要跟大家解释一下？ 这个特别有意思。 就是因为我们中国寺院的历史不是一直持续的，时常有中断。 我看大家前两天也在讨论历史上的灭佛，很多阶段，这个寺院就是荒废无人的。 我们就都知道中国是49年之后，因为我党是无神论，所以就有神论有信仰的这些地方都在逐渐的衰落。 当然彻底的把和尚赶去还俗，可能是发生在60年代。 然后到80年代，应该是1982年，胡耀邦他们当时提出过，重新推动了一个宗教政策，寺院重新大规模建立。 这个政策好像特别的宽容，以至于很多赵普初就说过： “这真是一个菩萨一样的政策。” 就可能大批的承认寺院的存在，承认僧侣的存在。 然后当时就有一个非常明确的规定，我是在西安大臣寺听说的： 这个寺院如果是有和尚，这个寺院就归宗教界所有了。 82年那个点，如果有和尚，对，就是那前后，可能是当时已经开始出现这种情况了。 就是有大批寺院重新建立了，重新恢复了。 然后这个寺院如果说找不到和尚，确实一个都没有了，所有的和尚都还俗了，或者说已经荒废了。 然后这个寺院又很出名，它就归旅游部门所有。 我们现在都看到很多寺院门口的收费的人不是僧侣，那都是工作人员。 正好前一段去大同，大同有两个非常著名的寺院： 寺院一个是华言寺 一个是上华寺 它已经完全就是在他们上一任那个很红的网红市场，在它开发的过程中，就把本来的佛教协会牵走了，就是你们去另外的地方。 这个完全变成文物，归文物和旅游部门所有，它就成了一个旅游职能。 那你说你现在去的华言寺，你去的是一个寺院吗？你更应该说我们去的是一个博物馆，去的是一个寺院遗迹，没有和尚的遗迹。 所以这就是一个特别复杂的事情。 但少林寺呢，跟这个又不太一样。 因为首先它是个寺院，而且它寺院规模很大，它名声很大。 我看释永信那个回忆里面，就是他80年代初去到少林寺的时候，虽然很破败，但里面还是有20多个僧人在里面出家的，还种地。 少林寺的历史就特别复杂，特别有趣。 因为要跟那个昌儿做节目，这两天还一直在研究，包括我们自己当时去，也看了很多资料。 因为80年代的时候，特别穷困的时候，少林寺的公开说法，就那个电影还没拍的时候，少林寺的自己寺院的历史，他也会说他可能有15、6处的篱笆是没有的，是他们连篱笆都修不起。 然后我们有一个开国中将，叫皮定军，他40年代的时候在那边打游击，然后44年在那边建立了抗日武装的根据地。 然后他有过几本回忆录，都反复说过少林寺。 大家现在看到很多少林寺的资料，都是从他那里来的。 是。 也特别有意思，他44年时候去的时候，他有第一本回忆录是60年代出的，就文革前期出的。 然后里面说，当时少林寺就有一群人，武僧就不怀好意地看着他。 然后芝克僧就拖住他，好像是一定要他吃东西，还要让他喝茶什么，让他喝茶。 其实他当时是想参观大殿的，就是参观后面，就是没有被毁坏。 因为虽然经过直奉大战，但少林寺还有一些大殿还留存了。 传说什么28年当时十有三火烧少林寺，确实是破坏了他的主体建筑，基本都破坏了，但是也留了很多明代建筑，也有很好的壁画，藏宁阁这些全都是后修的。 但是你想我们现在去少林，也能看到很多很好东西，比如说后山的塔林，还有一些宋代人物的什么黄庭坚的提字什么。 对对对，其实还是有大量的东西留存的。 所以他44年的时候，皮定军去参观的时候，他就说有知科僧不怀好意的拖住了他，然后有一群武僧拿着棍棒，虎视眈眈。 然后他就立声喝制了他们说： “我们是抗武装，你们不要跟日本人搞在一起。” 但是大家想想，这本书是60年出版的，就是我们刚刚土地改革的时候，50年代。 它是明确的有一部分僧人被定成地主了，有一部分人还是贫农，就是僧人大概也分等级嘛，也能够想象。 然后而且五几年，他那边就大概有些环俗的僧人，就成了生产队，好像叫登峰什么城官大队，锅庄生产队，大概都是僧人组成的。 但是确实他也有一部分没有还俗。 然后到86年，这个皮定军再出回路的时候，对上林寺就很多的表扬了。 就是上林寺里有很多抗日的僧人，跟他们结成了抗日武装，而且当时那个地方，就是成立一个抗日的政府。 有一个副区长，就是一个武僧，已经还俗了的一个武僧。 他没有还俗，因为四几年嘛，你说对他们没有还俗的需求。 当时还有记录就是说，都是他86年那本传记里面说到的是，当时有一个抗日的县长，晚上夜宿少林寺，而且有人来抓他，还被少林的老僧人和保护，然后悄悄派人送他下山了。 所以我们可以想象，就是少林的武装力量一直都有，而且他有大量的田地，因为在过去的漫长的时间里，中国很多寺院是有自己的地产，有自己的产业的，这个其实也不奇怪。 大家如果看看历史书，就会发现很多寺院是有自己的产业，它才能生存。 对。 而且比如你前面提到，像当年什么三五一宗，他们搞灭佛运动，其实背后都有这种经济层面的和寺庙争夺人口、夺取这些财富这方面的一些考量。 我们从历史学的角度来看，一定是这样的。 但当时意识形态宣传一定不是这样的，一定是有别的宗教，或者说你怎么样怎么样。 说起来很搞笑，我高中买过一本书，我觉得可能就是90年代以后释永信和他的团队整理出来的开始宣传了。 对。 就是那本书是个合定本，叫做武当三十六宫与少林七十二义，里面记载了108门武功。 而且他说没本武功，当然记录的很简单。 是密集吗？ 是密集。 里面就教你，比如说易纸禅怎么练，还有我记得当时有个什么锦拳弓，我当时还练过几个礼拜。 那个锦拳弓就是怎么样，说你找一口锦，就往里面每天挥拳。 比如说一开始两三年，可能你就发现挥拳下去毫无反应，不要气馁。 等到四五年的时候，开始紧紧，水波动，那个水开始有一些波澜。 到了你练到七八年，然后你一拳挥下去，那个水就咕咚咕咚的，就很大的响动了，说练到十五年以上此功已成。 说你一拳出去，如巨石投井，然后那个水能溅到井栏上。 说这个时候你就练成了。 然后你如果对人使用这招，可以在四五米开外隔空将人击倒。 说还好我只信了两个礼拜，这事练不成，都是因为你没练十五年。 十年很难去验证。 但后来我在想，这个东西到底是真的是少林寺当年的一些，还是说就是九十年代这些地摊坐着们自创的，我觉得特别有意思。 因为武功是一个无法证实的事情。 我前两天还在看徐好峰的小说，就是二十八年的时候，我们是有全国的国术大赛的，是武功比赛。 然后之后这个事就没了。 然后你现实中不存在说，我去踢武馆，我去打架，我去群殴，那个都属于要被公安制止的行为。 然后包括少林寺的武功呢，也都成了一个仅限于传说的事情。 我们一零年去的时候…… 因为专门也是要问这个嘛，也是对这个很好奇。皮定军一九五七年的时候重回少林寺，然后我看皮定军回忆录说，有一群老僧人，老武僧，还专门给他表演了武功。 然后大家知道，共产党我们的一个开国大将许世友，他是少年时代在少林寺，好像学过吧。是民间传说，说什么许世友能飞檐走壁什么的，那至少说明在民国的时候，在少林寺学习武功及练习武功，它不是一个完全虚假的事情。 对，它是有一定基础的。就跟我们一零年去的时候发现，因为当时少林寺做了三大块： 一个是练武 一个是药局，就是医药局 一个是修禅 这是当时世友信跟我们说，这是他的软件升级，就是他做了三大块。我觉得他很聪明，他就从中国的传统文化中找了三大块三大产业。 现在这三大产业大家弄得也很嗨，都可以上市了。对，所以一个是医药，也就是少林药局。他说武术这块，少林武术呢，有一个很大的原因，当然我们都知道，是因为那个香港不断的传播。 对，我们知道民国的时候的武术记者，平江不悄生啊，王杜卢啊，然后最典型的当然我们后面都说，梁文生和金庸啊，因为这个系统在香港没有断绝嘛。 然后当然对大陆来说，影响最大的肯定是八十年代初的少林寺。但少林寺刚才跟陈良在说，也不是在少林寺拍的。 对，少林寺那个取景是在国庆寺，在台中拍的。而且这个很有意思，我就是去年去到国庆寺的时候，当地人跟我讲这个故事，说因为少林寺首先，它确实地位很高，它是那个禅宗的祖庭。 然后国庆寺是那个天台宗的祖庭，但是天台宗在日本的话非常盛行嘛。所以当年田中角荣，据说是他的母亲是天台宗的信徒，所以田中角荣就提出要来国庆寺，就是进来参观一下。 那这个事传到周总理，他就让浙江这边汇报一下现在国庆寺的状况。但是后来得知说国庆寺在那会儿变成了一个纺织厂，就是僧人都没有了。 然后他就立刻让这些纺织厂赶紧挪个窝，把这些僧人都招回去，然后从北京特批了一百多件文物送到国庆寺去装点。这还不一定是国庆寺自己流程，肯定不是了。 那就是可能把北京的一些东西可能就运过去了，反正装点门面。这个就导致国庆寺就拿到了一笔装修的资金，它就变成了70年代中国最美的寺院。 当然国庆寺本身，它那个条件也还不错，大环境非常好。其实很多寺院这么保留的。我最近刚去大同永安，有个很小寺院叫浑原寺，当时就是因为他们被当做了凉库，所以寺面的壁画全部留下来了。 就金代和明代的壁画留下来了，堆在骨子里面，它就被遮住了，就特别好看。就真是大家会觉得怎么就都没有损毁，其实它出于种种偶然性，也被留下来的还是挺多的。 对，所以当时应该是拍少林寺的时候，可能从取景的角度，是没有取少林寺的景。我看到一些，就塔林那些还是少林寺，也有些河南的别的文物景观。 对，释永信就是在那前后，应该81年的时候就投奔了少林寺。他是从安徽引上线到少林寺，所以这个也很奇怪。 其实因为他自己都说得不清楚，你有问过他吗？没有。你在采访他的时候，他只会说自己想说的，他绝对不会顺着你的思维走。 我觉得他已经酒精考验、酒精沙场，所以非常奇怪。他会说他大概是一个过年的时候去的，家里人都不在，他就偷偷拿着衣服去了。 但是实际上，这个引上线跟河南登峰是隔得很远的。对，不知道为什么他会去那边，而且当时那个电影并没有放映，所以他去的算是早的。 就是他的信息是从哪来的，还是说他其实没有那方面的信息，就是误大误壮。误大误壮，或者说民间一直有传说。 就是因为民间传说的力量有可能比我们知道的更持久，或者就是一些更私人的原因，比如家庭里面有谁对去过，或者说那边有人。 但是因为他社会不清楚，但我们知道他去的时候少林寺是非常破败的，跟我10年去的时候完全是不可同流而喻。 我去的时候，其实我们是被他请到方丈室去接受采访。我跟我们的李宏谷李大人，我是被豪华所震撼了。 你知道他那边有一个巨大的那个我们拿玉石雕的那种巨大的假山，不是叫山子吗？我只有在故宫才看到过那么庞大的一个山子，只是说雕工当然不能跟北京故宫比了。 然后可能因为很热吧，我记得他给我们每个人递过一把这个檀香扇，就是真的是非常豪华。 他也不忌讳对普通记者展现他的豪华，而且我觉得他很有分寸。你看我们去那，我们是自己住的，寺院没有招待我们说你们住或者吃。 当然他们会说你们可以住在附近的民宿里，这按武侠小说不应该有个什么知客僧来引导一下你们？有知客僧，但是你肯定是事先开始就联系好了。 但是他就把你们安排在附近的民宿里，而且是自己付钱的。我还记得我跟李大人还每天在附近的民宿还每天喝点啤酒。 就是他做的很有分寸，不卑不亢，他没有讨好你。 你当时觉得这个人怎么样？我觉得他太像一个传说了。我们对于传说中的人见到真人，其实很难有清晰的判断。 而且我相信他一定见过太多人，所以他的整个呈现给我们的人，其实距离他真实的人，我觉得有非常厚非常厚非常厚的一层铠甲。 这是一切名人都有的手段，他就更加有了。但有一些名人是那种，比如你之前听说过他名气很大，你见了一面之后会立刻很失望，你觉得这个人怎么这么贫用。 他会给你这种感觉？没有，他有一定的复杂度。而且我觉得他至少给了你一个真诚的面貌。 比如说我们就前面说到这个收费不收费的时候，盛永迅立刻就说： “我们特别不愿意这个收费，为什么呢，因为门票收入太贵，对我们的信徒进来是有很大的问题的，就是信徒需要拿规疑证，然后报批。他就会觉得是防止扩大我们的影响力的。” 当然他也是从一个现实角度来提出这个问题的。 事实上很多寺院的方丈也会普遍有这个说法，就是说你们的这个会妨碍我们的信徒进入。 所以现在很多这种区域也会有一个说法说，你有皈依证可以免费进，他就非常明确地说，这样防止我们更多的发展。 而且他特别好笑，你会觉得他当时已经很出名了，是个强者，但是呢，他很愿意展示出一个我们其实是被欺负的对象。 谁欺负他？我举个最简单的例子，刚才我跟长尔也在说，特别好玩。 因为大家都说适用性注册了很多少林寺的商标。对，但事实上最早注册都不是适用性，都是社会各界，因为大家把它当作一个公共的名声在注册了。 我们都说少林武校有无数的武校，所有少林寺外围的武术学校都是当地农民办的，都跟他们没有任何关系。 然后这个还不足为奇，我当时听到一件事，我是真觉得他们被欺负了。 大家知道八十年代最出名的有个火腿肠产品，而且是最早的，就叫少林寺。火腿肠叫少林寺。 我想如果我是寺院里的和尚，我也会生气的。 因为，当然他私下偷不偷吃肉，这个我们不知道，但至少公开来说，你凭什么你跟火腿肠注册少林寺呢？ 这还是近乎侮辱的一种注册吧。因为他们八十年代去少林寺的时候，少林寺还很贫困，是在他们的努力下，少林寺一步步起来了。 所以我觉得他会天然地认为这个商标是我的，只能我用，你们不能用。我觉得他的这套想法的逻辑是很清晰的。 那会儿你有感受到，比如或者他跟你呈现出来，或者你通过任何渠道有打听到，比如当地的这些，比如登峰政府。 登峰政府就是他的，我们说用仇人这个字也不合适，但至少他们是一个非常紧张的商业关系，他们是有摩擦的，非常清晰。 当时我记得很多报道把这个摩擦讲得清楚的，只是我们现在不记得了，因为发生15年了。 因为我听说过一些小道消息没被验证过的，说当地的一些，比如政府呀，或者当地的一些势力，也在试图推动适用性来做一些上市之类的，这些商业化的一些举措。 因为少林寺作为一个寺院很难上市，他除非说他合资了很多公司。但是那个阶段他刚刚起来，所以更多的是已经上市的港中旅和登峰政府合作，我们再做一个旅游公司来上市。 我觉得这个更合理，就是说他的竞争对手更想上市，所以这个方案里面就排除了少林寺，排除或者只分给他很少的钱。 但是他觉得你们都是因为我，你们才有这个，你凭什么给我这么少的钱。 所以当时是一个非常公开化的矛盾，可能时间久了大家就真的忘了。 当然怎么回事，都觉得少林寺挣了钱，至少我们10年去的时候，他还处在一个想挣更多钱，但是在不断规划他的武功系统、他的医疗系统、他的禅修系统，因为这是他更精通的买。 而且当年有一个产品很好玩大家可能现在不记得，叫少林素饼，就是一种很圆的一种小素饼。 所以当时我也是去了之后，还挺好吃的。然后少林药局提供很多各种各样的保健品。 其实大家想想在10年，是比较先进的。什么助眠的，你知道少林最擅长的是什么吗？医学方面，推拿吗？针灸，有点像骨科，因为它跌打损伤太多了。 精创药什么的，他们会有一种黑膏，就是自己号称是流传下来的。不是黑欲断食膏。 ——黑欲断食膏叫黑欲糕，专门就是春节，或者什么时候，就会分给寺里的每个外夫用的。 外夫用的，因为他们当时28年就石油山去烧少林的时候，号称是当时的医方全部都烧毁了，就是方子全部烧毁了。 然后他们留下来的呢，就有很多乡村的，有大批的推拿医生。 我去的时候遇见了一个老医生，这个老医生叫士行真。他说他19岁的时候，那时候还在寺院里，就是赵不出，老见过他，并且动员他说你去上中国佛大大学。 然后他就去了那个北京上佛学院，然后之后就去西安一直在当一个医生，然后到他五六十岁，又回到家乡，七十多岁，被释永信就招到他们这来了。 他们那个针是两头针，说少林过去有一本经典叫血头经，根据血流的速度来给人扎针。 所以释永信，我觉得是吧，所有乡村的有关少林的这些人，都笼到了自己的身边。挺有意思。 然后这个老头，我记得他给我掰骨，当然我们就一边采访一边聊，他把我头一掰，我都觉得我头他掰掉了，整个啪，我还吓一跳。 后来想想真不敢随便让人来掰。 然后就是因为这种人的存在，所以少林是有一批他所谓的秘方。 我们不管这个秘方是真是假，不管这个传统会怎么样，他可能至少是流传在登峰一带、流传在河南乡间的大批的民间医学。 然后他说他们当时的黑膏是特别厉害的，然后还有用什么碎补和青皮等十几种中药，做了一个少林十三方。 然后用他们的药引，什么红药散、白药散，然后做出来，就是少林医局的基础。 所以他们是有东西的，不是说完全没东西。 他们武术也是这样的。 然后他们这个医学找的这个人叫士严林，严林法师。 当时我觉得很奇怪，我说这个你也不太像和尚。 他就是因为家里有海外关系，对士用心找来说，你帮我负责医学这一块。 他就因此出了家，他可能很早就跟士用心有关系，他也是士用心的土地，因为盐自备，好像都是士用心的土地。 然后这个人呢，我觉得他特别像个公司总经理。 他当时就说，因为大家都不肯来学嘛，然后就找了那些年轻的僧侣说你们谁来药局，然后来了三十多个人。 然后很多人都不坚持下来，后来他就说用那个禅修的方法要他们坚持下来学学学。 而且还跟当地的医学院合作，办了一个少林药局禅医功夫学院，然后初中生进去，然后经过三年的中专学习，他就有护校的文凭了。 其实他们是合法行医的。 我发现这是很多寺院在做的一件事，可能我们大家都忽略都不知道了。 比如说我去江西的那个曹洞宗祖庭，有办法的寺院，他申请了一个国家二类的什么学院，就办了一个当地最大的医院。 其实他们是有合法的行医资质的。 那你有那种大型医疗设备吗？X光机能拍片，他们还是偏于中医。 很多人去带着自己的片子，就是已经都有了。 当然中医骨科是有一套他的系统的，靠摸呀靠什么来做，当然他们也会结合厉害的人，是会看的。 所以我觉得他那个少林药局是试用性在走所谓现代化道路，他就很清晰，我该怎么做，我弄什么人，我怎么卖药。 他卖药也是合法的。 很久以前就有听众来问过我，咱们在节目里面提到过的这么多图书，尤其是一些比较珍贵的原版书，能不能直接通过节目来进行下单。 在经历很长时间的筹备以后，互作互有的卖书业务终于上线了。 我们在微信上建立了一个橱窗，这次在橱窗当中不仅汇集了节目当中深入探讨过的书籍，也包含了一部分我个人精选的图书，其中还有少量的签名版和稀有的原版。 完整的书目信息和购买方式，各位可以前往微信公众号忽左忽右left right，回复买书两个字即可获取。 我们这个橱窗中的书目将会不定期地更新，尽量挑选我认为比较好的版本和图书，欢迎各位常来刷新常来看看。 然后他的武术也很好玩，就我们前面说过那个皮定军去看过他们的武僧，然后我们去见到的几个教头，什么世炎傲什么这些人，他们全部都是80年代，就是少林寺那个电影刚放的时候偷偷去习武，然后就留在当地，然后慢慢地练练练练练练。 也有一些武校，因为80年代你记不记得特别多的武术杂志？ 我记得我们小时候还有各种杂志。 所以80年代初在电影上看到了少林寺，然后去到真实的少林寺。那少林寺里面已经有人可以传授武术了吗？我觉得他们就留下来出家了。 对，我见到的人有一个是十多岁少年，就那时候他就开始学了。然后还见到一个是本身在外面做武校，做不下去，他会觉得他徒弟都得想了，怎么我还不行，就去了少林寺，然后就当了教头。 而且少林寺的武术体系非常奇怪，有一套自己的体系，而且有出路。它特别像个学校。你知道有一大部分的武警部队的教头是少林寺的练传的人。 请去的吗？ 对。而且90年代的时候，深圳不是那时候发展起来，有很多人请保镖，大批是少林武校出去的。除了那种外面的农民办的武校，他们自己也有大批。他们还跟我说过，有一个最出名的两个人，一个叫誓言虎，一个叫誓言报，是深圳整个保镖系统的大老级人物，后来还做得很大，真的少林派弟子。 那你说他是80年代从你起来的，这个少林派究竟是怎么回事？我们根本没法去确定，因为他们不跟外人交手。 出于好奇心，我还问他们跟谁打了，他们会说： 我们一般都不打。 然后那边那个易经经是公开学习的，我们听着易经经，完全是金庸老师编造出来的，结果人家是真的有的。当时我们2010年去的时候，好多人在里面学易经经，还有那种得了癌症的僧侣，也在学，可能是希望戒指能够好。 这个易经经，比如说这些历史，是以前就有的吗，还是说七八十年代重新发明的东西？就跟医学那个一样，一大部分被销毁，但是我们民间乡村周围，还有大量的存在。 所以我认为是被重新恢复的，不一定是真有一步点击，但他一定有一些东西，加上这个名字，不是说全部空穴来风。我觉得如果全部空穴来风太奇怪了。 你看里面那个施阳奥就跟我说，他说： 很多人请他去比武。 他说有一次，外面开武校的俗家师弟找他，说有个外地来的师父很厉害，大家都打小空拳，结果他两次把那个人打败了。 然后也有什么山东的拳师来交手。施阳奥说对方是个山东拳师，喜欢用手顶住对方的手，双方近距离交战。 我对他说： 近距离交手容易受伤，隔开距离一把。 然后隔离一丈远，这个师父大吼一声，哼，用了施后功，然后没有动手。 施后功，对方就不站而退。 你听起来很神奇吧，但是他后面讲的时候，又真是有据可查，我就特别有意思。 就是普京去少林寺的时候，少林寺就指定这个施阎奥，在大殿为他表演心意霸，也是一套少林传统功夫。 然后在表演过程中，施阎奥不由自主地发出了各种施吼声，明代大殿里回音震耳欲聋。表演结束的时候，普京给他去立功。 我觉得这不是他编的，这一定是当时有记载的。普京毕竟自己也练柔道，施阎奥很开心地说： 普京练的是俄罗斯的传统摔销，也练柔道，他不会喜欢那种花哨的表演套路，他反而会觉得你是有传统功夫的。 所以你说他们这传统功夫到底是有，还是彻底没有？ 我觉得这真的是个谜。 我觉得会是个谜。 你要说纯粹是编造的，那我觉得也不尽然。 你要说完全是从传统一步步学下来的，那都是看金融小说来的套路，我觉得那个也是假的。 而且当时少林寺要担任一项非常重要的外交工作，就是武僧团表演。因为自从少林寺出名之后，他们要去世界各国表演。 我们去的时候，他们武僧团好像分了四支，那时候有一支正好在上海师伯会，我记得特别清楚，然后有一支在非洲。 然后他们还跟我们说，就是他们表演的套路不一样的： 在非洲的，可能就要展示力量型的武功； 在师伯会的，可能就展现一些套路型的武功。 他们还是非常迎合各地需求，制造了各种武术的套路。 对，但是释永信给他们解释的也很好玩。 释永信会说： 这个传统的少林功夫，要讲产武合一，你不能光练那种好看的，一定要修心。 这是释永信给他们讲的。 所以我们看到的时候，他们经常有那种打坐、静坐。 我们现在看那个NBA，好多球员是毛尼尔他们去学的，我觉得大家是不是都是首先被打坐静坐这一招给震住了。 你得先静坐两年，你再开始练。 所以大家都默默地觉得： 哦，好高深，好什么。 就挺有意思的。 释永信一直会觉得产武这个很重要，练到最后，都是靠修心。 这个倒是，我记得以前是赵普初还是谁，就点评过一句： 说少林称第一，是禅不是武。 他好像有个特定的背景，就是80年代。 就是说你的禅其实更重要。 对，因为80、90年代，少林寺龙蛇混杂，很多社会上的人都去投奔它。 当年什么王宝强什么的年轻人，也往那跑，又出了个名人王宝强。 出现了很多的一些意外伤害事件，就在少林寺里面投奔少林寺这些学武的弟子，什么互相谁把谁捅伤了，还有什么泼油要烧这个大殿的。 所以那会儿，可能有点乌烟瘴气。 如果有谁能把那首拍下来，多么精彩，是吧？ 就是他们说90年代的时候，那首就开始分了，就是武功。 他们觉得就有几类： 一类是电影里表演的那种，就基本上挂着绳索，就是那种吊威亚的； 一类是公安机关教授的那种很实战的； 但是少林寺的是最传统的，他们会强调说他们的动作是有一定的杀鸡的。 我开始还听着觉得： 哇，好诡异的一个词，要有杀鸡。 然后他们解释，他们的杀鸡其实就是内在的爆发力和速度感，他们认为这是通过禅修才能得到的，而不是光练武。 然后他们也解释，80年代的时候，他们刚刚走江湖的时候，还表演过撞石头、对刀枪之类的，那气功那一流的。 我觉得他们也是与时俱进。 然后后来他们也觉得不对，也说少林功夫肯定不是撞石板的，那个太跑江湖。 胸口碎大石。 对，我们去的时候，就是2010年代的时候，大家已经都学会了一套说法，就是： 大象无形 大因牺牲 所以你们看不出来，但我们是有的。 释永信自己会武术吗？问过，但是不说这种不语。 但是我也看过他练武的照片，白国pose。 对，还有很多，就是他没那么胖的时候，年轻的时候，不止一张，大家如果在网上搜都可以搜到。 然后他们这个武僧团可能就不断地给不同的摆引方式，然后慢慢地就从民间转向了庙堂，就是我们说的，给世博会、给海外，然后这种东西也找人编排过。 但是他们突出的还是那种最后那一拳的力量。 我看到后来，他在非洲也开了非洲少林寺，还有什么澳大利亚的少林。 可能恰恰就是因为他们的武术的名声，当他出名之后，他就被国家要求必须得代表国家出外宣传。 因为我们知道释永信在英国的时候，是被英国女王请去吃饭的，中国的僧侣界都知道这个事实。 就是普京的孙女是正式地拜过释永信为师的，就是那种磕头拜师的那种模式。 所以他这方面的能量和人脉确实是挺大的。 通过武术变成了世界高僧，所以也承担了这部分中国软实力输出的、官方赋予的使命。 因为在很长时间内，我们的输出就是一文一武： 文就是孔子学院； 武就是少林的产武。 但是孔子学院大家都知道好像有各种问题，然后少林产武现在其实仍然是非常主要的一个输出方式。 我记得他们跟我们说，他们基本上是不盈利的，是不会的，有文化部的补贴，河南政府的补贴，还是缺几百万元。 但是他们会觉得特别有荣誉感，比如说他们去俄罗斯表演的时候，是普京自己带着他们参观克里姆林宫，还请他们吃饭的。 那也确实是一种非常高的礼遇了。 所以他们这个是少林寺的武功与外界所接触的唯一的一个方式。 然后这个方式给少林寺带了非常多的名气和资源。 然后有很多人是因为这个又来到了少林寺出家。我记得当时见到了一个释永信的侍者，一个二十多岁的人，就是因为学习少林武功所以出了家。 我觉得你想，对于一个农家子弟来说，当你没有别的资源，你进入一个寺院学习武术，有可能受到一个大国总统的接待，这都是什么样的光荣啊，这个还真的是挺神奇的。 其实我们仔细想，少林寺成为某些人的近身之阶。 对，我感觉好像就是在那些年就出现了很多，可能从90年代以来就有了。 就是我进少林不是为了终身出家，而是我是需要在少林寺获得一技之能，比如说我练一身这种武功，不管是套路的也好，实战的也好，最后比如说我去当一个保镖，或者说我最后进入影视界。 你像那个世行语，他也是，香港演的还挺多的，他演很多，现在还很活跃，徐克那些电影，包括周星驰很多电影都有他。 所以我相信他们至少在他们的训练方式上有一套自己的系统。 对，因为我有一次在上海，我还踩过世行语，我当时在杨浦的一个片场，在一个房车里面跟他聊天，他跟我讲了很多。 我很好奇，因为那段时间我也在打拳，我就问他： 你每天怎么度过？ 他说他每天早上练什么，下午练什么，晚上再去怎么样。 我说你这个怎么都是现代搏击，你不是少林寺出来的吗？ 他说那个只是过去的一个来路，现在可能更多的还是各种现代搏击的科学训练。 其实就是，我相信就我们前面说的那个药局负责人，他可能就是因为生意而出了假，当然是用信邀请他出了假，但是他也是为了生意把这块生意做大。 有人是因为武术而出了假，有人是为了很多各种各样的原因而出了假。 就少林寺在凡间存在，他还是按照一个世间的逻辑在运转。 我觉得我们中国人对寺院老有一些特别美好的幻想，就是寺院的僧侣是与世隔绝的，大家都在里面默默念经，然后替你祈福，要么就是那种特别黑暗的想象，要么就是特别高尚的想象。 宏联寺其实在世间，你肯定要行世间法，你怎么可能隔离。 我这些年接触的各个寺院都有生意，都有跟大公司的广泛连接。 我在广东见到一个寺院特别有意思，就是那些有些研究生在里面做义工，我觉得非常奇怪，我说你们怎么毕业来这里做义工？ 他们做几年有可能就因为有跟他们特别好的知名上市公司，可以把他们推荐到那直接就去工作。 因为那个公司的董事长是这个寺院的信徒，甚至这个董事长的很多中层高层的任命，都来找这个寺院的方丈来看，说： 你帮我看一下，这个人到底靠不靠谱。 当然后来当我看到这个董事长也出事，说我想哇都没看出来，这个也挺好玩。 但就可见这个寺院和世俗经济的非常密的勾连。 当然他们还有一部分会做一些真的寺院独坐的慈善。 因为少林寺有他们所谓的班首，一个老和尚叫永钱，我记得挺深刻的。 其实我对这个人印象很好的，他说他是60年代出家的。 我现在对他们的这种叙述都有点存疑，因为60年代少林寺还能接受出家。 但是这个人，他在少林寺里负责孤儿，他跟我们一边说话，一边就非常焦虑，因为他说有个小孩是有哮喘被扔的，九个月他得回去给他喂奶。 他们还是要照顾大批的孤儿，因为宗教团体是有这个任务的。 所以他说他60年代出家，他不识字，也什么都不会，但是他只学会了住恶魔作，众善奉行，只会种地做饭以及照顾孤儿。 他现在职位很高，是奸怨，但因为他出家的年份，所以我觉得很有意思。 因为他叫永钱，跟释永信是一辈的。 他说他们就是叫永的这批人，就是辈分很高的，大概只有十个人，他们都是少林寺的前任方丈、行政老和尚的徒弟。 因为国家宗教局有规定，这种扳手必须受戒十年以上的人才能担当。 所以虽然这个永钱既不识字，也不练武术，也不关心少林寺怎么发展，但他还是少林寺的扳手。 这可见少林寺重新恢复之后，这种老和尚没那么多。 你刚提到那个老的那个寺庙方丈、行政老和尚，是这几天大家po出来，1987年举报释永信的那位吗？ 我对那个举报信不知真假，不知来源，我也看到了，但不确定。 我相信在释永信想做寺院主持的过程中，一定有大量的争夺。 对，因为你一定不止你一个人想做，那凭什么你做？ 我觉得他一定有各种各样的争夺和利益，所以不断会有人举报他。 我觉得这件事一点都不奇怪，一个单位负责人怎么可能没人举报呢？ 对。你刚刚竟然提到了那些武僧群体，尤其是言自辈、永自辈，他们后面的言自辈。 2015年的时候出现过一个誓言录，就是第一次举报。 我们去的时候并没有见到这个人。 然后据说他后面出来，也办了一个武校，据说他基本不见人。 那个事情后来是不了了之了吗？ 有公开的调查结果说这个私生女是释永信弟弟的小孩，然后说并没有贪污。 我觉得是当时释永信管理中出现的问题，比如说利益分配不平等，就导致了这样的结果。 这次大家也不提那件事，这次还是很多人提了，只不过官方没提。 我就说官方不提这件事，因为你如果真的要提，你肯定就当时是谁调查的，谁负责的，当时为什么处理结果是那个样子。 对，但是如果说到他有这些问题的话，我不相信那时候没有。 他不可能是这十年突然有了这个问题。 只能是说明这个中小街，就是都管是用信条大和尚，只能说明大和尚的能量很大。 确实是很厉害的一个人。我觉得他身上也符合我们很多对于他这种角色的，我们按一个现代企业来想象，就更容易理解。你本来放在寺院想象的，你就觉得更神秘。 你就把他当成一个河南的宗庆后，对，而且你对他的要求就更高，你就觉得你就应该这样，你就应该那样，但是他就不是那样一个人。但他毕竟是个僧人，对，那当然他是有很多问题，这个我们不用去怀疑。 但是我后来也想过，如果他是个日本的寺院，他不也可以结婚，那他不也就结了，合法生孩子，还可以传承这个寺庙。但是我觉得他是一个非常聪明的经营者，我觉得他是在把这个寺院当做一个非常好的企业在经营。 我觉得他真的特别多的卖点。我记得因为少林寺是禅宗的祖庭，大家知道达末祖师一伟渡江，就是在少林寺，在松山那里面壁的。然后所以少林寺的禅堂恢复得也很好，很多人去少林寺坐禅，因为觉得他是禅宗的祖庭。 中国当时几十万的僧侣，坐禅的人只有一两千人。然后这个负责的好像是永信了吧，然后他就在全国各地去学习怎么做禅，怎么做禅，包括西安的沃隆寺，可能是佛教界地位特别高的一个。 这些遗轨其实本身都丢失了，他就到处找，就重新把他找回来。2004年的时候，他们就恢复了禅宗入庭，然后集体坐禅。我记得当时我一听，我都觉得这还挺痛苦的，我是不行的。 他是凌晨三点起床，上店，然后行乡坐乡，然后五点钟吃早饭。早饭的时候只能吃粥，不能吃油炸类的，他就会上火，会引起你的心火。 然后行乡就是拿着香火，围着菩萨转圈，脚步要不急不缓，走四十分钟出汗，然后又开始打坐。中饭是上午九点吃，也只能吃粥。 然后有个国家领导人去说， “哎呀你们应该吃那牛奶。” 他们还觉得不行，会饱，就不能让他饱，牛奶应该绝对不能喝了吧，对，我不知道为什么。然后就接着让他行乡坐乡打坐，然后到晚上五点吃饭，然后九点前一定睡觉。 他们是有一套非常严密的那个坐禅的规矩的。然后少林寺当时在中国佛教界影响就变得更大了，所以为什么就这次我听到宗教界在聊，大家会觉得圣语心在宗教界的名声没有那么不好，可能跟这个坐禅有很大的关系。 就是少林寺的禅堂的恢复是非常厉害的，他们当时还找了一个85岁的云居山的老和尚来传规则。云居山这个老和尚是虚云的弟子，这个很有意思。 我上次也是跟一个宗教界朋友在谈，就是文革的时候，不是大批人让他还俗吗？对，比如说这个云居山就有五个人，所谓没有还俗，他们没有还俗。 怎么回事呢？因为山顶有一个农林茶场，然后这些没还俗的和尚就在那当农工，所以这批人后来到寺院重新恢复的时候，他们的地位就非常高。 他们就算没有还俗的，没有断裂的和尚，但是这批人现在可能也都去世了，也都圆寂了，没有了。 然后就是你看人释永信多么厉害，还找着这种老和尚来给他们恢复禅堂。据说这个上林寺的禅堂非常厉害，有很多那种去坐禅的人，一到那就立刻去叩拜。 所以我觉得他很会做功夫，你不管你这个功夫是不是表面功夫，就是他至少是个非常会做功夫的人。 然后我听说就是他们或真或假的记载，因为82年的时候就是我们的宗教条例恢复了之后，就可以重新建庙了。然后他们虽然很穷，然后尚永信上位之后，还是做了挺多事的。 就据说： 不许赌博 不许饮酒 私立的僧人禁止这些行为 然后到87年的时候，我觉得他跟这个登峰县的矛盾早就存在了，他87年的时候他清退了当时有吃空想的挂名僧人，就在登峰的宗教局里是有这个档案的，说他做过一件这种事情。 然后他89年的时候，到处去建立少林寺景区，而且当时还在中央台打广告。政治日报89年报道过，说少林寺跑到中央台打广告： 中原第一名煞 少林功夫故乡 然后90年的时候，少林寺游客量突破50万，门票收入从1986年的8万涨到了1990年的240万，这都是有数据的。 但当时就有人骂他说， “你们不能搞旅游。” 然后他就说， “那不搞旅游，我们连修大雄宝殿的钱都没有。” 所以他的大部分钱，也都是他自己慢慢的去挣出来的。 然后包括现在大家骂的，他不让什么少林祖宿开拍，当时他好像还是拿了票房分成的，他还是授权了的。包括我们知道后面还拍过一个什么少林的电影，那完全是他主导的。 就是刘德华和成龙主演的一版新少林寺，对对对。那拍的就是十有三，环境不太成功，那一版就影响不大，但是那个完全是他主动的。我记得就是有很多合影，什么都可以证明他。 然后我们去吃觉得那很好吃的那个少林素饼，是2004年做的。 然后他也很荒谬，他到06年还做了一个什么少林产酒，按说也不能喝酒，水果也没有动，然后还在纽约时代广场做少林素饼广告。 所以我觉得他是一个非常成功的企业家。我前两天在，尤其他16岁就出家的时候，我觉得他特别像那个了不起的Gatsby，一代巨商。 然后2011年，我们去的时候是因为上市的这件事，但事实上就是中国类似于他这种的寺院，我觉得并不算少，你像那个庐山东陵寺，然后包括九华山，当年是金地藏的那个道场，里面还是肉身店。 我们说更出名的林隐寺，大家也会去，觉得林隐寺的门票收入，但其实好像林隐寺的门票也有一部分是被旅游部门拿走的。 中国宗教和这个现实的关联太密切了，以至于大家分不清。 而且少林寺就是说，他不必然是今天这样的一个规模和影响力，我们很多人会觉得说，因为有一部少林寺的电影，但实际上我们看前面的数据，就那个电影放了之后，可能也没有带来那么大的收益。 对，而是必须得有一个很厉害的人去经营去发展。我觉得特别好玩，就是他特别会做IP，他还真是前现代IP达人。 你发现了吧，他那时候就把IP各种都注册了，因为据说他注册的时候，通过有几百个注册都跟少林寺没关系，他就一一去告了，然后说这只能我们注册。 然后后来我还看到一个材料，不知真假，就说有个谁拍少林，人家说我们跟南少林授权，结果南少林的那个也被他注册了。 他后来反复去澄清说： “根本就没有南少林。” “对，是我注册的，你们要用都得跟我合作。” 但是他自己说他心中的伤口，就是源自于当年那个火腿肠，我不知道这个或真或假，挺逗的。 但是其实中国的寺院有一个特别隐秘的收入，我们是不知道的，而且国家也是不知道的，就是捐赠。这个你没法去查他。 我听说是现在才规定说，那个二维码上不能有私人，就只能捐赠到，比如说少林寺有个账号，但那个也就这十几年的时间。 2000年以前，中国人普遍都没什么钱，据说90年代的时候，已经开始出现大批的捐赠了。 但这部分就是财务数据，我觉得是很多人不知道的，这个是一个非常隐秘的收入。 比如说来个人突然自己扫给他，你不可能监控到他的那个地步。 然后少林寺的商业收入确实非常高，他在2020年的时候当时公布过，说全天收入1.2个亿。 我相信只是公布吧，因为来自旅游啊，来自教育啊，来自于影视啊。 然后他也会公布说，比如说我有部分用在文物修缮了，有部分用在孤儿院了。 然后到疫情的时候，他还去做了直播。 所以我觉得他一直是一个非常敢做的一个僧侣。 包括我们说的，在海外做了很多，巴黎有什么少林文化驿站，你知道非洲有那个五月学校，就很多地方都是有他的学校的。 对，但这样的僧侣，我看其实中国近代挺多的，那个台湾佛光山的星云法师，就是你觉得有可比性吗？ 挺有可比性的。 因为我这两天看资料，我觉得就是可能很多人会觉得他是一个商业上的成功者，然后他是一个政治僧侣。 但中国的僧侣，什么时候脱离过政治呢？ 我觉得中国的出名的僧侣，就但凡我们知道的僧侣，一定不会在深山老林里被大家所知。 要么就是跟政治人物混在一起的政治僧侣，要么就是那些跟文化人物混在一起的这种所谓师僧。 对，包括什么苏曼殊这些人啊。 对，因为永信的，我觉得他建立了大量的社会关系，我觉得也跟我们前些年的一些国内的宗教政策有关系。 我记得我们当时也采访过，就是国家宗教局的前局长吧，看一下啊，就是他怎么说的。 就说： “就算有一亿的宗教徒，但是我们有几千万的党员，几千万的共青团员，十亿人里应该是信什么的都有，不可能让大家都信仰共产主义嘛。” 所以当时这是前宗教局长叶小文的直接说法。 哇，是那个年代， 对对对。 “你们把这些话登出来了？” 登了，大家现在在网上都可以看到。 而且我觉得他说的很清楚，他说：中国共产党人对宗教整体是宽容的。 比如说延安时代到五十年代，都比较小心，因为毛泽东在湖南农民运动口上报告说， “菩萨在农民心中，要等农民自己去搬走，我们不能越主带跑。” 他们认为真正乱的时候是文革。 然后到82年的时候，出了关于我国社会主义时期宗教问题的基本观点和基本政策。 提到了几个，就他觉得他是长期的、然后很复杂的，短期内不可能说让大家都不信教。 林环说的很好玩，林环说他从天津掉到北京的时候，他妈天天念经。 他说不是因为他在天津工作好，群众推荐党中央用了他，而是因为他念经念的。 然后林环说： “我也不能跟他吵架，我也只能说，好，就是你念经念的。” 所以只是反当时，说了一句很有趣的话， “就不能把不同世界观的对立，看成人与人的对立。” “我为什么要跟我母亲对立？我永远爱我的母亲，他信他的教，我不信，我是无神论者，我是共产党员，但我们是母子，我要孝顺他。” 我觉得他解释的非常好，其实。 然后因为宗教发展太快，所以我们很多宗教政策滞后的90年代的时候，是宗教特别发展的时候。 所以到了后面又出台过不断条例，2024年时候，我们才出台宗教事务条例。 其实是从法律上管理这些宗教的社团，管理寺院，管理各种东西，依法管理。 这个是一个，我们到现在还在用的一个东西。 然后宗教事务管理条例到现在，我觉得也是一个重点的法规。 然后特别好玩，就是05年的时候，国家发贵还发过一个通知。 关于与宗教场所有关的游览参观点对宗教人士实行门票优惠问题的通知。 就是说，有归正的人可以不买票。 我觉得这是当年很大的一个矛盾。 就凭什么你这个钱是给庙里，或者凭什么这个钱被旅游部门独占？ 所以估计这是一个长期存在的矛盾。 然后还有一个特别好玩，我也是看资料看到的。 你比如说一个园林里有一个庙，那这个哪个部分算庙，哪个部分算园林？ 当时有一个原则，叫滴水为界。 就是那个门沿，就是那个屋沿上滴水的地方，就是你们的，里边是你们的，外面都是属于园林。 所以我国的其实是做了很多很细节的管理。 那像北海公园的白塔寺，那个已经完全是个旅游景点了。 就像我们前面说的，如果当时是归旅游部门里没有僧侣的，那雍和宫里面还有喇嘛吗？ 雍和宫有啊，雍和宫是大量的喇嘛。雍和宫是一个非常纯粹的宗教场所。 但是我们会发现就很多宗教场所，我们并不知道他的方丈是谁，我们并不知道他的方丈大和尚是谁。 对，这是为什么？ 我个人觉得是一个个性问题。 因为很多僧侣非常德高望重，大家都是从内部知道他的，而外部不知道他。 但是永信是内外都很知道，我觉得这个很有趣。 就是上次我们俩还聊过，也是我被骂过，就叫“寻僧记”。 对，就是我写了各个寺院的各个方丈，里面有很多名头很大的人，比如辅导协会的副主席啊，或者哪个省的辅导协会主席。 大家建的寺院都很豪华，大家都有各种各样的，比如说做医院的方式，做孤儿院的方式，也有很多跟企业家沟通的方式。 但是大家不知道那些方丈是谁。 所以也可能有些人会注重韬光养晦，有些人会注重说我个体的名声只在宗教界。 但永信始终是一个不管不顾的人。 但是我觉得也不是孤立。 民国时候，我们有一个特别出名的僧侣，叫太虚。 大家看鲁迅，看胡适，当时都拼命地跟他吃饭。 你只要看他日记里面，就经常有大量的记载。 然后最近台湾的有一部分蒋介石的档案，叫泰西岛，里面大部分是蒋介石的私人通信、家族照片什么。 你们就找到很多太虚跟他的通信。 因为太虚和尚就是民国时候最公认的政治和尚，大家觉得这个人就是个政治和尚。 有些人特别不喜欢他，有些人特别喜欢他。 但是事实上，他是非常重要的中国民国时候最重要的一个僧侣界人物。 他等于说带领中国的佛教界进行了现代化改革。 那个什么人间佛教这四个字好像最早就是他提出来的，就是他提的。 因为事实上我们晚清到民国的时候，中国佛教特别衰落，我觉得跟现在有点像。 各个事件都是号召大家去捐钱赎罪。 当时地狱关罪突出，就不断说你们要捐多少钱，就有点像欧洲中世纪时候那个赎罪卷的那种感觉。 所以当时寺院的整个地位特别低，就整个中国宗教其实在当时是有一个危机的。 然后太虚的出现就特别有意思。 这个太虚本来是个记者出家的，然后出家之后，他跟蒋介石有很深的关系，因为他是蒋介石浙江老乡，在僧礼界的名声很大。 蒋介石据说是想让太虚帮他改掉一些…… 性格上很毛躁的问题，所以太虚就不断地用这层关系，不断地给蒋介石写信。因为当时有很多人去争夺寺院的财产，太虚是帮助僧团维护财产。当然如果说仅仅从这个方面来做比较小，他其实当时帮中国佛教做的现代改革，就是建立人间佛教。 就是告诉你在家也可以修行，你不一定要把钱捐给私院，你只要做个好人，做个善良的人。所以一下子当时中国的佛教图就扩大了很多，就很多人会愿意去做这个。 我们现在知道那个星云是太虚的徒孙，星云也写过太虚的一些故事。我看过南怀锦对太虚的回忆特别夸张，特别好笑，而且特别能见到太虚这个人的性格。 就是他从外地去南京的时候，就有大批的人去火车站迎接他，然后太虚下火车就开始撒尿，当着有很多人，包括很多女的信众。当然佛教徒的解释说太虚已经做到了，“就是对此都无所谓”，就可见他的修行，然后我觉得这个也是一种美化的说法。 你知道有本书，我不知道你做过没有，《银元时代的生活史》。我知道，陈存仁的记载里也记载过太虚，记载也特别有意思。 就是太虚看了陈存仁那些东西，就请他吃饭。然后陈存仁当时正好投资药厂，就说： 我帮你介绍一个人 然后帮他拉来一个朝鲜人，然后这个陈存仁就跟这个朝鲜人还没有合作。反正就有些乱七八糟的事，然后这个朝鲜人就被他的朝鲜同党给暗杀了，因为他同时又是个什么日本特务。有此可见太虚当时的社会交往是多么的复杂。 然后他跟讲的史有交往，跟林森有交往，跟各种名人的交往，可能更像一种手段。 就是我觉得他做了很多事情，都是帮助佛教复兴，以至于星云到万年的时候，就一直在说： “哎呀，这个真是中国历史上的高僧。” 我觉得他可能第一是去世比较早，1947年的时候他就去世了。第二，因为他跟蒋介石、跟林森这种交往，你不可能在后面会得到大规模的宣传吗，可能大家就遗忘了他。 但是我看过一些西方的研究中国佛教的论文，也会说太虚是帮助中国佛教现代化的一个很重要的高僧，甚至有人会把他跟历史上的那个著名的高僧并列。 他很大的功劳在于说，他当时派了大量的汉地僧侣去藏地，然后在重庆那边建了汉藏佛学院，然后他不断地派各地僧侣去各地弘法。 大家说如果不是太虚，中国的佛教可能就变成了日本的佛教。 我们知道日本佛教到现在基本上，僧侣在寺院里只管死亡或者结婚的事情，就管东西很少了。他们的主要盈利模式就是做法会，他只管死人的事情，就变得特别狭窄。 当然他们有他们的历史原因，比如明治维新的时候是允许他们结婚，也许他们结成财产，但是另一方面，他也把他身上的光环全部都抹掉了。 你就是一个世间人，其实就是一个处理葬礼业务的。 对，其实挺惨的。 所以我们现在去看太虚，就是太虚最大功劳是替中国佛教扩大了影响，让它进入了现代化领域，不仅仅是那种敛财机构，而进入了一个帮助世间人解决问题的阶段。 你看大家会知道，台湾有很多所谓的高僧大德，大家不都是觉得他们有这个功能吗？ 太虚还有一个很传奇的事情，就是他钱不过手，所有人给他钱，他从来不知道是多少，不过他的手就放在那儿，用的时候都拿上用，比如说救济孤儿。 他在抗战的时候组织了很多那种抗战的僧团上前线，他是这么一个人。 所以我觉得政治不是问题，社会交往也不是问题，经济上的扩大也未必是问题。 真正是在于说你对佛教做了多少，你做的东西越界不越界，你是不是个延迟戒律的僧人。 好像释永信并没有后面我们说的这些功劳，至少我们没有听说过，比如说他对中国佛教阶段推动，我就没有听说过。 当然他可能有孤儿院，但是你这个东西你真正去细看，你其实如果说他做足够好事应该很早，我们大家都知道的。 对。 所以我觉得很有意思，就像我们俩前面说的，中国的很多僧侣，大家对他们是有幻想的，要么就把他们想得特别好，要么就把他们想得特别坏。 其实真正的有行动力的人，他可能跟世间保持密切关系，但是与此同时，他对佛教界是有功劳的，是有推动的。 对。 其实两岸都有这样的人物，刚刚提到那个星云法师，他不仅是台湾的佛教领袖，他还是国民党的中常委，对吧，是国民党党员。 除了太虚和尚，还有一个民国时候我经常听说的虚云法师。虚云后来就到我们前面说过的云居山，原记的，大家还都知道好多，包括红一。 我觉得其实民国时候是有很多高僧的，但是我也觉得他们跟世间保持着大量的交往。 你看红一办学校，那凤子凯那些不都是他的徒子徒孙吗？ 对。 南京还有一个叫什么八指头驼，去北京教育部要经费的时候脑溢血发作。对，太虚和了，也是因为在玉佛寺讲经的时候说他有两个很好的土地，刚刚去世，大概伤心过渡一下脑溢血。 就在上海玉佛寺。 对，去世。 所以我也觉得很有趣，如果他们没有去世，在民国的时候，他们会怎么样呢？也许他能做更多的事情。 对。 我们回头再看，释永信他们这批人出家，他们确实也面对着一个所谓的系统的断裂之后重新出家。 你说他一个16岁的农家子弟，他相对学习修为还是要浅一些吧，可能也没有一个很好的什么师诚来辅导他。应该就没有，包括我觉得他的师傅，大家说行政老和尚，我们现在都很尊敬或者什么样，那可能也就是普通的一个农村僧侣。 对。 包括星云那些，我看星云的回忆，他是能够接到太虚的那个系统的，他的师傅就是太虚的四大弟子之一。 对。 可能现在新出现的这批可能稍微好一点，比如说佛学院毕业啊，比如诗诚稍微好一点，但是我觉得以释永信那一批，因为他也六十多了吧，可能诗诚都会差一些，然后学术学院都会很差。 我们看到各个寺院印的那种书或者什么，也没有看到一些说特别被推崇的。 所以现在也有很多人会说，他就是因为栽在美文化上，贪欲无穷，这个从严格的角度来说也没有错吧，但这种因为贪腐或者生活作风问题倒下了人，那也有很多是很有文化的。 所以我觉得可能有没有文化在这事怎么说都行，但确实我们可以看到他的修行不够。 是的。 这个特别有意思，但另一方面，这代人我觉得他们的这种发迹的时期，那个机遇又确实特别好。 他处在一个中国的这些民间宗教重新百川入海，重新又开始八千过海各显神通的这样的时代，在一个传奇时代产生了无数传奇的人。 就在于说你能不能站得住或者立得久，特别有意思。 我看过一个日本的思想家的研究，就说中国人会认为： “你这些佛教寺院的财富，虽然你是方丈，但是并不归你所有。” 这是中国很根深蒂固的一个思想。 你只是暂时的包括人，包括皇家的财富，大家也不认为属于皇帝啊，大家也认为你只是暂时的管理者。 可能我们中国的民间思想就是这个观念已经根深蒂固，就是他不可能做到像日本的僧侣一样：“这是我的家产，这是我的家庙，所以这个钱我随便动。”他不是。 他只是这个时代被选择出来的一个临时的寺院的管理者，那这个财富凭什么就能够被你据为所有，而且你可以为所欲为呢？ 包括我们中国的僧侣不能结婚，当然你可以说我们没有经过明治维新，没有那么一套法律，但是我觉得中国本身大家的民间习俗也好，大家的宗教幻想也好，本身对僧侣是有很多要求的。 对。 所以他违反的，我觉得不是短暂的说一个法律或者一个宗教管理条例，他确实触犯了很多民间禁忌。 或者你从另一个角度上来讲的话，像这次适用性的一个垮台，他恰恰也满足了很多另一部分中国网友对于佛教的那种偏暗黑、负面的这种想象。 就是果然是这样。 对。 我们在那些小说里面，你看当年那个平江不孝生写那《江湖家传》里面“火烧红莲寺”那么精彩的桥段，这个江湖传统终于在21世纪重生了一次。 所以我觉得确实有很多人是非常喜闻乐见的。 是。 大家现在的讲述可能也都是按照这个系统来讲述的，只是说我们今天讲述可能更把它放在一个80年代时代中来讲述。 其实它更复杂更有趣。 就还是我们回到那个我们开头说的，如果有人能够做一个80年代开始的中国佛教的研究，甚至把它联系到从建国以来的这种宗教的一个衰落，到他80年代以后的一个复兴，然后把释永信这样的一个角色在中间作为一个主角。他是怎么利用河南的这样的一片土地，这样的一个遗留下来的，走到了世界舞台之上，传宗祖庭只有20来个和尚，对吧，40亩薄地，然后你怎么通过官商勾结也好，或者自己的一些聪明才智也好，在改开的大浪当中对吧，把它做成一个市值，我不知道多少亿，但变成了一个航母。 对。 宗教永远是在世间，行世间法，做世间事，大家把它放在当时的历史社会背景中去考察，是非常有意思的。 是的。 而且它一直延续到了今天，就我们今天又生活在一个，其实各种民间崇拜，甚至民间的这些宗教，不管大家承认不承认，这相比起四五十年前绝对是今天已经是相当蓬勃。 你看我们前面说的是一亿，我觉得现在可能还不止一亿，肯定不止了。 现在，我听到还听到他一个就有的佛教徒说他好话，我觉得也挺有意思，说他“吃锅边素”，我好像也听说。 就因为这个，大家认为他是个很随和的人，就是说他跟你们一起吃饭，不要求你们都是。 对。 就锅里面煮了肉食也没问题。 对。 我觉得他一定是一个复杂的，表演性强的，有智慧的，也有能力的一个，我们时代的一个不法之徒，一个安徽冒险家，安徽进入河南的冒险家。 对吧，本来大家争夺一下他的省份，现在也不用争夺了，还给安徽引上线。 好的，那我们这一期就到这。 谢谢程老师，谢谢大家，感谢各位的收听。 下期再见，拜拜。 window.tocIndex = { "index": [ { "index_sentences": "可以说是最近40年来对少林寺这个品牌，我觉得算是鞠躬至伟的，但是备受争议的一位人物——释永信方丈。", "section_level": 1, "section_title": "释永信的陨落与一个时代的终结" }, { "index_sentences": "所以我们今天要找到王凯老师好好聊一聊这个话题。为什么聊这个话题？", "section_level": 2, "section_title": "播客缘起：三联生活周刊2010年的封面报道" }, { "index_sentences": "那会是因为一个什么样的原因你们会写它，而且是用直接封面的形式来写？", "section_level": 3, "section_title": "2010年少林寺上市传闻" }, { "index_sentences": "但不可否认的是，在20世纪，少林寺变成今天成为一个产业，达到这样的一个规模的话，那确实是伴随着释永信80年代进入这个寺庙，并且一手操盘，通过各种政商关系，包括他自己的一些聪明才智，把它营销到了今天的这样的一个规模。", "section_level": 1, "section_title": "释永信与少林寺的商业化之路" }, { "index_sentences": "所以这样的一个人轰然倒塌，深陷淋遇的这样的一个结局，挺符合改革开放时代的很多故事的。", "section_level": 2, "section_title": "释永信：改革开放时代的一个缩影" }, { "index_sentences": "就像这几天大家在猜测它倒塌的原因，到底是因为那些外交层面的：你见了那些不能见的人，还是说更政治层面的，你背后的谁是谁，发生了什么变化？", "section_level": 2, "section_title": "释永信倒台原因的社会猜测" }, { "index_sentences": "好像是discovery去采访过少林寺的时候，他们不知道怎么翻译方丈，那时候就给释永信了一个翻译，然后就说怎么翻译呢，然后就把释永信翻译成了少林寺的CEO。", "section_level": 1, "section_title": "少林寺的商业模式与收入之争" }, { "index_sentences": "其实当时解释很清楚，而且我觉得他的解释是解释得通的，因为大家知道中国有一句俗话：“什么名山大川，大概都被名寺所占了。”", "section_level": 2, "section_title": "少林寺上市传闻的真相" }, { "index_sentences": "所以当这个寺院开始变成旅游景点的时候，一个最真实的问题就浮现在眼前了：这些钱该谁收？", "section_level": 3, "section_title": "门票收入归属：寺院与旅游部门的冲突" }, { "index_sentences": "其实这里面很多钱不是寺院收的，是文旅部门收的。", "section_level": 3, "section_title": "寺院收入归属权的普遍现象" }, { "index_sentences": "我看释永信那个回忆里面，就是他80年代初去到少林寺的时候，虽然很破败，但里面还是有20多个僧人在里面出家的，还种地。", "section_level": 1, "section_title": "少林寺的历史与当代发展" }, { "index_sentences": "然后我们有一个开国中将，叫皮定军，他40年代的时候在那边打游击，然后44年在那边建立了抗日武装的根据地。", "section_level": 2, "section_title": "皮定均回忆录中的少林寺（1940-1980年代）" }, { "index_sentences": "所以我们可以想象，就是少林的武装力量一直都有，而且他有大量的田地，因为在过去的漫长的时间里，中国很多寺院是有自己的地产，有自己的产业的，这个其实也不奇怪。", "section_level": 2, "section_title": "寺院的历史经济实力" }, { "index_sentences": "因为武功是一个无法证实的事情。", "section_level": 2, "section_title": "少林功夫：真伪难辨的传说" }, { "index_sentences": "就跟我们一零年去的时候发现，因为当时少林寺做了三大块：一个是练武，一个是药局，就是医药局，一个是修禅。", "section_level": 1, "section_title": "释永信的“软件升级”：武、医、禅的商业化" }, { "index_sentences": "所以一个是医药，也就是少林药局。", "section_level": 2, "section_title": "少林药局：传统医学与现代实践" }, { "index_sentences": "我去的时候遇见了一个老医生，这个老医生叫士行真。他说他19岁的时候，那时候还在寺院里，就是赵不出，老见过他，并且动员他说你去上中国佛大大学。", "section_level": 3, "section_title": "高僧释行真与少林医学传承" }, { "index_sentences": "他说武术这块，少林武术呢，有一个很大的原因，当然我们都知道，是因为那个香港不断的传播。", "section_level": 2, "section_title": "少林功夫的全球传播与演变" }, { "index_sentences": "对，少林寺那个取景是在国庆寺，在台中拍的。而且这个很有意思，我就是去年去到国庆寺的时候，当地人跟我讲这个故事，说因为少林寺首先，它确实地位很高，它是那个禅宗的祖庭。", "section_level": 3, "section_title": "电影《少林寺》的取景地与影响" }, { "index_sentences": "然后他的武术也很好玩，就我们前面说过那个皮定军去看过他们的武僧，然后我们去见到的几个教头，什么世炎傲什么这些人，他们全部都是80年代，就是少林寺那个电影刚放的时候偷偷去习武，然后就留在当地，然后慢慢地练练练练练。", "section_level": 3, "section_title": "少林寺武术教头与训练体系" }, { "index_sentences": "而且当时少林寺要担任一项非常重要的外交工作，就是武僧团表演。因为自从少林寺出名之后，他们要去世界各国表演。", "section_level": 3, "section_title": "武僧团的外交角色与国际表演" }, { "index_sentences": "可能恰恰就是因为他们的武术的名声，当他出名之后，他就被国家要求必须得代表国家出外宣传。", "section_level": 3, "section_title": "少林寺：中国软实力的输出工具" }, { "index_sentences": "我去的时候，其实我们是被他请到方丈室去接受采访。我跟我们的李宏谷李大人，我是被豪华所震撼了。", "section_level": 2, "section_title": "释永信的采访：奢华与务实" }, { "index_sentences": "因为大家都说适用性注册了很多少林寺的商标。对，但事实上最早注册都不是适用性，都是社会各界，因为大家把它当作一个公共的名声在注册了。", "section_level": 3, "section_title": "少林寺商标权的争夺战" }, { "index_sentences": "登峰政府就是他的，我们说用仇人这个字也不合适，但至少他们是一个非常紧张的商业关系，他们是有摩擦的，非常清晰。", "section_level": 3, "section_title": "与登封政府的紧张关系" }, { "index_sentences": "我觉得我们中国人对寺院老有一些特别美好的幻想，就是寺院的僧侣是与世隔绝的，大家都在里面默默念经，然后替你祈福，要么就是那种特别黑暗的想象，要么就是特别高尚的想象。", "section_level": 1, "section_title": "寺院与世俗生活的交织" }, { "index_sentences": "因为少林寺有他们所谓的班首，一个老和尚叫永钱，我记得挺深刻的。", "section_level": 2, "section_title": "僧人永乾与少林寺的孤儿救助" }, { "index_sentences": "你刚提到那个老的那个寺庙方丈、行政老和尚，是这几天大家po出来，1987年举报释永信的那位吗？", "section_level": 1, "section_title": "释永信面临的争议与挑战" }, { "index_sentences": "2015年的时候出现过一个誓言录，就是第一次举报。", "section_level": 2, "section_title": "2015年释延鲁的举报事件" }, { "index_sentences": "确实是很厉害的一个人。我觉得他身上也符合我们很多对于他这种角色的，我们按一个现代企业来想象，就更容易理解。", "section_level": 1, "section_title": "释永信：现代佛教中的复杂人物" }, { "index_sentences": "我觉得他真的特别多的卖点。我记得因为少林寺是禅宗的祖庭，大家知道达末祖师一伟渡江，就是在少林寺，在松山那里面壁的。", "section_level": 2, "section_title": "少林寺禅宗传统的复兴" }, { "index_sentences": "然后我听说就是他们或真或假的记载，因为82年的时候就是我们的宗教条例恢复了之后，就可以重新建庙了。", "section_level": 2, "section_title": "释永信的早期改革与商业手腕" }, { "index_sentences": "我觉得特别好玩，就是他特别会做IP，他还真是前现代IP达人。", "section_level": 3, "section_title": "释永信：一位早期的IP运营专家" }, { "index_sentences": "但是其实中国的寺院有一个特别隐秘的收入，我们是不知道的，而且国家也是不知道的，就是捐赠。", "section_level": 3, "section_title": "隐秘收入：寺院捐赠的灰色地带" }, { "index_sentences": "对，但这样的僧侣，我看其实中国近代挺多的，那个台湾佛光山的星云法师，就是你觉得有可比性吗？", "section_level": 1, "section_title": "释永信在“政治和尚”历史中的位置" }, { "index_sentences": "但中国的僧侣，什么时候脱离过政治呢？", "section_level": 2, "section_title": "中国僧侣与政治的密不可分" }, { "index_sentences": "我记得我们当时也采访过，就是国家宗教局的前局长吧，看一下啊，就是他怎么说的。", "section_level": 3, "section_title": "原国家宗教局局长叶小文的观点" }, { "index_sentences": "民国时候，我们有一个特别出名的僧侣，叫太虚。", "section_level": 2, "section_title": "太虚大师：中国现代佛教的先驱" }, { "index_sentences": "这个特别有意思，但另一方面，这代人我觉得他们的这种发迹的时期，那个机遇又确实特别好。", "section_level": 1, "section_title": "结语：时代造就的复杂人物" }, { "index_sentences": "我看过一个日本的思想家的研究，就说中国人会认为：“你这些佛教寺院的财富，虽然你是方丈，但是并不归你所有。”", "section_level": 2, "section_title": "中国人对寺院财富和僧侣操守的看法" }, { "index_sentences": "或者你从另一个角度上来讲的话，像这次适用性的一个垮台，他恰恰也满足了很多另一部分中国网友对于佛教的那种偏暗黑、负面的这种想象。", "section_level": 2, "section_title": "释永信倒台与公众对佛教的负面想象" }, { "index_sentences": "就还是我们回到那个我们开头说的，如果有人能够做一个80年代开始的中国佛教的研究，甚至把它联系到从建国以来的这种宗教的一个衰落，到他80年代以后的一个复兴，然后把释永信这样的一个角色在中间作为一个主角。", "section_level": 2, "section_title": "呼唤对释永信与中国现代佛教的深入研究" }, { "index_sentences": "我觉得他一定是一个复杂的，表演性强的，有智慧的，也有能力的一个，我们时代的一个不法之徒，一个安徽冒险家，安徽进入河南的冒险家。", "section_level": 2, "section_title": "释永信：一个时代的复杂冒险家" } ] }; window.faq = { "qas": [ { "answer": "Shi Yongxin's primary contribution was transforming Shaolin Temple into a large-scale industry and brand. He is considered controversial because of widespread allegations and controversies surrounding his management and personal life, leading to his recent downfall.", "index_of_source": "可以说是最近40年来对少林寺这个品牌，我觉得算是鞠躬至伟的，但是备受争议的一位人物——释永信方丈。", "question": "What was Shi Yongxin's primary contribution to Shaolin Temple, and why is he considered controversial?" }, { "answer": "A long-standing conflict existed because local tourism departments often collected most of the gate fees for temples, not the temples themselves. Shi Yongxin, as a national人大代表 (NPC representative), argued for the temple to receive a fair share, or for tickets to not be charged at all, as he believed the income should be part of the temple's revenue.", "index_of_source": "当这个寺院开始变成旅游景点的时候，一个最真实的问题就浮现在眼前了：", "question": "Why was there a long-standing conflict regarding Shaolin Temple's gate receipts, and who typically collected these fees?" }, { "answer": "The text states that Shi Yongxin himself couldn't clearly explain why he went to Shaolin, only that he came from Anhui and secretly left home. It suggests it might have been an unintentional journey (误大误壮) or influenced by persistent folk legends, as the famous movie hadn't been released yet.", "index_of_source": "他是从安徽引上线到少林寺，所以这个也很奇怪。", "question": "How did Shi Yongxin, a 16-year-old from Anhui, end up at the impoverished Shaolin Temple in the early 1980s, especially before the famous movie was widely released?" }, { "answer": "Shi Yongxin implemented what he called \"software upgrades\" for Shaolin Temple by focusing on three main areas: martial arts (练武), medicine (药局), and Zen cultivation (修禅).", "index_of_source": "因为当时少林寺做了三大块：一个是练武一个药局，就是医药局。", "question": "What \"software upgrades\" did Shi Yongxin implement for Shaolin Temple, and what were the three main areas he focused on?" }, { "answer": "The true nature and authenticity of Shaolin's traditional martial arts remain a mystery because, according to the text, they \"don't engage outsiders\" (不跟外人交手) for verification. There's a blend of traditional legends, modern training methods, and performance-based martial arts, making it difficult to ascertain their precise historical lineage or practical efficacy.", "index_of_source": "那你说他是80年代从你起来的，这个少林派究竟是怎么回事？", "question": "Despite its global reputation, why is the true nature and authenticity of Shaolin's traditional martial arts a \"mystery\"?" }, { "answer": "Beyond spiritual devotion, some individuals joined Shaolin to acquire practical skills like martial arts, aiming for careers as bodyguards or in the entertainment industry. Others, like the \"corporate general manager\" of the medicine department, joined for business opportunities, viewing the temple as a path to social mobility and even global recognition through its growing influence.", "index_of_source": "我觉得想，对于一个农家子弟来说，当你没有别的资源，你进入一个寺院学习武术，有可能受到一个大国总统的接待，这都是什么样的光荣啊，这个还真的是挺神奇的。", "question": "What were some of the non-traditional motivations for individuals to become monks at Shaolin Temple under Shi Yongxin's leadership, beyond spiritual devotion?" }, { "answer": "Besides ticket sales, a significant and \"secret\" source of income for Chinese temples that is difficult to track is direct donations from individuals. This revenue stream was particularly opaque before recent regulations requiring donations to go through official temple accounts.", "index_of_source": "但是其实中国的寺院有一个特别隐秘的收入，我们是不知道的，而且国家也是不知道的，就是捐赠。", "question": "Besides ticket sales, what is a significant and \"secret\" source of income for Chinese temples that is difficult to track?" }, { "answer": "The public often holds highly idealized or extremely negative views of monks and temples, imagining them as either completely isolated from the world or entirely corrupt. In reality, Chinese temples and monks actively engage with the secular world, operating according to \"worldly laws\" (行世间法) and often extensive business connections, differing significantly from these simplified perceptions.", "index_of_source": "我觉得我们中国人对寺院老有一些特别美好的幻想，就是寺院的僧侣是与世隔绝的，大家都在里面默默念经，然后替你祈福，要么就是那种特别黑暗的想象，要么就是特别高尚的想象。", "question": "How does the public's perception of monks and temples differ from the reality of their operations, especially in modern China?" }, { "answer": "Despite his strong business acumen and influence, Shi Yongxin often portrayed Shaolin Temple as a \"bullied object.\" He expressed frustration over the unauthorized use of the \"Shaolin Temple\" trademark by various businesses (like a famous ham sausage brand) and disputed revenue sharing with tourism departments, feeling that others were unfairly profiting from the temple's fame.", "index_of_source": "而且他特别好笑，你会觉得他当时已经很出名了，是个强者，但是呢，他很愿意展示出一个我们其实是被欺负的对象。", "question": "Despite his strong business acumen and influence, in what ways did Shi Yongxin portray Shaolin Temple as a \"bullied\" entity?" }, { "answer": "Shi Yongxin demonstrated a talent for \"IP\" management by meticulously registering hundreds of trademarks related to Shaolin Temple, even claiming rights over historical entities like \"South Shaolin.\" He actively pursued legal action against unauthorized use of the brand, viewing it as his exclusive property, effectively acting as an early \"IP master.\"", "index_of_source": "我觉得特别好玩，就是他特别会做IP，他还真是前现代IP达人。", "question": "How did Shi Yongxin demonstrate a talent for \"IP\" management, long before the modern concept was widespread?" } ] };

2025/8/1
articleCard.readMore

Mary, Queen of Scots: The Royal Rivals (Part 2)

Mary, Queen of Scots: The Royal Rivals (Part 2) In marked contrast to her childhood treatment in Scotland, where she was considered at first a sickly child, unlikely to live, and later a pawn in a dynastic game, even at five years old, Mary was hailed as a figure of romance in France. A brave little queen who had been forced to flee the barbaric Scots, the cruel English, for the safe arms of all-embracing France. The stage was already set in French minds for the appearance of a childish heroine. To their satisfaction, Mary Stewart, with her charm, her prettiness, and the natural docility of youth, was ideal material to be molded into the playing of this golden role. So that was Lady Antonia Fraser in her celebrated biography of Mary, Queen of Scots. And she, Tom, is describing the impact that the infant—well, not really infant because she’s five years old—that the young Scottish queen makes on the French court after she arrives in France in the summer of 1548. In our first episode, you set the scene, didn’t you, by talking about the sort of turbulent politics of Mary’s childhood, the death of her father and so on when she was only a week old. Then all this complicated political game with Henry VIII and with the French court. So just remind us what went on there. Yeah, so the Rough Wooing. Henry VIII wanted Mary, Queen of Scots, to marry his son, the future Edward VI. He thought he’d got the deal. Then it got ripped up. He got absolutely irate and sent the English to burn and slaughter and loot, which they then did for eight years. The impact of this ultimately is that Mary, Queen of Scots’ French mother, Mary of Guise, who is part of France’s leading noble dynasty, she’s able to foil the English and secure a French match for her daughter. And when I say a French match, I mean, it’s the most brilliant French match imaginable. Because when Mary Stewart, after a perilous and storm-wracked voyage from Scotland—and you were very sniffy about her failure to succumb to seasickness, but I think it’s historical proof. I wouldn’t say sniffy, I’d say bracingly sceptical. Go on, continue. Young Mary, Queen of Scots, steps onto French soil, and she does so as a future Queen of France, because her betrothed is the Dauphin himself. And this is the eldest son of Henri II, or Henry II, François—or we’ll call him Francis, because we’re being very Anglophone in our treatment of French names. I read in your notes, Tom, you’ve written—I think this is massive punching down from you—you’ve written, “He wasn’t the hunk he might have been.” And then the next line, one year younger than Mary, so he’s four. Yeah. So I think that’s a bit harsh. And you mock him as being short. I mean, all four-year-olds are short, I think it’s fair to say. Well, he’s abnormally short, but Mary is abnormally tall. So there’s a disproportion in height. He’s got a stutter, and you laugh at him about that. What do you mean I laugh at him? I’m not laughing at him, I’m just setting the historical record straight. And you also mock him, very cruel remarks in the notes about his two left feet. Yeah. He can’t dance. Mary loves dancing. He can’t dance, Mary’s already a brilliant dancer. Brilliant dancer. She’s five, and she’s a brilliant dancer, and she can steer the ship single-handedly through a storm. Unbelievable. She’s very, very precocious. But actually, Francis is great. He’s weedy, but he’s plucky. Come on, he’s four. Actually, the two of them get on tremendously well. Yeah. Because he’s not just plucky, but actually very smart. And Mary already has a kind of instinctive sense of what is expected of her, how to play to the gallery. And so she and Francis, everyone agrees, they make an adorable couple. A charming couple. Charming couple. It’s five and four. No, Tom, don’t defend it. You love it. Listen, I’ll leave a second. The king of France, he’s like you, he’s drunk the iron brew, and he’s massively, like, he thinks Mary, Queen of Scots, is brilliant, doesn’t he? The most perfect child I’ve ever seen, he says. This is a historical testimony, Dominic, which you’re sneering at, and which I’m just laying it out as it is. It is what it is. And actually, he’s so impressed by Mary that he gives orders that she should have precedence over his own daughters in royal processions. And this isn’t just because she’s the betrothed to the Dauphin, but because she’s already a queen in her own right. And clearly, you know, she’s poised, charming and regal, all in one. She’s the complete package. And also, he says that she should get nicer clothes than his own daughters. She’d turn up in Scottish fashions. Well, we’ll come to the issue of Scottish fashions in due course. I mean, I get the sense that you already hate her because of this. No, no, no, I’m objective. And you might think that, you know, the king of France’s daughters would hate her, but not a bit of it. Because the French princesses find her so charming, such fun, that they all end up the most tremendous pals. Oh, lovely. And in fact, I mean, particularly in Lady Antonia’s biography, but more generally, the accounts of Mary, Queen of Scots and her chums, you know, the French princesses and so on. And it’s like reading about the kind of the poshest, funnest boarding school yarn. It’s a kind of Renaissance Mallory Towers. Right. And it’s great fun, but there’s also quite a lot of schooling. So Mary, like Elizabeth Tudor, her cousin over in England, is given a very, very good education. I think it’s fair to say that unlike Elizabeth, she’s not an intellectual. Okay. But that’s precisely what makes her so charming and fun, because nobody likes an intellectual, do they? No, no, no one wants a blue stocking. No. So she’s given an education that would conventionally be regarded as more appropriate to princes than princesses. But that’s because she’s a queen in her own right, and so therefore she has to be prepared for her role. And as John Guy in his brilliant, I think, definitive biography of Mary, Queen of Scots puts it, what she gets is the equivalent for a prospective ruler of a degree in business administration. So she works hard, but she’s not a nerd. Her real talent, much more, I think, than for kind of construing Livy or whatever, although she does that, is for having fun. So girls just want to have fun, and Mary, Queen of Scots is the epitome of this. And it’s such an amazing background for her, because this, of course, is the golden age of the chateau in France. Anyone who’s been to the Loire and seen the kind of the beautiful chateau there, this is what you’re getting. And Mary and Francis and the royal princesses and all their friends are kind of endlessly processing from Fontainebleau outside Paris to Blois and to Chambord on the Loire. She visits them all. And Antonia Fraser can’t get enough of it. So she exclaims, “beautiful gardens, beautiful galleries, so many other beauties. Oh, how charming.” But her biography is brilliant on this. I mean, it brilliantly evokes what she describes as the kind of the dreamlike quality of Mary’s upbringing in France. She’s got ponies, of course. She’s got lap dogs. She’s got falcons. She has amazing clothes. I mean, the king of France himself has said she has to have the best clothes possible. So she has gloves of dog skin and deer skin. She has velvet shoes of every conceivable color, you know, amazing gowns and dresses and trains. And she has so many jewels that she needs three brass chests just to hold them. So, as I say, the most tremendous fun. But, of course, you know, in boarding school stories, you’re away from home. And so the issue of homesickness is always there. And so people may be wondering, well, Mary’s a young girl who’s gone to a foreign country. Is she ever homesick? And I think the answer to that is that to begin with, yes, she misses her mother very badly. And, in fact, Mary of Guise will only once be able to make it out of Scotland and come and see her in France. And the reason for that is that she is literally holding the fort back in Scotland. You know, she’s got Stirling Castle and she’s surrounded by people who would quite like to see the back of her. And the person who particularly liked to see the back of her is the Earl of Arran, who is next in line to the Scottish throne. And there’s this guy who is kind of endlessly shifting and running away from defeats and things like that. So Mary of Guise thinks he’s a bit of a loser. And she strikes in the spring of 1544. And the reason that she does then is that by this point, her daughter, Mary, Queen of Scots, is 12 years old. And this is the kind of age when it’s felt that she is able to kind of start picking up the reins of power. So Mary of Guise has prepared her tactic well. She’s got people on the side and she persuades Arran to resign the regency in exchange for kind of various bungs and benefits and bribes. Mary of Guise, this foreign woman, takes his place as the regent of Scotland. And, of course, this is yet another consolidation of French control over Scottish affairs. And even as this is going on back in France, the Queen of Scots is becoming ever more of a French princess. Now, you say that, Tom, but am I not right in saying that she does dress sometimes in Scottish national dress to amuse the locals? So that’s nice, isn’t it? Yes. And this is defined in France as being basically animal skins, which is not Scottish national dress, but this is what they think. So it’s a kind of mockery. Right. And she’s complicit in it or they are dressing her up against her will or how does it work? A bit of both, I suspect. And I think it reflects the way in which Mary’s links to Scotland are starting to fade as she becomes, you know, enters her teenage years. So she had travelled to France with a substantial Scottish train. These are all the people who were seasick while she was being poised and regal and not being seasick on the journey. So most of her Scottish train has kind of been sent away. So the Scottish attendants who have come with her when Mary’s French grandmother, her Guise grandmother, Antoinette de Guise, comes across them. She’s appalled and she says, “we can’t have them.” And her exact phrase is “they are not as clean as they might be.” So essentially she’s offended by Scottish standards of hygiene. Mary has a governess who herself is an illegitimate daughter of James the Fourth and she’s called Lady Fleming. She’s stunningly gorgeous. So John Guy describes her as beautiful and voluptuous. And she was known by the French as la belle écossaise, so the beautiful Scotswoman. She has a fling with Henry the Second, the French king. And this in itself wouldn’t have been particularly a problem except that she ends up pregnant and thereby precipitates a scandal. And so she gets dismissed and sent back to Scotland in disgrace. However, it is not a completely Scot-free zone because Lady Fleming’s daughter does remain with Mary in France. And she is one of four Scottish girls, all of Mary’s age, all of noble birth, all called Mary, who constitute, I guess, I mean, we’ve now moved on from boarding school stories. These are now Mary’s squad. This is like kind of gaggle of teenagers in Beverly Hills or whatever. So you’ve got: Mary Fleming: daughter of Lady Fleming, Mary Stewart’s cousin, the highest ranking of all the four Marys. So you might call her Posh Mary. Mary Livingstone: loves hunting, loves dancing, loves archery. So you might call her Sporty Mary. Mary Beaton: the most beautiful. So you might call her Pretty Mary. Mary Seaton: knows all about clothes, brilliant at hairdressing, can tease any head of hair into an absolute miracle. So you might call her Fashion Mary. And the five of them are completely inseparable. They get up to all kinds of scrapes and japes. They’re actually a bit like Peter the Great and Marie Antoinette, that they love playing at being common people. Okay. So to quote John Guy, “Mary adored making Cotignac, a type of French marmalade, putting on an apron and boiling quinces and sugar with powder of violets in a saucepan for hours before laying out the slices of crystallized fruits.” And she gets the four Marys to help her. They have a special kitchen that is kind of created in their apartments. They put on aprons and pretend to be kind of servants or bourgeois women organizing, you know, the housework and the house routines and things like that. It’s all such fun. Right. All very charming. And it does enable Mary, Queen of Scots, to keep up with her native language because she and the four Marys are all chattering away in Scots, even though all of them by this point speak absolutely fluent French. Probably it’s their, you know, it’s their language of choice. But by the time she’s in her mid-teens, she can probably barely remember any of it. She left when she was five, so she can probably remember nothing of Scotland or virtually nothing. Yeah. And presumably that’s what the French want. They want her to become, you know, she’s the wife of the Dauphin. She’s going to become the Queen of France. They effectively want her to be French, no? Well, and specifically the Guise family. So her relations on her mother’s side, because she is absolutely a key part of the Guise plans, not just for France, but for Europe more generally. So while Mary’s been having this lovely time with her ponies and making marmalade and so on, her two uncles, we mentioned them before, Francis, who’s now the Duke of Guise, and Charles, who’s the Cardinal of Lorraine, they’ve been busy consolidating their grip on France. So the Duke, his nickname is Scarface, a battle injury. He’s very charming, but he’s also very hard-nosed, the man of action, veteran of the Italian wars, as we mentioned in the previous episode, and a man who in January 1558 shames himself by capturing Calais from the English. Robbing them of their rightful possession on the French mainland. So that’s very sad. And the Cardinal, who is, as we mentioned, the kind of brilliant and fertile schemer, the leading political figure in the French court. The pair of them remind me of the Duke and the Cardinal in the Duchess of Malfi, the Jacobean drama. There’s a kind of quality, I think, about them that conveys everything that Protestants tend to find sinister about Catholic powers and courts at this time. They have a kind of Borgia vibe to them. And why they want Mary is not just to, you know, if Mary and Francis have children, then their grandchildren will be kings of France, presumably. But their ambitions are higher than that. They want to forge a kind of Franco-British empire that will be under their thumb. And so step by step, they’ve been putting the building blocks in place. They have Mary, Queen of Scots, in their grasp. Her mother, their sister, Mary of Guise, is ruling Scotland as regent. On the 24th of April, 1558, their niece, Mary, Queen of Scots, marries the Dauphin. Mary writes to her mother back in Scotland on the morning of her marriage. She says, “All I can tell you is that I count myself one of the happiest women in the world.” And of course, she’s an absolute wow at the wedding. Very shockingly, very innovatively, she wears a completely novel colour at her wedding. And that is white. In France, unheard of to wear a white dress. And she does this because she knows that it will suit her complexion and her auburn hair. Great festivities, gold and silver coins tossed out into the crowds. The wedding banquet features six giant mechanical ships that kind of glide around the banqueting hall. And it’s all great fun and absolutely brilliant. And everyone has a wonderful time. Now, what does this wedding mean for Scotland and its continued independence? So objectively, publicly, nothing, it would seem. So nine days before she gets married, Mary signs and seals a promise to uphold, and I quote, “Scotland’s freedoms, liberties and privileges.” But behind the scenes, things are slightly different. So 11 days before she’d signed that pledge that Scottish independence will be maintained, Mary had been given another document that had been drawn up by her Guise uncles, which is top secret. Nobody talks about it. It’s been drawn up in very florid legal jargon, basically to dazzle Mary, who’s still only 15 years old. But when she puts her seal to it, she is effectively agreeing to a completely bombshell promise that if people in Scotland knew about it, it would cause outrage. Because what Mary says is that: If she dies without an heir, then her husband, Francis (who presumably was going to become the king of France), and All his successors, so all the future kings of France, will rule as well as the kings of Scotland. So Scotland will become part of the French kind of fiefdoms. They also write that Mary has had this wonderful education, all her business studies and stuff, her ponies, her dog skin gloves and everything. And therefore, the Scots owe France a million pieces of gold for it. What? So a very expensive boarding school education. That’s a very expensive boarding school. And effectively, it’s a blueprint for turning Scotland into a province of France. Oh, the French. God, who would have thought that they would behave like such snakes? And here’s the thing. They’re not just— that’s not even the limit of their ambitions. Because they’re also, and this is the really shocking thing, they have their eyes on a much greater prize, some would say, which is our own beloved country of England. Unbelievable. And again, it’s Mary who is the key to this plan because, and we haven’t touched on this yet, but it’s a very, very important part of Mary’s story. We’ve talked about how she’s a steward of the royal line of Scotland, but she also has the blood of the Tudors, the English dynasty, in her veins. This is because her grandfather, James IV, the guy who had died at the Battle of Flodden, had married the sister of Henry VIII, Margaret Tudor. And by 1558, so when Mary is marrying the Dauphin, this has come to take on a potentially enormous dynastic significance. So we left England under the rule of Edward VI, the young Protestant heir of Henry VIII, but he had died in 1553. And he’d been succeeded by his elder sister, Mary Tudor, who is a Catholic. But she, in the months following Mary Stewart’s wedding to the Dauphin, has fallen ill and potentially fatally ill. And so the question that is being asked in France, as in England, is, “if Mary Tudor dies, who is going to succeed her?” Now, of course, she has a younger sister, Elizabeth, but the claim of Elizabeth to the English throne is viewed in France, and particularly by the Guise, with contempt. So as the Cardinal of Lorraine, the subtle, one might almost say sinister, power player at the French court, points out, “Elizabeth is the daughter of a witch, namely Anne Boleyn, who had been put to death by Henry VIII, and her daughter, Elizabeth, had been declared a bastard by Act of Parliament in 1536.” So she seems to be ruled out of the succession, in the opinion of the Guise. And worse than that, of course, she’s a Protestant, and no cardinal is going to think that a Protestant should succeed to the throne of England. And so the Cardinal of Lorraine argues that the throne should not pass to Elizabeth, but instead to the next in line, who, of course, just happens to be his own niece, Mary Stuart. So shocking developments. OK, so that is potentially, I mean, that would be momentous. I mean, that would give the Guise family control over not one kingdom, but three kingdoms. And so when Mary, Bloody Mary, Mary Tudor dies on the 17th of November, 1558, they’re straight in there, aren’t they? They say, “Our niece, Mary, Queen of Scots, the Queen-in-waiting of France, is now the new Queen of England.” And they really mean this. I mean, this isn’t just a performative gesture, that this is a serious part of their ambitions, of their kind of geopolitical ambition, right? I think they do mean it seriously, but of course, there is a slightly performative element, because they want to treat Mary as though she is the Queen of England. And so the English coat of arms starts featuring alongside those of France and Scotland, you know, in the doorways and the crockery and the plates and things that the Dauphin is using in his palace. And when ushers clear a way for Mary Stuart, you know, going through the chateau or whatever, they are crying out, “Make way for the Queen of England,” which is obviously very heady stuff for Mary. I mean, brilliant. She’s Queen of Scots. She’s now the Queen of England. Soon she’s going to be the Queen of France. And that moment arrived very soon, because on the 10th of July, 1559, the French King, Henry II, dies of a sports injury. He’s been in a tournament and the lance goes through a gap in his helmet into his eye and he dies very soon afterwards. And two months later, on the 15th of September, 1559, Mary’s husband is crowned the King of France. And again, Mary is a massive fashion success. She wears white. She looks dazzling. All the other members of the royal family are in black. So they look very dowdy compared to her. And, you know, the Queen of Scots, who the French claim is the Queen of England, is now also the Queen of France. Wow. To quote John Guy, “Wherever the French court came to rest and whichever towns it visited, the heraldic arms of Francis and Mary were blazoned with those of England on the gates.” So she has done very well for herself. Queen of Scots, Queen of France, Queen of England. And she’s what, about 16 or something. And she’s very glamorous and all this. The only slight flaw I can see in this design is that she’s not actually the Queen of England. No. So this is a big problem for her because now the person who is the Queen of England is her cousin, Elizabeth, Elizabeth Tudor. And this is the person who, in all the films and the books and the operas and whatnot, is her arch enemy. And you can see why that would be the case, right? Because if you’re Elizabeth I and you’ve spent your life being declared a bastard and being locked up in various houses and all of this, the arrival of this new contender over the channel in France, who’s also very glamorous and whatnot. If you’re Elizabeth, you presumably absolutely hate the very mention of Mary Queen of Scots’ name. Well, I think that what you do initially is to ensure that powers across Europe do not recognise this potential rival to your throne. And this is exactly what Elizabeth and her diplomats do. So they ensure that, for instance, Mary Tudor’s husband, who was the King of Spain, Philip II, that he recognises Elizabeth as Queen. And even though Philip is Catholic and Elizabeth is Protestant, Philip II does recognise her. And the reason for that is that he’s desperate to keep England on board in his ongoing struggle with France. Yeah, of course. So power politics trumps kind of religious affiliations. This in turn means that the Pope has to recognise Elizabeth because the Pope is under Philip II’s thumb. And so he has to do what Philip says. And it’s also implicitly agreed, even by the French themselves, because on the 2nd of April, 1559, they sign a treaty with England, the Treaty of Cateau-Cambresi, which effectively is the treaty. I mean, it’s kind of complicated, but in its practical effect is that the English acknowledge that they have lost Calais. And the quo for the quid is that the French implicitly accept that Elizabeth is Queen of England. And obviously, this is terrible news for Mary, because not only has her claim to the English throne been sign-lined, but actually, the guys who have signed this treaty with the English are her own uncles, the Guise, the people who’ve been saying, “you’re the Queen of England.” Right. Because it was the Duke of Guise who had captured Calais, and so he wants to confirm his military achievement. And it’s doubly bad for Mary, because all the kind of claims to the English throne, as you said, it’s alienated Elizabeth, but it’s particularly alienated two key operators at the English court. One of these is a guy called Sir Nicholas Throckmorton, who is Elizabeth’s ambassador to France, who is a very, very committed, hot Protestant, and therefore very opposed to the idea of England having another Catholic queen. He has been very busy making copies of Mary’s offensive, heraldic arms. So all these ones with England’s English coat of arms being courted with the French and Scottish coat of arms. And he’s sending them back to London. There they are being read by Elizabeth’s most trusted servant, her chief minister, who is a man called William Cecil, who, like Throckmorton, is a very, very committed Protestant. He’s passionately loyal to Elizabeth, but he’s even more passionately loyal, I would say, to the Protestant cause. He’s also a man who’s long been interested in Scottish affairs, because he had fought with Edward Seymour, the Duke of Somerset, at the Battle of Pinky, the kind of disastrous defeat that, in a way, had precipitated Mary going to France. So these two Protestants, with the ear of Elizabeth, they loathe Mary, and they see her as a potentially mortal threat, not just to Elizabeth, but to the Protestant cause in England. They are completely committed to stopping the Catholic Mary ever succeeding to the English throne. They’re so committed to that, that actually it’s Cecil’s real ambition is not just to stop Mary becoming Queen of England, but to topple her from her Scottish throne. Mary, I think, is oblivious to this, but she has made herself an enemy who, as events will prove, is as subtle and dangerous an enemy as any queen could have. And while this has been going on, of course, there have been tumultuous events in Scotland, back in her homeland. Events that threaten not just the future of Catholicism in Scotland, but Mary’s own survival as queen. Tom, let’s find out about those after the break. This episode is brought to you by Indeed. Now, speed can be very, very important at times. A great inspiration for me is the Battle of Cape St. Vincent in 1797, when Horatio Nelson really made his name. He went out of the line to attack the Spanish. His ship got entangled with a Spanish ship. He jumped onto the Spanish ship and captured it. Then he realised they’d become entangled with a second Spanish ship. He leapt onto the next Spanish ship, capturing that one as well. A tremendous example of quick thinking, saving the day. And if you’re hiring, it’s also important to act fast. Thankfully, Indeed Sponsored Jobs can help get things done. It moves your post to the top of the page so it can stand out to candidates. So speed up your hiring with Indeed. Get your jobs more visibility with a $75 sponsored job credit at indeed.com/resthistory. That’s indeed.com/resthistory. Terms and conditions apply. Hiring, Indeed, is all you need. K-pop Demon Hunters is taking the world by storm. I don’t think you’re ready for the take. Don’t miss the global phenomenon. Everyone’s listening to it. Featuring the number one top Spotify album with hits like Golden. Boop, boop, boop, it’s our moment. Netflix and Sony Pictures Animations, K-pop Demon Hunters. Now playing only on Netflix. Rated PG. Hellman’s believes they can turn any mayo hater into a lover. So if you know someone who thinks mayo equals bland, show their taste buds the truth and challenge them to try Hellman’s flavored mayo. With irresistible flavors like Hellman’s spicy mayo, garlic aioli, chipotle, and Italian herb, they’ll be eating their words. Challenge haters to try Hellman’s flavored mayo and get them to eat their words. hellmans.com/eat-your-words. To promote a woman to bear rule, superiority, dominion, or empire above any realm, nation, or city is repugnant to nature. Now that was an archive recording of a Scot, an author of one of the most famous pamphlets ever written by a Scot. I mean, history is littered with famous pamphlets, but this is probably the most famous. And it’s the first blast of the trumpet against the monstrous regiment of women. And disappointingly, monstrous regiment there actually means the kind of unnatural rule of women. I’d always thought of it as being, you know, a military squad of women with bayonets marching towards cannon or something. Now, it’s an author you describe in your notes, Tom, as top feminist and funster, John Knox. So he’s the great sort of godfather of the Scottish Reformation. He absolutely hates Mary, Queen of Scots, doesn’t he? Yeah, he does. He does. But here’s the thing. His target in 1558, when he wrote that pamphlet, she wasn’t really his target there. His target was somebody else: Mary Tudor in England and Mary of Guise in Scotland, right? Yeah, principally Mary Tudor. Yeah, it’s not just women, it’s Marys. He’s not in favour of them at all. And he sees definitely Mary Tudor, but also by extension Mary of Guise, as kind of exercising a Catholic tyranny over the whole of Great Britain. And so that pamphlet is a kind of pain, cry of despair and frustration that Catholic rule under these women seems impregnable and that the Reformation threatens to be stillborn in the island. And yet, Dominic, you as a bluff Protestant will be thrilled to learn that only two years later, everything had been stood on its head. So by 1560, the authority of the Catholic Church in Scotland has been toppled and actually, as it turns out, decisively defeated. Protestant reformers like Knox himself, perhaps preeminently Knox himself, have emerged triumphant. And the measure of what Scotland has gone through in those two years is that Alec Ryrie, the great historian of British Protestantism, in his book, The Origins of the Scottish Reformation, he can describe those two years as not just one of the most extraordinary national transformations in European history, but he goes on to say that it’s arguably the first modern revolution. He disagrees with you, because you claim the first modern revolution took place hundreds of years earlier, and yet he’s a famous professor, Tom. How could you explain that? I think what he probably means there is the first revolution of modern Europe, which is a slightly different way of phrasing it. Is that what it is? I think that is what it is. So people may be wondering, what on earth has happened? What is this great revolution? And I guess that as good a way as any of answering that question is to look at the career up until this point of John Knox, who was probably the most famous of all Mary Stuart’s adversaries. And so he, like so many of the early Protestants, I mean, there’s such a feature of, you know, the series we did on Martin Luther. He had originally himself been a Catholic priest, and probably he was converted to Protestantism by Patrick Hamilton, who we mentioned in the previous episode was Scotland’s first Protestant martyr. He’d been burnt at St. Andrews in 1546, and the man who presided over that execution, who’d brought Hamilton to trial and then convicted him and then set him up on a stake in St. Andrews, strapped with bags of gunpowder. This was Scotland’s leading churchman, a guy called David Beaton, who was the Archbishop of St. Andrews, but was also Scotland’s last cardinal, as it will prove. And three months later, Beaton gets killed by five assassins who were out to revenge themselves on Beaton for the death of Patrick Hamilton. And they murder him in his own stronghold, the castle of St. Andrews, which is right on the coast, pretty impregnable. And they hang his body from the castle walls. It’s been stripped naked. Then they pull it down. One of the Protestants who’ve kind of gathered around to exult over his humiliation then urinates in the mouth of the corpse. They then salt his body and they throw it into the most notorious dungeon in the whole of Scotland, which is inside the castle of St. Andrews. And it’s a very deep bottle-shaped hole. So once you’ve been put in it, you can only be got out with a kind of rope. And it’s dug below the level of the sea right next to the, you know, whether you can kind of hear the distant booming of the waves through the rock. And this is, you know, a pointed humiliation of the preeminent Catholic churchmen in Scotland. And yet also a fairly standard night out in St. Andrews, I would say. Yeah, well, students at St. Andrews can take us up on that. So Knox is obviously thrilled by this. He’s one of the Protestants who goes to the castle of St. Andrews, which is now kind of a stronghold of Protestantism. He serves the garrison there as its chaplain. They’re able very easily to hold out against the Scottish authorities up until the point when Henry II sends a massive task force to Scotland to try and throw the English out. This task force has enough men that they can go and capture the castle of St. Andrews. All the garrison are taken prisoner and they are sent to the galleys, and Knox is among them. So he gets chained up on the bench. He has to pull the oars and so on. He does this for a year and a half. Finally, he is set free as a result of a personal intervention from the English Protestant king, Edward VI, who admires John Knox. So Knox is set free but can’t go back to Scotland, so he goes to England. He’s given a parish in Newcastle, which he chooses because it’s pretty close to Scotland. Protestant refugees can kind of amass there. It becomes a kind of center of resistance to Mary of Guise and her Catholic regime in Scotland. Knox thinks Edward VI is brilliant. He thinks England is brilliant. He sees it as, you know, the new Jerusalem. Brilliant. It’s absolutely wonderful. Hooray. And then, of course, Edward VI dies and is succeeded by Mary Tudor, who is Catholic, who will go on to be called by Protestants Bloody Mary because of her burning of Protestants. Knox thinks, “Ah, this isn’t a good place for me.” So he scrams and ends up in Switzerland, in Geneva, which has become a kind of godly city of reformed religion under the aegis of John Calvin, with Luther, one of the two great figures of the Protestant Reformation. If Knox had thought that England was brilliant, he thinks that Geneva is even better. He described it famously as: “the most perfect school of Christ that ever was in the earth since the days of the apostles.” Great key. But he does go back to Scotland from time to time, doesn’t he? A couple of times he goes back, yeah. Because there’s now a sort of coterie of Scottish lords who are prepared to protect him. Is that because they are Protestants themselves now? Yes. Basically, the Reformation has swept through Scotland, including the aristocratic hierarchy. In fact, the most significant of these is Mary Stuart’s own half-brother, one of the many illegitimate sons that her father, James V, had fathered. He is also named James, so James Stuart, Lord James Stuart. He’s often cast as a kind of Machiavelli. He’s certainly a very subtle man, capable of incredible duplicity, but also very sober-minded and, I think, very devoutly Protestant. All his political manoeuvring is, I think, in the cause of very strongly held Protestant convictions. Jenny Wormald, the great Scottish historian who’s very down on Mary Stuart, loves her half-brother. She thinks that Lord James Stuart is brilliant, describing him as a true Stuart, tough, able, masterful, self-interested. By implication, she’s suggesting that Mary Stuart is none of those things. He’s great pals with Cecil, the man who would become Elizabeth I’s great spymaster. This is a really important axis in Mary’s story. This man, who in due course will become the Earl of Moray—the name by which he’s better known—is a key player in the story of Mary, Queen of Scots. He is able to provide patronage to Protestants in Scotland. But it’s also because Mary of Guise, the regent, even though she’s Catholic, is not a natural fanatic like Mary Tudor. Her priority, I think, is to keep Scotland secure for her daughter, Mary, Queen of Scots, her dynasty, the Guises, and her country, France. If that means cutting a bit of slack to Protestants, then she’s happy to do that. Her commitment to preserving Scotland for the true Catholic faith is kind of fourth on the list, I would say. But then, presumably, what changes everything is the death of Mary Tudor in 1558, and the fact that Elizabeth becomes the Queen of England. Mary Tudor was not an old woman, so she could have lived for 10 or 20 years more. With her death and the accession of Elizabeth, the whole kaleidoscope takes another turn. Why does it take such a turn? Is it basically because England now, between Scotland and France, is ideologically different, and therefore there’s a new impetus towards Protestantism in Scotland? Why does Elizabeth’s succession make such a difference to the story? Having a Protestant on the English throne is obviously a massive destabilizing factor, certainly for Mary of Guise. And the reason that it’s a disaster is that it goes with the grain of trends within Scotland. Her rule, although it’s been very competent, and although she’s a very effective political operator, she is French, and this has become increasingly unpopular with large swathes of Scottish public opinion, and particularly with the Scottish nobility. And so, by the time that Elizabeth becomes Queen in England, there are nobles in Scotland who are starting to think that France is playing the role in the national demonology that the English traditionally have played. So, there is one resentful lord who describes the alliance with France, the old alliance, as “an iron hook that hath caught and killed afore now the most part of our ancestors,” which is a kind of massive rewriting of history. Part of the reason for this isn’t just dislike for the kind of heavy-handed character of the French garrisons that are providing Mary of Guise with her muscle in Scotland. It’s also that she is trying to provide Scotland with the kind of centralised rule that she, as a French aristocrat, takes for granted. She thinks this is what is required for a successful modern state. And, again, Scottish nobles feel resentful of this. They feel kind of cut off from her patronage. They feel that the traditional roads by which they can source power are being closed off to her. And then, on top of that, there’s the continued presence of their own queen, Mary Stuart, Mary of Guise’s daughter, in France, which is, I think, experienced by many Scots not as redounding to Scotland’s glory, but as a national humiliation. We talked about this when we did Marie Antoinette, the way in which often foreign queens in France are kind of hostages. You know, they’re held there in a gilded cage to ensure good behaviour from the country that the queen has come from. So a group of these guys have already formed. I mean, I’ll tell you what they love in Scotland: They love a covenant. They can’t get enough of a covenant. Covenants like tea cakes, iron brew—they love all that. They’ve pledged themselves to the Reformation. And with the arrival of Elizabeth, these sort of people, what do they call themselves? The Lords of the Congregation. Yeah. They now think, brilliant, they’ve got a Protestant monarch in England, and we can actually use this now. The time has come to kick the French out of Scotland and actually to completely turn the alliance system on its head and to have an ideologically based Protestant alliance with the old enemy, with the English. Yes. And Dominic, you might say the kindling has been laid, but the spark needs to be provided. And the spark is provided by the return to Scotland on the 2nd of May, 1559 of John Knox. And I think it’s fair to say that just as Mary of Guise is kind of unsettled by the accession of Elizabeth in England, so also, ironically, it’s John Knox, because John Knox is actually very, very pro-English. He had wanted to go to England rather than to Scotland because he thought that the English were naturally much holier. Always very grateful to Edward VI. Always kind of held a candle for England. But it’s unfortunately his blast of the trumpet against the Monstrous Regiment of Women, which he had targeted against Mary Tudor, comes out just as Mary dies and is succeeded by the Protestant Elizabeth. And Elizabeth is massively offended by it. To be fair, you can see why. I mean, of course you can. And so she bans Knox from England. And so, for de Muir, Knox ends up going to Scotland. So Knox actually, ironically, is a bit like Mary. His loyalties primarily are to a foreign country. Knox would rather be in England. Mary would rather be in France, but they both end up in Scotland. And Knox’s arrival in Scotland, as I said, has this kind of incendiary impact. In Perth, he gives a brilliantly inflammatory sermon that sets mobs roaming the city, attacking the city’s monasteries. Gives another sermon in St. Andrews, and the congregation rush out and start disassembling the local cathedral. And all across Scotland, he inspires these kind of waves of godly vandalism. Abbeys are stripped bare. People go into the orchards of friaries and chop them down and take all the apples. And my favorite detail is that lots of godly reformers go into the gardens kept by the monks and literally pull up all the flowers by hand. “These flowers are offensive unto the eyes of the Lord.” And I guess the key indicator that Knox and the lords of the congregation have the wind in their sails comes when the Earl of Arran, who, you know, we’ve been fingering throughout the story as Scotland’s supreme trimmer, a guy who’s constantly responding to the blowing of the wind, he changes sides. So initially, he’d been a supporter of Mary of Guise. Then by September 1559, he thinks, “Oh, it’s all up with her. I’m going to switch sides.” And he brings over a whole nother swathe of the Scottish nobles. So that by October, most of the lords in certainly in the Scottish lowlands—it’s different in the highlands, which, of course, is that much more remote—most of the lords in the Scottish lowlands are openly ranged against Mary of Guise, the French regent. And in fact, there are only two leading noblemen in the lowlands who stay loyal to her. One of these is Lord Seaton, who is the brother of Mary Seaton, Fashion Mary, the one who’s very into her fashion. The other is a guy called James Hepburn, who is the son of Patrick Hepburn, who we mentioned in the previous episode, the Earl of Bothwell, who was the hereditary Lord Admiral and Sheriff of Edinburgh, whom Mary of Guise had been flirting with. James Hepburn has succeeded to those titles and to the Hermitage, the grim castle in the debatable lands. He’s a very effective defender of Mary of Guise. His most spectacular stunt is to pull off an enormous gold heist, because Cecil, Elizabeth’s minister in England, who’s obviously delighted at there being a Protestant revolution in Scotland, sent a massive train of gold. James Hepburn, we’ll call him Bothwell from now on, ambushes it and makes off with all the gold. The news of this is brought to Mary Stuart in France, Mary, Queen of Scots. She’s terribly touched by this display of loyalty on the part of Bothwell. From that point on, I think it’s fair to say the Earl of Bothwell is a man who will have a place in her heart. She has a certain tendresse for Bothwell because she’s so grateful to him. She’s completely powerless in all this. I mean, I know she’s the Queen of France, but she presumably can’t do anything to help her mother, Mary of Guise. She sits there impotently in France watching events in Britain, where it just gets worse and worse from her point of view. So in February 1560, her half-brother, Lord James Stuart, negotiates a treaty in Berwick, the town on the border of England and Scotland, on behalf of the Lord of the Congregation. This is an overtly Protestant alliance. The English send a fleet that blockades Mary of Guise in Leith. Elizabeth commits England to protecting Scotland’s ancient rights and liberty. This is not cast as a kind of English invasion and English takeover. But the consequence of this is that the old alliance, the alliance between France and Scotland that had been the foundation stone of Scottish foreign policy for 300 years, is now effectively defunct. Whether of a broken heart or not, Mary of Guise gives up the ghost on the 11th of June. She dies, only 44. With her death, everything is kind of thrown up in the air because Scotland now effectively has no ruler, certainly no Catholic ruler. Then, one month later, a further calamity for Mary Stuart, who’s now lost her mother. A new treaty is signed in Edinburgh between England, the lords of the congregation, and in a completely stunning volte-face, her two uncles, the Duke of Guise and the Cardinal of Lorraine, the very people who have been pushing her to take up the rule of England throughout this. The terms are completely humiliating for Mary Stuart: - France now officially recognizes Elizabeth as Queen of England. - Mary is obliged to drop her claim to the throne of England. - All French troops are to evacuate Scotland. - Any failure by King Francis and Mary, his Queen, to ratify the treaty licenses Elizabeth in England to intervene in Scotland to uphold its terms. So France has signed, sealed, and delivered their recognition of Elizabeth. All French troops that Henry II sent, which Mary of Guise had been using as her garrison and cutting edge, have to withdraw. This is a complete triumph for the English crown and the Protestant insurgency in Scotland. I guess Mary, Queen of Scots, at this point has basically only one consolation: She’s lost control of Scotland insofar as she ever had it, which she didn’t. She’s not going to be Queen of England. But at least she’s Queen of France. She’s got this husband—who you said he’s got two left feet, he’s too short, and he’s got a stutter, which you don’t rate. He’s plucky and smart. He’s a nice guy. And thank goodness nothing has happened to him, and hopefully it won’t. Right, Tom? Right. So then Mary of Guise has died in June. Then in mid-November, Francis goes out hunting and he comes back and says, “Oh, I’m feeling a bit dizzy and I’ve got a kind of strange buzzing in my ears. I hope it’s nothing serious.” By the end of the month, he is suffering a series of violent fits. By early December, weird matter is starting to ooze out of his ears. Then it starts coming out as kind of dribble from his mouth. And obviously, this isn’t a good sign. And on the 5th of December, he dies. Crikey. I mean, you never rated him, to be fair. plucky and I said he was smart. That’s all you need. But also, the other thing you need is not to die of matter coming out of your ears. But he’s gone. And for Mary, who it said wept for a month on the news of her mother’s death, in a way, this is an even more devastating blow because she’s still only 17, but she’s now been orphaned, she’s been widowed, and now she’s no longer the Queen of France. And so her whole future, which had previously seemed so glittering, I mean, now, you know, what’s she going to do? Her uncles, these arch manipulators and Machiavelles, wanted to stay in France because what they want to do is basically to marry her off to Francis’s younger brother, Charles, who is now the new king. But Mary is smart enough already to know there is no prospect of this. And the chief problem she faces is her mother-in-law, who Dominic is a link to one of the previous series we did, which is the Medici. So Mary’s mother-in-law, the wife of Henry II, is Catherine de’ Medici. So she’s from the Dukes of Florence. And Catherine de’ Medici had always disliked Mary as a kind of tool of the guise. You know, Catherine de’ Medici is a very, very smart operator. I mean, a ruthless operator. And she does not trust Mary. She wants her cleared from the chessboard. And she makes this very clear to Mary. So just one day after Francis’s death, she sends a message to Mary saying, “Hey, give me back all the jewels. You know, they’re not yours anymore. They belong to the Queen of France. You’re no longer Queen of France.” Right, she’s not messing around. Yeah, she’s not. And so Mary knows there’s no way that Catherine de’ Medici is going to allow her to marry another of her sons. The other reason why Mary isn’t happy to go along with the schemes of her uncles is that, you know, she views them as having completely stabbed her in the back because they’ve signed this Treaty of Edinburgh. You know, they’ve deprived her of her right to the throne of England. And she feels that they had betrayed both her and her mother. And so because of that, the obvious solution is that she goes back to Scotland because there at least she is Queen of Scots. Do you know what? She should have stayed in France. She should have knuckled down, just kind of had a little palace of her own, had a quiet life, a lot of dancing, and she’d have lived to a ripe old age. This is a great mistake that she makes. Do you know who wouldn’t agree with you? It’s Jenny Wormald, the great Scottish historian who also, I think, was herself Catholic. She feels that Mary should have gone back immediately to Scotland because the sooner she gets back, the better the chance there is of reversing the recent successes of Protestantism in Scotland. And the reason for this is that although many of the leading nobility are now Protestant, the vast mass of people in Scotland are still Catholic. So very like England under Henry VIII. And her status as Queen is still unchallenged. As Jenny Wormald puts it, “There is no doubt about the strength and prestige of the House of Stuart and the profound unwillingness of the Scots to challenge it.” So it’s what we were talking about in the first episode. Mary, by virtue of being a Stuart, can demand a kind of loyalty that is instinctive to the vast number of Scots of her subjects. And so that being so, Wormald argues, I think entirely convincingly, that Mary could have played the part of a Mary Tudor in Scotland. You know, she could have gone back. She could have sought to reverse the Reformation. And Wormald is very, very down on Mary because of this. I think having myself had two girls who at one point were 17, I think it’s quite a big ask to expect a 17-year-old girl who hasn’t been to the country that she’s being asked to go back to since she was five years old, to go back and kind of reverse these seismic developments. I think it’s quite tough to be down on Mary for having done the challenge. And I am more sympathetic to Mary for essentially prevaricating. So rather than hurrying back to Scotland, she goes into mourning for Francis. So that lasts 40 days. She then goes on a kind of massive tour around France saying goodbye to all her, you know, her relatives and her pals and so on. And she is also, throughout this, engaging in kind of massive slanging matches with Nicholas Throgmorton, the English ambassador, who is trying to pressure her to sign the Treaty of Edinburgh, which she refuses point blank to do. However, she is preparing to go back to Scotland. And the key figure in her plans is her half-brother, Lord James Stuart. And the reason for that is that I think Mary instinctively, when in trouble, looks to her… Family, even though her mother had not let her down, her uncles have done. So there is a problem with Lord James Stuart, which is, of course, that he’s a Protestant and not just a Protestant, but he’s basically been the kind of the leader of the lords of the congregation. You know, but as I say, he is family. And Mary is obviously hoping that blood is thicker than water. And sure enough, between them, they arrive at a kind of an agreement compact that, when it’s announced, appalls both her Guise uncles and John Knox. The reason for that, the reason that the compact appalls the Guise is that Mary promises to recognise the Reformed Church, the Protestant Church that, you know, John Knox is kind of now the leading light of, as being Scotland’s official church. In return, she will be allowed to celebrate the Catholic Mass in her own chapel at Holyrood, which is the Royal Palace in Edinburgh. And I think there’s no real question as to who gets the better of that deal. I mean, clearly, you know, Mary is selling Catholic Scotland down the river, really. Right. But at least she’ll be able to go back as Queen, right? So, you know, it’s a big win for her. And as a Catholic Queen. So she leaves Calais on the 14th of August, 1561, and with her go her four beloved Marys, and they stand on the deck saying adieu, adieu, bidding tearful farewells to France, mourning the fact that they worry they will never see France again, which is entirely accurate. Mary will never see France again. They’re a bit worried because Mary still hasn’t signed the Treaty of Edinburgh with Elizabeth. And so they’re worried that the English may be looking out for her to try and take her hostage. And in fact, as they are going up the Northumbrian coast, one of her transport ships, which is carrying her horses, gets picked up by an English patrol ship on suspicion of piracy and taken into Newcastle. So these horses are impounded and it will take three months for them finally to be sent north to Edinburgh. Mary herself, she gets to Scotland, to Leith in record time. It only takes her five days. No seasickness, of course, because she’s so poised. They arrive off Leith and the harbour is thickly veiled by mist so that Mary, in a kind of it’s pregnant with symbolism, can’t actually make out the contours of her kingdom through this sea mist. And John Knox, you know, I mean, he’s a man ready to see it as an omen. And he wrote: “The very face of heaven at the time of her arrival did manifestly speak what comfort was brought unto this country with her, to wit sorrow, dolor, darkness and impiety.” But Dominic, was he right? Well, we love a bit of darkness and impiety on the rest is history. And I’m happy to say that next time there’ll be loads of both as we explore the melodramatic story of Mary, Queen of Scots on her return to Scotland. Now, if you cannot wait to hear that story, which is one of the most remarkable we’ve ever done, the great news is that you can hear it and the remaining three episodes of this series by joining our very own group of evangelical reformers, the Lords of the Congregation, the Rest is History Club at: therestishistory.com So what are you waiting for? And on that bombshell, goodbye. Bye-bye. window.tocIndex = { "index": [ { "index_sentences": "Thank you for listening to The Rest Is History. For weekly bonus episodes, ad-free listening, early access to series, and membership of our much-loved chat community, go to therestishistory.com and join the club.", "section_level": 1, "section_title": "Introduction and Sponsors" }, { "index_sentences": "In marked contrast to her childhood treatment in Scotland, where she was considered at first a sickly child, unlikely to live, and later a pawn in a dynastic game, even at five years old, Mary was hailed as a figure of romance in France.", "section_level": 1, "section_title": "Mary's Arrival and Impact at the French Court" }, { "index_sentences": "Yeah, so the Rough Wooing. Henry VIII wanted Mary, Queen of Scots, to marry his son, the future Edward VI.", "section_level": 2, "section_title": "The Rough Wooing and Alliance with France" }, { "index_sentences": "I read in your notes, Tom, you've written—I think this is massive punching down from you—you've written, \"He wasn't the hunk he might have been.\"", "section_level": 2, "section_title": "Mary and the Dauphin, Francis" }, { "index_sentences": "And in fact, I mean, particularly in Lady Antonia's biography, but more generally, the accounts of Mary, Queen of Scots and her chums, you know, the French princesses and so on.", "section_level": 2, "section_title": "Mary's Education and Social Circle" }, { "index_sentences": "But, of course, you know, in boarding school stories, you're away from home. And so the issue of homesickness is always there.", "section_level": 2, "section_title": "Mary of Guise's Regency in Scotland" }, { "index_sentences": "Now, you say that, Tom, but am I not right in saying that she does dress sometimes in Scottish national dress to amuse the locals?", "section_level": 2, "section_title": "Fading Scottish Links and the Four Marys" }, { "index_sentences": "Well, and specifically the Guise family.", "section_level": 1, "section_title": "The Guise Family's Grand Ambitions" }, { "index_sentences": "On the 24th of April, 1558, their niece, Mary, Queen of Scots, marries the Dauphin.", "section_level": 2, "section_title": "Mary's Marriage to the Dauphin" }, { "index_sentences": "Now, what does this wedding mean for Scotland and its continued independence?", "section_level": 2, "section_title": "Secret Plans for Scotland" }, { "index_sentences": "Because they're also, and this is the really shocking thing, they have their eyes on a much greater prize, some would say, which is our own beloved country of England.", "section_level": 2, "section_title": "The Claim to the English Throne" }, { "index_sentences": "So shocking developments. OK, so that is potentially, I mean, that would be momentous.", "section_level": 2, "section_title": "Mary Tudor's Death and Mary Stuart's Proclamation" }, { "index_sentences": "And that moment arrived very soon, because on the 10th of July, 1559, the French King, Henry II, dies of a sports injury.", "section_level": 2, "section_title": "Mary Becomes Queen of France" }, { "index_sentences": "The only slight flaw I can see in this design is that she's not actually the Queen of England.", "section_level": 2, "section_title": "Elizabeth I's Rise and Opposition" }, { "index_sentences": "And it's doubly bad for Mary, because all the kind of claims to the English throne, as you said, it's alienated Elizabeth, but it's particularly alienated two key operators at the English court.", "section_level": 2, "section_title": "Sir William Cecil and the Protestant Cause" }, { "index_sentences": "This episode is brought to you by Indeed.", "section_level": 1, "section_title": "Mid-Episode Sponsors" }, { "index_sentences": "To promote a woman to bear rule, superiority, dominion, or empire above any realm, nation, or city is repugnant to nature.", "section_level": 1, "section_title": "The Scottish Reformation and Mary's Return" }, { "index_sentences": "So people may be wondering, what on earth has happened? What is this great revolution?", "section_level": 2, "section_title": "John Knox: From Priest to Protestant Reformer" }, { "index_sentences": "But he does go back to Scotland from time to time, doesn't he?", "section_level": 2, "section_title": "Knox's Return and the Protestant Lords" }, { "index_sentences": "But then, presumably, what changes everything is the death of Mary Tudor in 1558, and the fact that Elizabeth becomes the Queen of England.", "section_level": 2, "section_title": "Elizabeth's Accession and Scottish Unrest" }, { "index_sentences": "So a group of these guys have already formed. I mean, I'll tell you what they love in Scotland: They love a covenant.", "section_level": 2, "section_title": "The Lords of the Congregation and English Alliance" }, { "index_sentences": "And I guess the key indicator that Knox and the lords of the congregation have the wind in their sails comes when the Earl of Arran, who, you know, we've been fingering throughout the story as Scotland's supreme trimmer, a guy who's constantly responding to the blowing of the wind, he changes sides.", "section_level": 2, "section_title": "Shifting Loyalties and Bothwell's Act of Faith" }, { "index_sentences": "So in February 1560, her half-brother, Lord James Stuart, negotiates a treaty in Berwick, the town on the border of England and Scotland, on behalf of the Lord of the Congregation.", "section_level": 2, "section_title": "The Treaty of Berwick and French Defeat" }, { "index_sentences": "Then, one month later, a further calamity for Mary Stuart, who's now lost her mother.", "section_level": 2, "section_title": "The Death of Francis II and Mary's Widowhood" }, { "index_sentences": "Her uncles, these arch manipulators and Machiavelles, wanted to stay in France because what they want to do is basically to marry her off to Francis’s younger brother, Charles, who is now the new king.", "section_level": 2, "section_title": "Catherine de' Medici's Influence and Mary's Options" }, { "index_sentences": "And so because of that, the obvious solution is that she goes back to Scotland because there at least she is Queen of Scots.", "section_level": 2, "section_title": "Mary's Decision to Return to Scotland" }, { "index_sentences": "And the key figure in her plans is her half-brother, Lord James Stuart.", "section_level": 2, "section_title": "The Agreement with Lord James Stuart" }, { "index_sentences": "So she leaves Calais on the 14th of August, 1561, and with her go her four beloved Marys, and they stand on the deck saying adieu, adieu, bidding tearful farewells to France, mourning the fact that they worry they will never see France again, which is entirely accurate.", "section_level": 2, "section_title": "Mary's Return to Scotland" }, { "index_sentences": "But Dominic, was he right? Well, we love a bit of darkness and impiety on the rest is history.", "section_level": 2, "section_title": "Conclusion and Next Episode" } ] }; window.faq = { "qas": [ { "answer": "Mary, Queen of Scots, was hailed as a figure of romance, a brave little queen forced to flee the barbaric Scots and cruel English for the safe arms of France. She was seen as ideal material to be molded into a childish heroine due to her charm, prettiness, and natural docility.", "index_of_source": "In marked contrast to her childhood treatment in Scotland, where she was considered at first a sickly child, unlikely to live, and later a pawn in a dynastic game, even at five years old, Mary was hailed as a figure of romance in France.", "question": "What was Mary, Queen of Scots' initial impact on the French court upon her arrival?" }, { "answer": "While Mary publicly pledged to uphold \"Scotland's freedoms, liberties and privileges\" nine days before her marriage, she secretly signed a top-secret document drafted by her Guise uncles 11 days prior. This document stipulated that if she died without an heir, her husband and his successors would rule Scotland, effectively turning it into a French fiefdom, and also claimed Scotland owed France a million pieces of gold for her education.", "index_of_source": "So nine days before she gets married, Mary signs and seals a promise to uphold, and I quote, \"Scotland's freedoms, liberties and privileges.\"", "question": "How did the Guise family's ambition for Mary contradict her public pledge regarding Scottish independence before her marriage to the Dauphin?" }, { "answer": "Philip II recognized Elizabeth as Queen because he was desperate to keep England as an ally in his ongoing struggle with France. This demonstrated that in this instance, power politics, or geopolitical strategy, took precedence over religious affiliations.", "index_of_source": "And the reason for that is that he's desperate to keep England on board in his ongoing struggle with France.", "question": "Despite Elizabeth I being a Protestant, why did King Philip II of Spain, a Catholic, recognize her as the rightful Queen of England over Mary, Queen of Scots, who was Catholic?" }, { "answer": "Mary's uncles, the Duke of Guise and the Cardinal of Lorraine, who had been vigorously promoting her claim to the English throne, signed the Treaty of Cateau-Cambresi on April 2, 1559, which implicitly accepted Elizabeth as Queen of England. Later, they were also signatories to the Treaty of Edinburgh, which officially recognized Elizabeth and obliged Mary to drop her claim to the English throne, betraying her ambitions.", "index_of_source": "And it's also implicitly agreed, even by the French themselves, because on the 2nd of April, 1559, they sign a treaty with England, the Treaty of Cateau-Cambresi, which effectively is the treaty.", "question": "How did Mary's own Guise uncles, who had initially championed her claim to the English throne, later contribute to its abandonment?" }, { "answer": "The death of Francis II, Mary's husband, was a devastating blow as it left her orphaned and widowed at just 17. Consequently, she was no longer the Queen of France, which had been the foundation of her glittering future. Her mother-in-law, Catherine de' Medici, immediately asserted her authority, demanding Mary return the royal jewels, signaling her drastically diminished status.", "index_of_source": "And on the 5th of December, he dies.", "question": "What immediate and significant impact did the death of Francis II have on Mary, Queen of Scots', future and status?" }, { "answer": "John Knox, despite his preference for England and gratitude to Edward VI, ended up in Scotland because his pamphlet \"First Blast of the Trumpet Against the Monstrous Regiment of Women,\" intended to target Catholic Mary Tudor, offended the newly crowned Protestant Queen Elizabeth I. As a result, Elizabeth banned Knox from England, compelling his return to Scotland.", "index_of_source": "But it's unfortunately his blast of the trumpet against the Monstrous Regiment of Women, which he had targeted against Mary Tudor, comes out just as Mary dies and is succeeded by the Protestant Elizabeth.", "question": "Why did John Knox, a staunch Protestant, unexpectedly find himself forced to return to Scotland instead of remaining in England, a country he held in high regard?" }, { "answer": "Mary went into 40 days of mourning for Francis and then embarked on a grand tour of France to say her goodbyes. During this period, she engaged in disputes with the English ambassador, Nicholas Throgmorton, refusing to sign the Treaty of Edinburgh. A key figure in her plans for returning to Scotland was her Protestant half-brother, Lord James Stuart, with whom she forged an agreement: the Reformed Church would be recognized as Scotland's official church, but Mary would be permitted to celebrate Catholic Mass in her private chapel at Holyrood.", "index_of_source": "So rather than hurrying back to Scotland, she goes into mourning for Francis.", "question": "How did Mary, Queen of Scots, prepare for her return to Scotland, and who was a crucial figure in these preparations despite significant ideological differences?" }, { "answer": "The Treaty of Edinburgh, signed between England, the Lords of the Congregation, and Mary's own Guise uncles, was profoundly humiliating for Mary Stuart. Its terms officially recognized Elizabeth as Queen of England, compelled Mary to drop her claim to the English throne, and demanded the withdrawal of all French troops from Scotland. It effectively meant a complete triumph for the English crown and the Protestant insurgency in Scotland.", "index_of_source": "The terms are completely humiliating for Mary Stuart:", "question": "What were the key terms of the Treaty of Edinburgh, and what devastating implications did it have for Mary, Queen of Scots?" }, { "answer": "The Rough Wooing was Henry VIII's aggressive attempt to force a marriage between his son, the future Edward VI, and Mary, Queen of Scots. When his initial deal was rejected, Henry VIII sent English forces to conduct eight years of burning, slaughter, and looting in Scotland. This prolonged conflict ultimately enabled Mary of Guise, Mary's French mother, to successfully thwart the English efforts and secure an exceptionally advantageous French marriage for her daughter with the Dauphin.", "index_of_source": "Yeah, so the Rough Wooing.", "question": "What was the significance of the \"Rough Wooing,\" and how did this conflict ultimately lead to Mary, Queen of Scots', highly advantageous French marriage?" }, { "answer": "In France, Scottish national dress was inaccurately perceived as being \"basically animal skins.\" Mary, Queen of Scots, was complicit in this perception, sometimes dressing in these outfits \"to amuse the locals.\" This practice reflected her increasingly fading connections and identity with her Scottish heritage as she assimilated into French court life.", "index_of_source": "Yes. And this is defined in France as being basically animal skins, which is not Scottish national dress, but this is what they think.", "question": "How did the French court's understanding of Scottish national dress differ from reality, and what was Mary's role in perpetuating this particular image?" } ] };

2025/7/23
articleCard.readMore

The Demise of Late-Night TV Is an Omen for American Culture

The Demise of Late-Night TV Is an Omen for American Culture So I was on vacation in Maine last week with some of my best friends. And as often happens among us geriatric millennials, the conversation turned to why everything is worse now than when it was in our teens and twenties, which I know is something that no middle-aged person has ever said about their youth. This was, I recognize, a totally original insight on our part. One of the things that we lamented as we engaged in this cliché was the decline of adult comedies. We grew up on funny movies that became a part of our vocabulary: Old School 40 Year Old Virgin Anchorman Superbad Talladega Nights The Hangover Bridesmaids I guarantee you, I have quoted each of these movies conservatively a hundred times in my life. Anchorman one-liners accounted for most of my social relations when I was a college freshman, which you should feel free to make any assumptions you wish about how popular I was when I was 18. But as many people have noticed and mourned, the adult comedy barely exists as a genre anymore. It is genuinely difficult to name five big, popular mainstream original adult comedies in the last five years. Or ten years. Or fifteen years. You can say the genre migrated to streaming TV, but it’s not as if there are so many spectacularly successful sitcoms there either. Or you could say it migrated to stand-up. What matters for my purposes here is: it’s gone. The adult comedy is gone. A comedic institution that lasted for decades just kind of died. I was reminded of that conversation last week when CBS announced that it was canceling Stephen Colbert and The Late Show. I was listening on Monday as Bill Simmons made an interesting point that struck me as obviously true about this moment. Comedians used to think of jobs like hosting a late-night show as the apex of their field. Now, nobody under 65 seems to watch the show live, and nobody under 50 seems to want the job. Younger comedians today have totally different paths to success. They don’t want to make movies, sitcoms, or fill the shoes of a late-night host. No, they want to: Make stand-up specials: Just them, a stage, and a contract with Netflix. Make video podcasts: Just them, a mic, and YouTube. Success in comedy today is more solitary than it used to be. You don’t have to join big organizations to get famous. You can just strike out on your own. I don’t want to force a trend that doesn’t exist here, but I think you can see the resonance. The demise of the adult comedy and the demise of the late-night host are, to me, two pieces of the same story. Comedians used to think that fame and fortune required joining big organizations, and now they’re finding both fame and fortune by working often alone, or alone-ish. In comedy, as in so much of our culture and our economy, the age of institutions has handed off to the age of individuals. Today’s guest is Lucas Shaw, a reporter for Bloomberg and frequent commentator on The Town Podcast. We talk about all of this: The cancellation of the Late Show The demise of late night The death of the Hollywood comedy The retreat of sitcoms on TV But what I’m really after here is something that I think I’m still struggling to put into words. Why comedy as a field has become more of a solo business, and what that says about entertainment culture and society more broadly. Lucas Shaw, welcome back to the show. I do not add that in my title for any episodes. That’s incredibly rude and gauche, and yet somehow I’m very grateful that you brought it up. So I wanted to bring you on because I’m really interested in the cancellation of the Late Show in terms of what it says about television and comedy, and really more broadly, media entertainment. And it seems to me like the interpretations for the cancellation of this show really broke down into three categories: This was political. CBS did it to genuflect to the president as they’re looking forward to prepare for this merger. It happened because Colbert isn’t very funny as the host of the Late Show. Conservatives, I think, preferred this point — that the show got woke, and this is yet another case of go woke, go broke. And then the one that I’m mostly predisposed to believe is that this isn’t about politics. It’s not about cultural comedy. It’s about economics. This was a very expensive show. It cost too much money, and CBS, heading into this merger, wanted to save money. Who or what do you think killed the Late Show? So we’ll just dispense with the second one first, because it’s worth getting it out of the way. The go woke, go broke thing makes no sense here. It was the highest rated of the late night shows. We can get into particulars around social views and all of that, but that really has nothing to do with it. Comedy is a matter of taste, but people at CBS are very proud of the show. The show gets nominated for awards all the time. Stephen Colbert is objectively a talented comedian, whether you like him or not. Anyways, I think the economic rationale is indisputable. You know, the ratings for late night have gone down pretty much every year. His show has held up better than some in part because the audience is so old. But, you know, you think about from when he started to now, it’s lost at least a third of its audience. When he started, the audience for late night was already much smaller than it was. The cost doesn’t really go down. Because, you know, he is a highly paid talent. He gets paid about $20 million a year. They have a lot of people working on it. There’s a full band, lots of writers, producers — all told about 200 people. Whether we’re going to fully accept the numbers that people are putting out there — $40 million in losses this year, $50 million losses next year — that’s what people are saying. So I have to go with it. There could be some classic Hollywood accounting in it, but there’s no way that this is a profitable show and it’s not a growing show. So it does make sense to want to cancel it for that reason. At the same time, it is almost impossible for CBS to get the benefit of the doubt right now, given everything else that has happened around their Paramount merger with Skydance: Settling with what everyone considered a frivolous lawsuit with Donald Trump Everything that is coming out around David Ellison’s meetings with the chair of the FCC and the promises they’re making There are all sorts of compromises and changes being made to satisfy the Trump administration to get this deal done. So even if politics had nothing to do with this decision, people don’t believe them when they say it. I want to present not a conspiracy, but maybe like a little half conspiracy when it comes to the politics here: Media companies are absolutely sucking up to Trump. I mean, that’s not a matter of opinion. That’s a matter of fact. Amazon buying the Melania doc Disney paying $15 million to Donald Trump to resolve the lawsuit over comments made by George Stephanopoulos The president is absolutely using the power of the office to elicit payments and promises from media firms. Sometimes the threat of extortion can be convenient for companies. Companies sometimes like to have cover for cost cutting. I remember during the Great Recession, it was practically a meme that big companies would hire McKinsey to do a strategic evaluation of, let’s say, Condé Nast. And then McKinsey comes in and they look at the company and they’re like: "Hey, we're McKinsey. We're really smart consultants. We think you should..." Cut 10% of your workforce. And the head of Condé Nast is like, McKinsey told me to.” Or during the pandemic, when people used that as an excuse to make all sorts of changes to their business, some of which were completely legitimate and some of which were manufactured. Precisely. And if Paramount wanted to make certain cuts to improve their profitability or just to make the company more ideologically culturally in line with the new owners and “Oh, those cuts just happened to appease a president who we know is litigious and thin-skinned.” Well, then you have a situation where the motivation of the cuts is, to your point, maybe 90% economic, but the appearance of the cuts, if it turns out that those are 90% political, well, that’s just fine. It’s fine. If the cuts seem political because it’s nice to have that sort of cover, that overlay, that excuse. How do you feel about my sort of half conspiracy that CBS is allowing the political crackle to exist because it’s kind of a nice way to ensure that the president and a political FCC are going to approve a forthcoming merger? It is a very fun contrarian theory. The problem with it, I guess, is in the case of the recession and the pandemic and some of these other sort of exogenous political circumstances, usually, they are convenient excuses for doing difficult things to try to soften the blow when you announce that you’re doing it. In this case, it’s sort of the opposite, where because fans of Colbert are by and large not fans of Donald Trump, if you’re using kowtowing to the press, and they’re obviously not—they’ve explicitly said it’s not a political decision. But if you are sort of covertly helping to create this perception that you’re doing it for Trump, it doesn’t help you or benefit you in any way. It just alienates the Colbert viewers. So I don’t know how they gain from that. This is an economic decision. Some media are reporting that it’s a political decision. It’s useful for CBS to have the Trump administration think that an economic decision is political because it makes it seem like “They’re really trying so very hard to get this through and appease the president.” That’s the half conspiracy. I mean, sure. I think that’s fine. They were going to get, they were going to get… I think they knew that it would be perceived that way, no matter what. And so it would probably benefit them a little bit. Important to note that the person who, one of the three Paramount CEOs, one who’s in charge of CBS, George Deeks, is the one who will stay at the new company. He certainly has a vested interest in ensuring that his new owners are happy with him and that the administration is happy with him because he’s not going anywhere. One question I had is—I was hearing Matt Bell and he reported some of the numbers here: A hundred million dollar show, which is just unbelievable. Ad revenues plunging from— Is it unbelievable? Well, okay. Tell me why it’s believable. Because a lot of people have responded negatively, like, “Oh my God, how can the show cost a hundred million dollars? It’s just a guy sitting behind a desk.” But think about it: it’s a show that’s on every night, most of the year, and it costs about as much as one 10-episode show on Netflix. It doesn’t seem that crazy to me. When you have, I guess people would be surprised to know that: There are 200 people working on it. Stephen Colbert gets paid $20 million a year. When you have the band that he has with John Batiste, like, that’s not cheap. When you have some very experienced and probably reasonably well-paid writers, that adds up. When you have to film any stunts or skits or what they have to do for makeup or to help get people in to promote their stuff, there are just a lot of costs that add up. I don’t find that number shocking. Okay. Let’s assume the number is not shocking. Let’s assume the number is absolutely real. One thing I still don’t understand about this decision is: if it’s just about finances, why couldn’t the network just cut costs to keep the show in the air? I know this question has been asked a lot, but I don’t think I’ve actually heard a satisfying answer. You’ve got a show that’s bringing in tens of millions of dollars of ad revenue a year. In 2018, it brought in $121 million. That’s a $20 million profit just seven years ago. Why not just say the show as it currently exists is economically unsustainable, but the legacy could be profitable. We’re going to make some painful cuts. And oh, by the way, that could also be seen as useful to the president because it looks like we are chopping down to size a show that is ideologically predisposed to make fun of him all the time. Why not just cut? You know, when I asked the folks at CBS that, the answer was that the losses are so significant that there was no way to cut to profit. And so why do it? That was the most satisfying answer than that. I guess I would almost flip it and say, you might have to make those costs, but the other part of it that they struggled with, and I’m sure you want to get here, is like, why can they not make more money from the show’s existence elsewhere? Right? Because the show makes most of its money from advertising—and to, I mean, maybe they give it credit for some affiliate fees and stuff like that from linear television. But as everybody knows, these late night shows are primarily consumed on the internet now. And we’re not talking about streaming on Paramount Plus. We’re talking about YouTube, we’re talking about Instagram. We’re talking about kind of fans around it. And there have to be innovative ways to make money from a property that people still like that can keep shows like this on the air. Interestingly, I don’t think that’s a mystery at all. There’s the old cliché about how you’re trading sometimes the analog dollars for the digital pennies. But as you move attention from live television to YouTube, each impression gets so much less valuable. The kind of shows that succeed on YouTube, Instagram, and TikTok are not filmed in the Ed Sullivan Theater. They don’t have 200 employees. They don’t have—maybe they have a star making $20 million, but he or she is like the only person making over $500,000 in the entire group. So this is what’s happening. What you’ve already described is that attention is moving from these enormously expensive institutions to these smaller solo proprietorships, these little podcasts, these little celebrity interview shows that are thriving on YouTube because they’re getting all these views without the burden of organizational infrastructure that 20th-century legacy shows have. This is where I want to move. You wrote in your newsletter: “Late night talk shows are dying and have been for a long time. The Late Show CBS is the first of the three major late night talk shows to call it quits. It won’t be the last.” Why do you think it won’t be the last? Because the numbers that we have seen and heard on this show are similar for all these other shows and the trend lines are the same. The audience is going down. The revenue is going down. The costs aren’t going down commensurate with those other metrics. Eventually these media companies will decide it’s not worth it. The last one to go will probably be The Tonight Show because it has the longest history. Fallon has the biggest audience on the internet. If I were to be uncharitable, I’d say of those three hosts, I think Fallon has the least clear path after that. Jimmy Kimmel’s contract is, I believe, up the same time as Colbert next year. Would I be shocked if at some point in the next couple of months we heard that he wasn’t coming back? I wouldn’t be. There’s a greater chance because he’s more woven into the fabric of Disney that they will maybe extend it for another year or two and they’ll find ways to save money. His show’s also not as expensive as Colbert’s. It’s not filmed at the Ed Sullivan Theater. It doesn’t have quite as big a staff. I just think it’s inevitable that this particular type of late night topical talk show filmed in the theater with a big writing staff and band—the whole format—is a relic of the latter 20th, early 21st century. And to your point about these kind of smaller, nimbler shops: If you asked anyone under 30 whether they spend more time watching Hot Ones or one of these late night shows, I would assume they spend more time watching Hot Ones. Or Chicken Shop Date. Or any number of these shows that take the interview format of celebrity and reinvent it for the internet. It’s not the exact same format. There are parts of that topical humor that now exist in podcast form. There are parts of that, especially the interview part, that exist in podcast form but on YouTube. It just has splintered, and there are many eras to it. And, and, and, and as you said, they’re the newer ones, the YouTube and podcast versions of these shows are far less expensive. Another quote from your newsletter, which I thought was really important is that there’s not a long list of people who want to host these shows: “Major comedians can now make more money and reach more people by touring and filming standup specials.” What I’m hearing here, and Bill Simmons made this point as well with your friend, our friend, Matt Bellany on his Monday show, is that like in the 20th century, early 21st century, like to be a comedian meant to a certain extent to aspire toward having these kinds of jobs, toward being able to fill out these kinds of shoes. And I think it also meant wanting to, you know, star in adult comedies, which is a trend line I also want to trace with you in a second. But nowadays comedians just find it much easier to make an enormous amount of money as sort of solo entrepreneurs, rather than a part of this like huge legacy organization. Like that seems like a really important piece of this as well. That like, it’s not just that Netflix and CBS can’t make talk shows work. It’s also that the kind of people who would be excellent at these kinds of talk shows, if the year were 1975, now that the year is 2025, don’t actually want to do these jobs. They want to be one man bands like Shane Gillis and do their own thing everywhere without being tied down to a big institution. Yeah. If you were, I had an interesting conversation about this with Bert Kreischer, who’s a popular comedian, a couple of years ago, maybe last year. He was saying how at the start of his career, if you’d asked him where he wanted to be, it would have been: hosting a late night show starring in your own TV sitcom making your own movie Those were sort of, those were sort of the three, the top of the pyramid for comedians. Now he can make more money going on the road and then filming his tour for a Netflix special than he could from any of those. He’d have to take a pay cut to make a sitcom. And that’s just a fundamental change in our media ecosystem that I don’t think is going back anytime soon. So you would have: John Mulaney is not hosting the equivalent of a late night show on Netflix because of how it pays him. He’s hosting it because it’s the format that he loves and he wants to try it. And Ted Sarandos, the co-CEO of Netflix, is a huge comedy nerd and is totally willing to go with it even though nobody watches it. But I, if you ask me, do I think that John Mulaney is going to be hosting that show for 20 years, like some of these people host a late night show? I’d be shocked. I want to go back to the comment that that comedian made because it so clearly makes the point that I haven’t been able to stop thinking about since the Colbert news came out. I really do see this as part of a larger story about the shift from institutions, entertainment to individuals, entertainment. That is to say, if you wanted to become a star, it used to be the case that you had to move through institutions. But now in the age of fragmented media and the internet and YouTube and algorithmic media, it’s weirdly easier to become a star many times and in many ways, not by going through institutions, but rather by building this kind of one man band with a direct relationship to your audience. And how do you, how do you define an institution in that case? I mean like a group, I mean like a group of dozens or hundreds of people, right? You just told me the late night show has 200 people supporting Stephen Colbert. That’s an institution, not only because it’s lasted for decades, but also because it employs a large group of people around a star. And the name of the show is weirdly more significant than the star itself, right? You can have different hosts of The Tonight Show or The Late Show through the decades, but the show remains the same. The institution remains the same. Like different people living in the same house. So that, so I think of institutions as being both about: groups — about large groups of people required to make products but also brands that exist that are bigger than the individual, rather than the individual being bigger than the brand Like, you know, you think about just something like a really, really obvious example: Mr. Beast doesn’t make sense to that Mr. Beast. Like, it doesn’t even make sense to say that sentence. And so that is an individual, not an institution. But so in the case, in the case of a couple of unpretested, so with Mr. Beast, so, but we’re not counting sort of YouTube as an institution. No, no, that’s a platform. Right. In the case of Hot Ones, would we count it because it was sort of birthed within First We Feast, which was part of Complex? That is an institution. It’s a small one, but it sort of came of age in a digital media institution, right? It is now since gone independent. So, yes, I mean, look, there’s going to be some blurred lines here. But, I mean, you just look at the way that these shows are accessed. People don’t search for Stephen Colbert. They tune in at a specific time to CBS in order to see a spot that CBS is paying to Stephen Colbert, right? He is almost like he’s being used to rent space within the CBS structure, rather than someone searching Mr. Beast to find their videos directly. That also seems like a distinction between institutions and individuals. Yeah, well, I would argue that, and I don’t think you disagree, but they’re sort of stuck in the middle of this, right? So, when I was curious and wanted to look up for my newsletter, just for writing in general, the social followings of the different late night hosts, I went to YouTube and did not look up the late show. I looked up Stephen Colbert. I think that they have become about the hosts and the institutions matter less and less, which is why it’s so important for their longevity and for the future of those ideas that you build around the talent. Unless you’ve created a brand that in some way outlives that, but I don’t think people watching online are now like that. SNL is different. I do think people go and look up SNL because that is the brand. There are, every once in a while, performers on it who matter, but the brand is SNL. I think for those late night shows, the brand is now the host. And so, you need those hosts and the affiliated show to be treated almost like they are a YouTube influencer. I actually, in thinking about your really good question—and this is the nice thing about having a good journalist be the guest who sometimes becomes the host—I want to go back again to the interview that you had with that comedian. Remind me of his name? Bert Kreischer. He said: “I used to want to be a talk show host, have my own sitcom or star in a comedy. Now I realize that I can just go on the road.” That is actually the clearest example of what I’m trying to describe. He’s saying, I used to think that to become a star, I had to join a group. Whether that group was an already existing talk show, an enormous team of people putting on a sitcom, or an even more enormous team putting together an adult comedy. But those aren’t the paths to success anymore. The path to success now, he’s saying, is: Me with my United Frequent Flyer account Getting hired by a bunch of people Putting me in theaters Putting me in hotel rooms Then flying back home That is a major difference in terms of interpersonal relations. What he’s saying is: I used to need to be around other people in order to become a star. Now I can make much more money essentially by myself. And that strikes me as a trend that, yes, is about economics and media, but also, there’s something deeper here. It’s about society. And that’s why I think it’s so interesting. Well, I would qualify that a little bit, only in so much as this: for those unfamiliar with his story, a big part of his success was, one, that he had a sort of a bit from a Showtime comedy special. The comedy special itself, I don’t think, did that well. But the bit went viral on the internet. The other is, he is sort of an extended part of the Joe Rogan universe. Joe Rogan has helped propel a lot of stand-up comedians and other podcasts to fame. And he is not an institution in the traditional sense. It is more of an informal network of creators that sort of propel one another forward. So, it is decentralized, but there is still, in many ways, a core to it. This episode is brought to you by Zendesk, introducing the next generation of AI agents built to deliver resolutions for everyone. With an easy setup that can be completed in minutes, not months, Zendesk AI agents resolve 30% of interactions instantly, quickly giving your customers what they need. Loved by over 10,000 companies, Zendesk AI makes service teams more efficient, businesses run better, and your customers happier. That’s the Zendesk AI effect. Find out more at Zendesk.com. This episode is brought to you by Indeed. Hiring someone new for your business can be a big move, and I understand you probably want to take your time to make sure you found the right person. But playing the waiting game could do more harm than good. Because that’s extra work and extra stress you’re putting on you and your team. It’s not a healthy work environment. When it comes to hiring the right people fast, Indeed is all you need. Their sponsored jobs move your job posts to the top of the page, letting you stand out first to relevant candidates. It makes a massive difference. According to Indeed data: Sponsored jobs have 45% more applications than non-sponsored jobs. You’re only paying for results. No monthly subscriptions or long-term contracts. There’s no need to wait any longer. Speed up your hiring right now with Indeed. Listeners of this show will get a $75 sponsored job credit to get your jobs more visibility at Indeed.com/plane. That’s Indeed.com/plane right now. Support our show by saying you heard about Indeed on this podcast. Indeed.com/plane. Terms and conditions apply. Hiring? Indeed is all you need. What does it mean to live a rich life? It means brave first leaps. Tearful goodbyes. “I love you so much.” And everything in between. With over 100 years experience navigating the ups and downs of the market and of life, your Edward Jones Financial Advisor will be there to help you move ahead with confidence. Because with all you’ve done to find your rich, we’ll do all we can to help you keep enjoying it. This is slightly a tangent, but it’s a tangent I’m really interested in. When the news that Colbert was eventually being let go by CBS broke, I had just had this long conversation with friends during a main vacation about the decline of adult comedies. We’re a bunch of men and women in our late 30s. So take this for what it is. When we grew up, this was the beginning of the Judd Apatow universe. We watched and quoted old school: Anchorman 40-Year-Old Virgin Superbad These movies were our lexicon. “Throw in Wedding Crashers, even though it wasn’t a Judd Apatow movie.” Before I force a connection here between these two developments—the Late Show cancellation and the demise of the adult comedy—I would actually really love to understand your expert opinion on why these kinds of movies aren’t made anymore. I know this is something that’s been talked about a lot, but maybe some kind of consensus has settled in the industry that explains why, if you go back over the last 5 to 15 years, it’s actually very difficult to build a Mount Rushmore of influential adult comedies that people go to each other anymore. Yeah, it’s really unfortunate. What happened? What is the movie that I’m going to watch after a long week on a Friday night when I just want to turn my brain off and laugh? You end up having to go back to the 2000s instead of the 2010s. I think the simplest answer, which is different from one related to some of what we’re talking about, is the globalization of the movie business. Studios are going for massive blockbusters that work all over the world. Comedy is generally local. Certain comedians work everywhere (big stand-ups touring globally), but by and large, comedies tend to be more localized. If you’re looking for a movie that’s going to make 400, 500, 600, 700 million worldwide, there are very few comedies. The Hangover is one of the only ones that’s ever done it. So I think that’s probably the biggest factor. Some of it might also be that: Up-and-coming comedians are coming from outside the Hollywood system. They’re getting famous on YouTube or podcasts. It’s not clear the skills that make someone really funny and successful there translate to the cinematic medium. The one that probably has the most direct connection to what we’re talking about is just the loss of common culture, which makes it harder. Comedy is such a shared experience. You want to go to a theater and laugh together. You need shared references. We just don’t have as much of that. So it’s harder to find things that bond and bind people together. I really like those explanations. I want to fold in domestic consumer behavior as well. Like, how to say this best? It just seems like the typical American buys what? Three movie tickets a year? Like three and a half? It’s around there. It seems like we collectively are reserving those tickets for blockbusters. Then we’ve almost been implicitly trained by Hollywood to reserve those tickets… For the Oppenheimers, the Barbies, and the Marvels, in a world where we’re holding onto that ticket until there’s either a big franchise event or just like a really sui generis breakout hit like Sinners, right? That is not based on any IP, but just becomes its own micro phenomenon. Outside of those things, nothing seems to crack $70 million anymore. That seems to me to speak to not just changes in, yes, everything you said — the globalization of the movie business — but also changes in American audiences. They see different movies, donate their money to different kinds of films now than they did 20 years ago. Hollywood is going to see that message and respond to it. How do you feel about that? No, I think it’s a really good point. The movie business has become an event business. It used to be a habit business where people would just go to the theater on a Friday and some wouldn’t decide what to see until they got there. Now you pretty much only go to the theater to see something in particular. As televisions at home have gotten better, and entertainment options at home have gotten more plentiful, people are only going to the theater for something that they feel like they have to see, something that merits being on the big screen, something that merits them leaving their house. There’s just too much they can watch and laugh at at home. I’m personally mystified by why horror has navigated this transition better than comedy. I would say that neither one of them is about the quality of cinematography or visual effects or anything like that—something that needs to be seen in the theater on a big screen. But the reason to go and see them in a theater is because they are both communal experiences, one out of fear and one out of kind of laughter and delight. For some reason, there is a little more of an emphasis feeling that you should go and be scared together than you should go and laugh together. I’m always confused by that. I think it helps that horror movies are cheaper than comedy movies by and large. If people could make a bunch of comedies for $5 million, maybe we’d be having a different conversation. But horror has generally held up better than comedy in theaters. That is really interesting. I never thought about that. I’m actually curious what your theories are for why horror would hold up better than comedy. My guess, as you were talking, is that horror is spectacle in a way that comedy is not. Some of my favorite comedies are kind of quiet comedies — awkward comedies. That’s simply a matter of taste. But there’s nothing spectacular about awkward comedies. The humor almost exists entirely within you rather than being shared with people laughing hysterically at crazy antics. Whereas horror demands to be expressed. It demands to be externalized. So there is something about being in an audience around other people. There’s also something I think about watching. People sometimes are afraid to watch horror alone and less afraid to watch horror with other people. So there’s something about the genre of horror that has a collective function of safety when you’re watching it with other people. Especially because horror tends to over-index a little bit with women. Maybe I’m confusing horror and true crime. But I feel like there is something about that. But you’re right that if you’re going to be scared, it’s more fun to laugh with other people. But it’s very comfortable to sit and laugh at home. It’s not as comfortable to sit and be petrified in your own home, especially if you’re alone. Well, the other thing I guess I’ve been thinking about is, because you were talking, we were talking about the kind of societal role of the theater and when and why you go to the movie theater. One of the reasons is certainly that you can entertain yourself more at home. But I find it interesting that it’s not just film comedies that are in decline. It’s also television comedy. If you look at Nielsen’s recent data for the most watched streaming shows of the first half of the year: The acquired shows, basically reruns, are filled with comedies like: Family Guy Bob’s Burgers South Park Young Sheldon Friends If you look at the original shows, there are no comedies. That’s so interesting. There’s this meme that all these Gen Z-ers are still watching Friends and The Office because nothing’s been made in the last 20 years. But there isn’t really a new sitcom — except maybe Ted Lasso. So a lot of the most popular comedic quotes today come from shows that are not strictly comedies, like The Bear, which is not really a comedy. The Max is a comedy. The White Lotus is a comedy-ish. Like there’s a lot of these things that are comedy and drama. And I, people assume that after the success of Ted Lasso, every streaming service was going to say, “we need our own Ted Lasso.” Well, where are they? Is it just that people aren’t watching the sitcoms? They can’t find the right thing? Whatever it is, there are not a bunch of new sitcoms working. And so I, I think it’s something that happened, is happening across Hollywood. Yeah. You, you triggered a couple thoughts here. One is which, one of which is that, um, you know, I had this, the last chapter of my book Hitmakers was called Empires and City-States. And I said, you know, one vision of the future of entertainment is that certain kinds of media will just get bigger and bigger and bigger. Those are the empires. And then certain kinds of media will just get smaller and smaller and smaller. And those are the city-states. And listening to you, again, and this is just coming together for me for the first time right now, movies are becoming empires and comedy is becoming a city-state. Like, movies are getting bigger and bigger. Like, that market is globalizing. And some of that is about marketing costs and some of it is about labor costs. But there are a lot of pressures moving movies to become bigger global businesses. And meanwhile, comedy in a lot of ways is just getting smaller. It’s not that there’s no comedy on the internet. That’d be ridiculous to say, but comedy movies don’t really exist the way they did 15 years ago. Comedy sitcoms don’t really exist. Popular sitcoms don’t really exist the way they did 15, 20 years ago. To the extent that comedy exists, it exists, again, at a very, like, individuated level. It’s smaller. It’s direct. It’s individuals. It’s more narrowly targeted. It’s often, like, more political because if you’re just talking to your own audience, you don’t have to worry about offending people, right? It’s your audience. You can be more conspiratorial. Again, you don’t have to worry about offending people. And that means that, like, it’s harder to have mainstream comedy if the evolution of that market is going toward individual comedians talking to individual audiences in a relatively intimate way. Or just memes. I mean, to tie it back to Colbert, right? So the biggest topic on the internet last week was the Coldplay concert where the CEO and the employee and everyone is going to know what I’m talking about. Which, by the way, can we pause there? It’s amazing that everybody knows what you’re talking about. Like, it is amazing when, as fragmented as our culture is, we have this, like, little tiny moment that comes to us from the past. That everybody knows about. That somehow, I was at a high school friend’s 40th birthday. And a friend of mine from high school who, like, lives in Brooklyn, teaches in Brooklyn, I just launched into a conversation about this Coldplay concert. I haven’t spoken to this guy for 10 years. I just assumed, well, of course, somehow this information would have got to you. So it is just funny how, like, once every, like, six months, there’s an event that just somehow cuts through the fragmentation and becomes this little tiny vestige of the mainstream. And then it dies. And we, you know, go back to fragmentation again. It’s just weird. It’s totally, and I brought it up because Colbert did something on it by the, you know, I think the next day. But by that point, it used to be that if you wanted to, first of all, if something like that happened, you might not even know, you wouldn’t really know about it until someone made a joke about it on late night, right? Like, they would have folks scouring the country for fun little local things to talk about that you would shine a light on. Instead, by the time Colbert talks about it and ropes in Jon Stewart and Jon Oliver and all these people to have fun with it, there have been millions of memes on the internet. There have been a ton of podcasts that have talked about it. There’s just been so much media about it that there’s nothing special about Colbert doing something. You have to, like, really nail the bit. And even if you nail the bit, it’s not the only way, it’s probably not the thing that people remember about that moment. And that is one of the reasons why late night is considerably less relevant than it used to be. And it’s also one of the reasons why I think one of the greatest ironies of this moment is that Stephen Colbert is going to be much better and funnier when he moves in the direction that comedy is moving, which is toward a guy and a mic being himself for an audience. I always felt like he was a, he was always a little bit of an awkward fit for this kind of like super like tie-up comedian that you have to be if you’re doing a late night show. That is theoretically purportedly speaking to like a hundred million Americans at once, even though it’s only talking to two million Americans at a time. He’s going to be so much better at being the kind of comedian that this age rewards—just like weird and talented and narrow and focused and specific and not trying to pretend to be, you know, keeping this vestige of 20th-century mainstream culture alive. So, what would you see him doing a year from now? Amy Poehler. He’ll be hosting a podcast. irreverent podcast that has him maybe blending a little bit of like character acting because he’s obviously so good at that. He’s an extraordinary improviser and I don’t think you can really, I mean, like as a comedian, he’s an extraordinary improvisational comedian. And I think it’s hard to really let that freak flag fly at whatever 10:30 PM on CBS at the Ed Sullivan Theater. The pomp and circumstance of that show doesn’t fit. I think his style—he’s weirder than it. And I think his weirdness will actually be perfect for the thing that he does next. Whatever it is. Like it goes, it, I guess it, it is to say that while there’s aspects of this sort of transition between institutions and individuals that I kind of bemoan, there’s parts of it that are like, so obviously wonderful, you know, more creation, more creativity. And I think the irony here is that five years from now, we’ll think that this is one of the best things to happen to Stephen Colbert. Okay. So you adopt, a lot of people seem to be adopting this belief that it’s the best thing to happen to him. Probably look, maybe not his bank account—$20 million is a lot to make in a year. I’m not worried about his bank account. He hosted that show for nine, ten years. He hosted The Colbert Report for a long time. He’s fine. He’s been, yeah, he’ll be fine. It’s, you know, I guess in a weird way, a good thing if you believe he was destined for that show to be canceled or end in the near future anyways. Doing so in a way where he comes out looking like a martyr and a hero is probably a good thing. And it gives, and the way it’s also been structured is he has time to figure out what to do. So, yeah, I know you’re right. There is something kind of interesting about how martyr culture is an important ingredient in modern entertainment success. Like Shane Gillis was better off being fired by SNL than hired by SNL, and maybe only in the 2020s could that be the case given all the things that he can do in the ways that he can go directly to audiences. But I think you put your finger on something quite important about how a little dose of martyrdom can go a long way in modern media. Oh, I think it, I mean, not to keep it all in the ringer family or talk out of school or anything. Like he leaves The Hollywood Reporter under this cloud where the owners of it wanted to influence the journalism. And so instead of having to leave at a certain point cause the magazine is losing money and it’s firing people and it just looks like a failing thing, he chooses to get out. He basically, you know, he leaves as the hero of the story and it paves the way for whatever he wants to do next. I totally think it helped him. Well, and The Ringer itself probably would not exist if someone had not shut down Grantland. Yes. I don’t know if Bill came out as the martyr. I guess he came out as the martyr there. Yeah. Well, I think to bring it back to the top point, this is a really interesting moment for comedy. And I think it’s really interesting the degree to which a lot of these contours that we see in the comedy market—the sort of miniaturization of the field as comedians go from wanting the biggest possible platform to a smaller platform, the fact that it’s becoming more tailored, the fact that it’s becoming more political—it’s just interesting how many of these trends seem to mimic and retrace things I’m seeing across media and entertainment. So we’ll leave it there. Lucas Shaw. Thank you very much. Thanks, Eric. Bundle and save with Expedia You were made to follow your favorite band and from the front row, we were made to quietly save you more. Expedia made-to-travel savings vary and are subject to availability. Flight-inclusive packages are not all protected. window.tocIndex = { "index": [ { "index_sentences": "Hey, it's Craig Horlbeck here to tell you that the NFL is back, whether you like it or not.", "section_level": 2, "section_title": "The Ringer Fantasy Football Show Introduction" }, { "index_sentences": "This episode is brought to you by Contentful.", "section_level": 2, "section_title": "Contentful Sponsorship" }, { "index_sentences": "This episode is brought to you by Indeed.", "section_level": 2, "section_title": "Indeed Sponsorship" }, { "index_sentences": "So I was on vacation in Maine last week with some of my best friends.", "section_level": 2, "section_title": "Lamenting the Decline of Adult Comedies" }, { "index_sentences": "We grew up on funny movies that became a part of our vocabulary: Old School, 40 Year Old Virgin, Anchorman, Superbad, Talladega Nights, The Hangover, Bridesmaids.", "section_level": 3, "section_title": "Legacy of Iconic Comedies" }, { "index_sentences": "But as many people have noticed and mourned, the adult comedy barely exists as a genre anymore.", "section_level": 3, "section_title": "The Disappearance of Adult Comedies" }, { "index_sentences": "I was reminded of that conversation last week when CBS announced that it was canceling Stephen Colbert and The Late Show.", "section_level": 2, "section_title": "CBS Cancels The Late Show" }, { "index_sentences": "Comedians used to think of jobs like hosting a late-night show as the apex of their field.", "section_level": 3, "section_title": "Shifting Career Paths for Comedians" }, { "index_sentences": "Today's guest is Lucas Shaw, a reporter for Bloomberg and frequent commentator on The Town Podcast.", "section_level": 3, "section_title": "Interview with Lucas Shaw: Introduction" }, { "index_sentences": "So I wanted to bring you on because I'm really interested in the cancellation of the Late Show in terms of what it says about television and comedy, and really more broadly, media entertainment.", "section_level": 3, "section_title": "Analyzing The Late Show Cancellation" }, { "index_sentences": "So we'll just dispense with the second one first, because it's worth getting it out of the way.", "section_level": 4, "section_title": "Debunking the 'Go Woke, Go Broke' Myth" }, { "index_sentences": "Anyways, I think the economic rationale is indisputable.", "section_level": 4, "section_title": "Economic Reasons for Cancellation" }, { "index_sentences": "There could be some classic Hollywood accounting in it, but there's no way that this is a profitable show and it's not a growing show.", "section_level": 5, "section_title": "High Costs and Lack of Profitability" }, { "index_sentences": "At the same time, it is almost impossible for CBS to get the benefit of the doubt right now, given everything else that has happened around their Paramount merger with Skydance.", "section_level": 5, "section_title": "Merger Context and CBS Credibility" }, { "index_sentences": "I want to present not a conspiracy, but maybe like a little half conspiracy when it comes to the politics here: Media companies are absolutely sucking up to Trump.", "section_level": 4, "section_title": "Political Cover for Cost Cutting" }, { "index_sentences": "One question I had is—I was hearing Matt Bell and he reported some of the numbers here: A hundred million dollar show, which is just unbelievable.", "section_level": 4, "section_title": "Unbelievable Show Costs and Alternatives" }, { "index_sentences": "Because a lot of people have responded negatively, like, 'Oh my God, how can the show cost a hundred million dollars?", "section_level": 5, "section_title": "Justifying the $100 Million Cost" }, { "index_sentences": "One thing I still don't understand about this decision is: if it's just about finances, why couldn't the network just cut costs to keep the show in the air?", "section_level": 5, "section_title": "The Question of Cost Reduction" }, { "index_sentences": "You know, when I asked the folks at CBS that, the answer was that the losses are so significant that there was no way to cut to profit.", "section_level": 2, "section_title": "The Digital Shift and Financial Realities" }, { "index_sentences": "Interestingly, I don't think that's a mystery at all.", "section_level": 3, "section_title": "Value Degradation in Digital Media" }, { "index_sentences": "'Late night talk shows are dying and have been for a long time.", "section_level": 3, "section_title": "The Demise of Traditional Late Night Format" }, { "index_sentences": "'Major comedians can now make more money and reach more people by touring and filming standup specials.'", "section_level": 3, "section_title": "Rise of Solo Entrepreneurship in Comedy" }, { "index_sentences": "I really do see this as part of a larger story about the shift from institutions, entertainment to individuals, entertainment.", "section_level": 2, "section_title": "Institutions vs. Individuals in Entertainment" }, { "index_sentences": "How do you, how do you define an institution in that case?", "section_level": 3, "section_title": "Defining 'Institution' in Media" }, { "index_sentences": "I think that they have become about the hosts and the institutions matter less and less, which is why it's so important for their longevity and for the future of those ideas that you build around the talent.", "section_level": 3, "section_title": "The Host Becomes the Brand" }, { "index_sentences": "I actually, in thinking about your really good question—and this is the nice thing about having a good journalist be the guest who sometimes becomes the host—I want to go back again to the interview that you had with that comedian.", "section_level": 3, "section_title": "Bert Kreischer's Career Path as an Example" }, { "index_sentences": "This episode is brought to you by Zendesk, introducing the next generation of AI agents built to deliver resolutions for everyone.", "section_level": 2, "section_title": "Zendesk Sponsorship" }, { "index_sentences": "This episode is brought to you by Indeed.", "section_level": 2, "section_title": "Indeed Sponsorship (Repeat)" }, { "index_sentences": "What does it mean to live a rich life?", "section_level": 2, "section_title": "Edward Jones Financial Advisor Sponsorship" }, { "index_sentences": "This is slightly a tangent, but it's a tangent I'm really interested in.", "section_level": 2, "section_title": "Revisiting the Decline of Adult Comedies (Deeper Dive)" }, { "index_sentences": "I think the simplest answer, which is different from one related to some of what we're talking about, is the globalization of the movie business.", "section_level": 3, "section_title": "Impact of Movie Business Globalization" }, { "index_sentences": "I really like those explanations.", "section_level": 3, "section_title": "Shift in American Audience Behavior" }, { "index_sentences": "I'm personally mystified by why horror has navigated this transition better than comedy.", "section_level": 3, "section_title": "Horror vs. Comedy in Theatrical Success" }, { "index_sentences": "One of the reasons is certainly that you can entertain yourself more at home.", "section_level": 2, "section_title": "The Decline of Television Comedy" }, { "index_sentences": "You, you triggered a couple thoughts here.", "section_level": 2, "section_title": "The 'Empires and City-States' of Media" }, { "index_sentences": "To the extent that comedy exists, it exists, again, at a very, like, individuated level.", "section_level": 3, "section_title": "Individualization and Political Comedy" }, { "index_sentences": "I mean, to tie it back to Colbert, right?", "section_level": 3, "section_title": "Mainstream Moments in Fragmented Culture" }, { "index_sentences": "And it's also one of the reasons why I think one of the greatest ironies of this moment is that Stephen Colbert is going to be much better and funnier when he moves in the direction that comedy is moving, which is toward a guy and a mic being himself for an audience.", "section_level": 2, "section_title": "Stephen Colbert's Post-Show Path" }, { "index_sentences": "So, yeah, I know you're right.", "section_level": 2, "section_title": "Martyr Culture in Modern Entertainment" }, { "index_sentences": "Well, and The Ringer itself probably would not exist if someone had not shut down Grantland.", "section_level": 2, "section_title": "Final Thoughts on Comedy and Media Trends" }, { "index_sentences": "Bundle and save with Expedia.", "section_level": 2, "section_title": "Expedia Sponsorship" } ] }; window.faq = { "qas": [ { "answer": "The speaker believes both phenomena are \"two pieces of the same story\": a shift from the \"age of institutions\" to the \"age of individuals\" in comedy and broader culture/economy, where artists find success working alone or alone-ish.", "index_of_source": "The demise of the adult comedy and the demise of the late-night host are, to me, two pieces of the same story.", "question": "What is the speaker's main theory connecting the demise of adult comedy and late-night hosts?" }, { "answer": "Comedians once considered late-night show hosting the \"apex of their field,\" but now, nobody under 50 seems to want the job, as younger comedians have different paths to success like stand-up specials or video podcasts.", "index_of_source": "Comedians used to think of jobs like hosting a late-night show as the apex of their field.", "question": "How did comedians used to view late-night show hosting jobs, and how has that changed today?" }, { "answer": "According to the folks at CBS, the losses for The Late Show were so significant that there was no way to cut costs enough to make the show profitable, which was the most satisfying answer for its cancellation.", "index_of_source": "When I asked the folks at CBS that, the answer was that the losses are so significant that there was no way to cut to profit.", "question": "Why couldn't CBS simply cut costs to keep The Late Show on air if it was purely a financial decision?" }, { "answer": "The \"half conspiracy\" suggests that media companies are \"sucking up to Trump\" and using the threat of extortion as a convenient cover for cost-cutting measures, making economic decisions appear political to appease the administration and facilitate mergers.", "index_of_source": "I want to present not a conspiracy, but maybe like a little half conspiracy when it comes to the politics here:", "question": "What is the \"half conspiracy\" regarding CBS's motivation for The Late Show's cancellation?" }, { "answer": "According to Lucas Shaw, the simplest answer for the decline of adult comedies is the globalization of the movie business, where studios prioritize massive blockbusters that work worldwide, while comedy is generally more localized.", "index_of_source": "I think the simplest answer, which is different from one related to some of what we're talking about, is the globalization of the movie business.", "question": "What is the simplest answer for why adult comedies are not made anymore, according to Lucas Shaw?" }, { "answer": "The Coldplay concert example illustrates that by the time late-night hosts like Colbert address a viral moment, millions of memes and podcasts have already covered it, making the late-night segment no longer unique or special, thus diminishing its relevance.", "index_of_source": "It's totally, and I brought it up because Colbert did something on it by the, you know, I think the next day.", "question": "How does the \"Coldplay concert\" example illustrate the diminished relevance of late-night shows?" }, { "answer": "The speaker believes Colbert's style—being weirder, talented, and an extraordinary improviser—is better suited for the current trend of comedy, which rewards individuals being themselves for a specific audience rather than trying to fit a 20th-century mainstream format.", "index_of_source": "He's going to be so much better at being the kind of comedian that this age rewards—just like weird and talented and narrow and focused and specific and not trying to pretend to be, you know, keeping this vestige of 20th-century mainstream culture alive.", "question": "Why does the speaker believe Stephen Colbert will be \"much better and funnier\" after The Late Show's cancellation?" } ] };

2025/7/23
articleCard.readMore

Advancing the Frontier of Silicon Intelligence: the Past, Open Problems, and the Future

Advancing the Frontier of Silicon Intelligence: the Past, Open Problems, and the Future So a few of my team members, although weekend, have been cooking a really good model and released last Saturday, on this Saturday actually. And if you haven’t tried it, you can open your chat, GT chat, GPT app, and if you don’t have it, please download it. If you click this, it’s part of discover, but if you click the button there, you will enter into the video, speech model. For example, we can try there on the far right. Yeah, it’s really hard to discover. I’m sorry about that. And hey, I’m actually giving a talk right now at Columbia about AI. Can you make a joke about Columbia, which is appropriate, a light one for you? Did the AI apply for Columbia? Because it heard even the algorithms get a degree of sophistication there. Good luck with your talk. Okay. That’s not bad. Okay, cool. Now let’s skip the A and let’s start the talk. So it all started around 1948-1950. During this time, the question arose: machine synonymy, right? So basically, we will come back again when people are talking about conversation synthesis. I really like the framing: we’re not trying to simulate an adult brain; we’re actually trying to simulate an infant’s brain and, subject to appropriate cause for education. So that’s exactly what machine learning is, and we will come back to this as well. And today, in this history, I will tell about the history of two cities: one is, you know, self-supervised learning and the other is reinforcement learning. So let’s go to self-supervised learning first, about 13 years ago. There’s, so basically that’s the first large-scale deep learning model, like using GPUs and a lot of data that achieves astonishing error rates on ImageNet. Much more data than before. One insight we got from this is with sufficient data and compute, new networks surpass humans in hand-engineered vision algorithms for the past few decades. So this was kind of a disaster for the people who were working on vision with hand-tuned features. It’s a nightmare because all their work for the past few decades doesn’t mean anything anymore suddenly. And this actually revived the interest in neural networks, and the deep learning revolution began that year. Most people see that as the year marking the deep learning revolution. Then in 2013, there’s a really fun funding from Google called word2vec. So basically, you can use a vector, an embedded vector, to represent words, and you can do arithmetic on this work. Like, if you take “king” minus “man,” you get “queen” minus “woman.” So you can have semantic meanings in the algebra operations. The other thing is, if you use the embeddings, they’re really good in the downstream tasks. This kicked off another two trends. One is from work to work to everything to work. Everyone wants the recommendations of the entity. You can represent the app when you’re recommending apps. You can represent a video when you try to recommend videos to users, and so on. The other thing is that the reinforcement of compute plus data performs much better than inductive bias. If you go back to Turing, right? Turing said, “we do not want to simulate an AI brain,” meaning we do not want our human inductive bias in the models. We want the model to be minimal, like minimal structure as possible. You want to build a model that just wants to learn, rather than building a lot of human prior into that. And this again proves that compute plus data is much better, better than several decades of human engineers. In 2014, there’s a really good paper called Al Nice. So basically, we have two nice networks; one is a generator, and the other is a discriminator. This has not a lot to do with self-learning, but it’s a really great idea that was applied almost everywhere later. Lemme skip that. In 2015, there’s a really good method for optimization called Adam. This really accelerated the progress in deploying models. The reason is you have a standard way of learning algorithms; you don’t have to hand-tune a lot of the parameters anymore. It’s especially good for large data sets, particularly for noise-frustrated scenarios. So they streamlined a lot of the training frameworks. Even till now, I think a lot of optimization methods address variations of Adam; they’re not the plan of Adam turning by this workload. In 2015, another really good paper came out titled ResNet. This is a really good paper that illustrates the problem we were facing with training very deep networks. Training very deep networks was really, really hard because you have a very deep network. The great advantage can explode, right? The genius idea is ResNet implies this, basically skip connections. Every layer can skip to the next layer as raw input, so you don’t have to. It’s basically an ensemble method. They ensemble every layer of the model from shallow to really deep networks. This is a really good illustration if you look at the area surface. A is the one that you don’t have the residual connections, C is very spiky. Those things are very hard to optimize, right? For B, this is a very smooth loss curve, so you can easily optimize that. It’s much smoother. Basically, this is almost all the networks; if you employ such a structure, it makes it much easier to learn. This is another fundamental paper that came out. Back then, there were a lot of thoughts about deep learning. People had a lot of doubts. Including myself, I started with a pure math background. I think there are a lot of things that intuitively are not correct, right? I was talking to a statistics professor yesterday, and I realized a lot of the things I learned in graduate school about statistics had wrong intuitions. The reason is that previously, people lived in such a lower-dimensional space. The intuition we got there did not generalize well in the hard and real space, where we have a trillion parameters. So I urge everyone in statistics, we should just study more of those problems rather than traditional ones. Because the intuition we give to students might be wrong. I had to spend years overcoming those wrong intuitions. One of them is different, like deep, this is a non-convex optimization. When you work on non-convex optimization, the first thing you worry about is being stuck in a local minimum. Depending on where you start, that’s really bad because how do you trust the result if the best solution is going to be stuck in a random local minimum? There are a lot of studies on this. I think one thing that makes me feel confident is that navigating a really high-dimensional space is actually really hard to get stuck in a local minimum. We live in a 3D space, right? When we see a 2D surface, there are a lot of local minimums, and they’re really bad. but it’s really hard to escape because they only have two degrees of freedom to do that. But when you live in a billion or trillion dimensional space, it’s really hard to stack in a local minimum because you have so many degrees of freedom. And the other thing people find out is even in that case, you’re stacking a local minimum. The local minimum, first, is not really bad. It’s actually a really flat local minimum. It’s actually a good one. It’s not very far away from the global minimum when you plug the data into the loss surface. And so those kind of two things I think make sure people don’t worry too much about the local minimum non-communication anymore. Our intuition just from 3D gives us a lot of fat, but those things do not hold in the redemptional space. A similar one is, if you are on a plane, the probability of random two vectors being aligned is almost zero, right? But if you look at, in dimensional space, the probability that random two vectors are near-aligned is almost one. They almost are, so almost. And those are different intuitions when you have a low versus high dimensional space. Am I going too fast or too technical, or this is okay? Good. Okay, a little bit too fast. Okay, I will slow it down. Thought number two. So if you are from traditional statistical analysis, right? You have a number of parameters larger than the number of data points you have, that’s a disaster. That leads to overfitting theoretically, right? And so that’s another thought because in most deep learning models, the number of parameters are usually, you know, more than the data points. And why doesn’t that lead to overfitting? That’s another thought. People, I mean, because I train as a math major and computer science major, so that’s always given me a thought. Then there are a series of studies. The good ones are, I think one thing that convinced me is even though when you have more parameters than data points, you can feed random labels, right? You can feed random noise. But what we discover is the deep learning model always learns the pattern first because we will come back to this when the deep stochastic gradient designs learn. They will learn the featured space where they have the largest A values. That basically means where you have the most pattern; they will learn that first before they learn the noise. So that basically means even though your prioritize doesn’t matter, they’ll learn the pattern. And what are the other good things? Then, there’s also this double descent paper, that similar idea, right? When you have a really over-parameterized model, once the network can interpolate, then it enters into a world where there is a big surface of zero loss points, and then the model tends to pick the best one. So that’s a convenience for even myself; they overcame that. This is not a bad thing; in fact, it’s actually a good thing rather than a bad thing. Then there’s the sequence to sequence model, learning and attention. In 2014, there is a familiar sequence of sequence model. This is used everywhere in all the applications, particularly in machine translations. And then in 2014, there is attention—the wait. Yeah, that’s the first attention paper. That’s not the attention zone need, but it’s attention on some other architectures. The major challenge point back then from 2014 to 2016 was that recurrent models were really hard to train in parallel because of their recurrent structure. You have to train the first step and then the next step and so on, which limits the size of the model and the training data since you can’t parallelize the process. Additionally, RNNs suffer from gradient diminishing. On the time horizon, we solved this depth dimension with residual networks, which can have the gradient diminishing issue, but on the time horizon, the problem persisted. Techniques like TM helped, but didn’t solve it completely. Then the transformer architecture came in, which is arguably the most important paper of the last decade. While it didn’t introduce a lot of new concepts, it combined many existing ideas into an architecture that solved most of the previous concerns. It eliminated recurrence entirely and relied solely on self-attention. Transformers stack multi-head attention and feed forward layers, offering much better data efficiency and parallelism compared to before. This allows for training much larger models with much more data, and transformers have become the backbone of nearly all cutting-edge NLP models and multimodal models. From there, something really interesting happened: in 2018 we saw GPT-1, in 2019 GPT-2, and in 2020 GPT-3—models that proved to be highly generalizable. This area is often called generative AI, though some prefer “gen” to stand for generalizable, which captures the essence better. Previously, you could build models for almost anything if you had data, but you needed to build a specialized model for each domain, which wasn’t scalable. The new regime of models is super generalizable; you can zero-shot or few-shot them to do your task, lowering the effort per model significantly. In 2020, a famous paper on the scaling law showed that as you increase compute, data, and parameters (all on a log scale), you get a lower loss in a clear linear relationship. The curve is beautiful and, while it will plateau at some point depending on the data, it’s not a physics law and may not hold everywhere. Notably, the scaling law almost perfectly predicted GPT-4 performance before it was released, holding across about 13 orders of magnitude—that’s a lot, up to 10 trillion. Returning to the starting point, this is the bitter lesson. If you haven’t read it, it’s highly recommended. The bitter lesson over the last 70 years is that AI started in 1950 with the goal of minimal inductive biases and building systems that can scale with compute. And the two measures that can scale with compute: One is search. The other is learning search; here doesn’t mean Google search. It means you explore new ideas and so on. For example, you can explore different moves. The gist of that is basically that compute ultimately outperforms the algorithms that leverage compute, ultimately outperforming those that rely heavily on human engineering inductive biases. We see that in image, we see that in NLP, we see that in everything. Basically, if you have data on compute, you just want to build a model that just learns rather than having to human engineer a lot of human inductive biases into the model. If you don’t take anything else from the talk, that’s the key takeaway. Then there are a lot of things that make you wonder again: where is scaling coming from? It’s just observation, right? People’s nature is that we want to understand where it’s coming from. There are many hypotheses. The major ones that resonate with me are: One is that scaling actually reflects the structure of the data distribution. If you look at the data, the data distribution follows the power law. For example, a really good doctor can solve rare diseases, while a mediocre doctor can solve the common diseases. Similarly, the higher the intelligence of the data, the less often it actually occurs. There might be millions of books on arithmetic, but maybe there are only a few on algebra and geometry. So that parallel distribution is kind of what might be underlying the scaling law; like you need extra 10x more compute to discover things that are rarer in the data or of higher intelligence in the data. This suggests that the scaling phenomenon is actually derived not from anything else but from the data’s intrinsic property. I really like that interpretation, and we’ll come back to it. We also discussed how a modeler is the most common pattern first. It is similar to when you do principal component analysis and use eigenvalue decomposition. You learn the gain features with the largest eigenvalues. That’s the nature of SGD (Stochastic Gradient Descent). The other thing that people have always wondered about, maybe two or three years ago, there were debates on the internet and Twitter about why those emergent abilities happen. You see this here, right? They don’t go smoothly or like massive capabilities; it just suddenly feels like the model can solve some massive problems. Why is that? I think it rules back to the scaling law. Even though the perplexity and loss are smooth, because of the power law in the data, capabilities suddenly come about when you have 10x compute; it’s when you finally understand calculus and can do calculus. So, it always depends on how you marry these concepts because you can always convert a discrete variable to a continuous variable in some way. It’s about how we view it, which is why people see those emergent abilities. I don’t think people should be surprised by that. It’s not sudden; it’s exactly just the nature of scaling. What reflecting in underlying data. And the one famous thing he has, the models just want to learn, I think at that stage. And we actually have a decent model architecture with this transformer that the model just wants to learn. You just want to feed the model data. And that’s actually a really good thing to have. So, okay, we overcome a lot of fat for the last 10 years. This is another one. I think a lot of people, including myself, do not have a lot of conviction. Now I do, but the thing is that compression of prediction leads to a really understanding and intelligence, and we don’t know, right? But there are two perspectives. One is from the information theoretical perspective; it is Shannon’s definition of information as unpredictability. If you can completely predict something, there’s no new information to you, right? So intelligence can be seen as the ability to reduce the price. Like if you are not surprised by anything in the world, which is hard, that means you’re really a wise person in many ways, right? And so if you think about what LLM does, LLM basically, by predicting the next word, they’re basically compressing a lot of the patterns so that it wants to get less price over time. In that way, you can understand, basically predicting the next word of compressing the patterns is kind of a form of intelligence similar to how humans predict the world. The other thing is from cognitive science perspective. As human beings, we have been trying to compress the infinite information in the universe, in this physics laws. Newton has low motion; we want the unified field series. Those are all compressions. We have a lot of observations and we want to compress them into minimal size of loss as possible. Similar for Max, we want the XM system, right? The other thing is our brain constantly compresses sensory input. So if you look, if nothing surprises you, you’ll have a reaction to that; otherwise, you’ll not. Your brain actually has a good way of compressing the things that don’t surprise you, but digesting only the new incremental information. In that sense, learning is compression. Like when you learn a new theory, it makes you excited and curious, and then you learn. But if you’re reading something boring, your brain already compressed that, and you get bored in many ways. Those two things, to me, are really good experience explanations why compression at least implies a large portion of intelligence, maybe not all. So that’s the first city, like, which is self-supervised learning. And then let’s go to the second, the reinforced learning part. This, the whole thing, the deep reinforcement learning, started in 2015. There’s a DQ in the DPU network that can play 2000 Atari games really, really well, much better than humans. I don’t know if I ever watched this video. Some of them are super fun because I grew up playing Atari games and they discovered a lot of strategies I never imagined. People call that alien intelligence because the way they learn is by playing, and they reinforce. But they don’t necessarily copy from humans. They invented other things that humans never thought of. And, 2016, I think that’s the biggest event maybe in the last decade about AI. People get really excited because goal is kind of the moment people say, “Oh, there’s really a lot of intelligence in these models.” The AlphaGo actually bootstrapped it from the human games. They basically combined the deep learning with self-play and Monte Carlo research to defeat the work campaign. In 2017, they refined that to AlphaGo Zero. So basically they didn’t bootstrap the deep neural network with any human data, all from self-play, which is just amazing. This is actually very interesting. I also read a lot of writings, and when I was growing up, I read a lot of this. There’s actually a Kung Fu master who is so good he cannot find anything that can defeat him. He started self-play, like he will just fight with a life-like arm and so on. It’s very similar to that. Then, you basically split your brains and your body, and then you fight each other. It’s pretty fun. In 2018, there’s AlphaZero. Basically, you can just, with minimal human inputs, tell the model the rules of the game and what constitutes a win and a loss, and no human data, domain-specific tricks. They can play chess and really, really well, not only Go. Then in 2019, people moved to digital video games like StarCraft and became the Grandmaster there. The interest along those lines died down because we actually proved we can play any game really, really well. We will come to a question about why it has less impact. Then there’s another branch at Berkeley and OpenAI, a group of people including John Schulman, who addressed the instability of the DQN algorithm. Then there’s PPO, which is behind the TBT algorithm, simpler to implement, smarter, and has better sample efficiency. And then there’s OpenAI, which has focused on OpenAI Five rather than just StarCraft. They played Dota 2 and became the world champion at Dota 2. The question is though, why did this impressive reinforcement learning treatment have limited direct impact on productivity in everyday life? They generated a lot of hype, right? People say, “The model can play Go,” but they generated minimal economic value. The reason for that is, they don’t have economically valuable generalizable players and environments. They’re very specialized in specialized AI or specialized super intelligence in some way. Things changed when we combined pre-trained reinforcement learning; I call this “part-time wine,” a pre-trained and low-computer RL. In 2022, there’s InstructGPT, which is basically training language models to follow instructions to be useful. If you don’t fine-tune the pre-trained language model, it’s just an auto-completer; it doesn’t really do the things that you want the model to do. In 2020, there’s TBT, utilizing a similar method with RHF. There was actually a low-key research preview by John Schulman, Barrett, Luke, and Liam, and people said, “Let’s just put it out there and see how people use it.” Now, there are over 500 million users using ChatGPT every week. I am always amazed by how helpful ChatGPT can be. This can save people then literally save people’s lives. There are users who are uploading their years of medical history and find out things that the doctor never told them before and saved their lives. This is not just a statement; this is a reality. And I use the video every day. Not only the voice mode, but I use deep researcher every week. My search volume has increased significantly, maybe by 10x, because I’m using the deep research model, doing all the searches for me; I don’t have to do that myself. It can be much better than an intern in many ways. But then you think about this: what changed from the game RL? We mentioned the game RL is not that useful. That’s an understatement; it’s really useful, but they didn’t generate economic value. Now it does, right? The reason is now RIL is combined with a much more generalizable pre-training, which can also apply to environments that can have a lot of economic value. But then you ask yourself, where is the general ability coming from? The majority of the general ability is still coming from the pre-training. Why is that? The pre-training is basically about the next word prediction. This approach has almost minimal inductive bias. Like Ilia said, the model just wants to learn a simple loss, and everything is trained rather than us teaching the model to do specific tasks. In the future, maybe we can do meta-learning or reasoning leads to generalization, potentially making the model even more generalizable. For the RL part, it’s not as generalizable as pre-training yet because, for RL, we have a lot of human-defined rewards where we introduce a lot of human inductive bias. I’m not saying humans are bad, but we want the model to learn. It’s like your kid; you want to teach your kid how to learn rather than teaching your kid specific things. Teaching our kids how to learn is much more important than imparting specific knowledge. Then there’s a concept people call young LA cake. The majority of the computation was spent on surprise learning. The icing is surprise learning, and the cherry on top is RIL. Basically, people are saying that in this regime, RL is a tiny amount of compute compared to the pre-training. But we believe that we need to change that. We believe more RL is required to build GI and SI and to adapt to entirely new environments that have not been seen by humans before. Let’s go back to VI Verizon again because I’ve read it several times, and I want to code it many times. Remember there are two things that we discovered in the last, say, 70 years that scale really well with compute. One is learning, which is pre-trained one search. We haven’t applied search that much, and that’s why we want to start paradigm two. Like OpenAI started this for pre-trained high compute IL in 2014, the O series came out and started the next paradigm of scaling. We want to scale the test compute, and if you look at the also 2025, there is an open-source version from deep. If you look at the performance on Amy, did anyone take Amy in this room? Only one, two, come on. More people take Amy. Oh three, cool. So this is the AMY score. When you give more compute in the train time, you can see from like 40 points to 80% percentile, right? And then for the test can be similarly from 20 to like 85 or 90. So giving the model more time to sync can significantly improve the AE performance, which is a pretty challenging task. It’s not that trivial. Uh, this is a new paradigm. We call that, you know, high computer. So basically, a lot of capabilities are really enhanced where personal learning, uh, but the caveat for there is on the public, you know, the papers publish; a lot of them only work well with verifiable rewards, meaning it’s a mess problem or it’s a coding problem. You know, that’s correct or not, right? In the end. And what we need to really do is to expand what is verifiable. And, uh, there’s a recent paper from David Silver and, uh, reset, welcome to Viral Experience. I think almost exactly, you know, illustrating the importance of this. And so the core idea, I think behind that, that paper is high quality. Human data is limited, like even though we have maybe centuries of civilization, right? And, but the actual data we accumulate is not that much. We already consume most of the intelligence text. The idea is, you know, how do we generate more, more data? But then you ask yourself a question: where does human data come from? Where is the human intelligence coming from? The human data comes from human brain thinking and getting reward or feedback from the real environment, which is the Earth, right? And then given that we know compute is gonna get cheaper and cheaper, we want to generate new knowledge and new data by using more compute from the computer rather than the human brains. By interacting with the environment and generating a lot of data, I think that’s gonna be faster than humans generating data in the future. That’s why I’m so bullish about AGI and ASI. The other thing is just anecdotal about why the chain also works, right? One thing that always puzzled me since the beginning of our model is why we spend the amount of compute in a token. It doesn’t make sense. Like there are our training corporates, right? There are tokens from the internet, the chitchat on Reddit, right? And there are also really intelligent conversations between two mathematicians, for example, right? Not every token we create is equal. And so we should not spend the same amount of compute in training and testing time for those tokens. And that’s where the chain source comes from, right? In channel salt, you can actually spend arbitrary long tokens for a really hard problem. And that gives you this adaptive compute. So the thing we want to do is basically save FLOPS playing tight rather than, you know, same flows per token. This is my personal journey. I can’t summarize it initially. I’m very optimistic about the AGI, because there are a lot of factors that non-com maximization. We know that can lead to very bad things, right? It’s our prioritization from a serial perspective before traditional statistics. That’s a bad thing. And, uh, the other thing that prediction leads to understanding, and, uh, we don’t know, right? But there’s a lot of good serious believing that that’s true. And our learning can also be most seeking. There are a lot of people worried about that. And the biggest one actually for me to overcome is I always believe human brains are special and human intelligence unique. Before, for example, all the things we talk about mathematically, they’re trivial. They’re basically tensor modifications and gradient design, and nothing fancy about that. So it makes one wonder, are we really that easy to replicate, to mimic, right? I believe life and life—so like human brains are so special as I understand a little bit more. You can simulate the human brains in many ways. Why not? Computers learn the same way as we do. So those are the next questions: why I’m more a G field, right? One is perhaps, brain cells aren’t uniquely special. It’s just basically a result of evolution. They’re already complicated, right? This is just a biological computer. There’s nothing special compared to a artificial silicon computer. The other thing is maybe scale is the thing that matters more fundamentally rather than how complicated the structure is. Intelligence may not arise from the complicated structure we have in our computations in the brain. The intelligence may, you know, rise from the data and the interaction we have with the environment. But maybe our brain, I mean, we talk about why those simple tensor or matrix multiplication can lead to intelligence. I don’t think our brain is doing something way harder than that or way more complicated than that. We’re not doing quantum computing with our brain in any way, right? Okay, come back to risk reset, right? Similar, Turing had the same idea 75 years ago. Turing basically says, “I doubt human mind is very complicated.” But we want to make sure we can simulate an infant brain and give that a course of education to make that really intelligent. This agrees with drawing. If you ever feel confused, I always go back to reset and drawing and read our papers. Those are the people who are actually ahead of their time for a few decades. That’s amazing. The other thing that made me a GIP over the last two, three years is you constantly observe, impossibilities become realities. Every couple of months, something you thought impossible becomes possible. Then you start doubting all the things you saw as impossible. They’re just, they’re just bs, right? Just ignore them. So even though there are a lot of open problems and lots of reasons to be concerned, there are also a lot of reasons to be optimistic. We will talk about the open problems in the next part of the talk. I’m checking how I’m doing on time. Good. So let’s talk about open problems. I want to make a few definitions to keep the discussion grounded a little bit because everyone has different definitions of AGI and AI and so on. So, when I say AGI, I mean a system that matches human level skill across virtually every domain. It’s basically the word that matters: generalizable. It’s not a special thing built for a goal for a video game. ASI is a system that surpasses the best human in every domain. Specifically, superintelligence is a system that surpasses humans only in specific domains. For example, AlphaGo and OpenAI Five, those surpass human capability in certain domains, right? We call that task-specific superintelligence. So what, what are the open problems? A lot of people say on the internet, “scaling law failed.” And I don’t think scaling for a scaling law failed. The data failed. Remember what I said—my personal perspective on the scaling law is the reflection of the data structure. So it can never fail. You just—it’s just a law, right? It’s the mapping. What really failed is actually the data. And so what do I mean by that? To fundamentally improve the capabilities, what we need to fundamentally improve is we need more or better or higher intelligence data so that the model can learn it through the scaling law. Learning, to me, is fundamentally data-bonded. If you have infinite data in every domain, we already saw the AI; given infinite compute—which I mean, here’s my way of thinking, right? Infinite compute is going to be here sooner or later. I don’t worry about that. But the thing that really bottlenecks us is the data. So then people ask, right? We can actually convert compute into data because human data are converted from human compute, right? From our brains. But why not convert the silicon compute to data? There are a few caveats that we haven’t solved yet: One is that right now, it’s only in limited domains where the results are verifiable. The current I/O may or may not effectively generate data outside the support of the previous policies. That’s exploration. And the other thing is, remember in AlphaGo, we did a Monte Carlo research and you do the random exploration, which can lead to win or lose the game. You cannot do that in language models. The reason is the language model is just many, many, many others. The space is many larger than the goal even. So random generational tokens can never lead to anything. I mean, that’s a strong word because when we have infinite compute, things change, right? People say you have a, you know, a cat; you know, “give me a typing machine if you tap Shakespeare at some point,” but that’s just going to be a super long time, right? By random luck. So that’s one aspect I mentioned. We want to improve the data we have. The other direction is actually to improve the data efficiency of learning. If the fundamental bond is data, we can either get more data or we can make our algorithm more data efficient. So that’s, we can talk about both of them. Better wine generating more and better data faster. We talked about this before, and if you think about how much data we accumulated: we have been accumulating printing press data for about seven years, right? In the beginning, it was very minimal. Then we have been collecting detailed data for the last two years or so. From the whole chain of where the human data is coming from, initially, the human first gives a task, right? Inspired by the environment, it can be solving polynomials. And I don’t know, does anyone know the history of solving polynomial equations? People were gambling around that and trying to get rich by solving more polynomial equations, right? I mean, the initial task we humans gave was to survive. That’s a meta-learning. The current generation of humans learns existing knowledge. And the third step is to think about the problem. For a long time, maybe some juniors came up, like Gs uh, gawa. and those people in new knowledge is right. And then you get feedback by interacting with the re environment, the peer-reviewed. And then you distill this new knowledge into funding into knowledge. You write a textbook, calculators, and then later write a paper on calculators, and then later generation you’ll learn that. And then you iterate. That’s how human journey, our day journey, our knowledge over the past so years, right? And, which aspect, you know, we can significantly accelerate by using AI or by using silicon compute. So about tasks, there are already a lot of open problems. For example, there are open conjectures in mass, like Riemann hypothesis and so on, but maybe there are two spars still. You don’t have a good clear room, right? And for learning. So that’s stage one for learning. I think model can learn very fast with weight or in contact. So that’s, we’ll talk about it later, how to make it even faster, data-efficient and for thinking. Models can think very fast. But the key question is can models generate new ideas? It doesn’t matter how fast you think if there’s no exploration, right? And then we ask ourselves, how does human discover new knowledge? Like I think the fundamental driving force is curiosity, right? We are curious. So we always explore new things, and so we want to really teach the model to be curious. There are different beliefs. My personal belief is maybe interpretation and exploration is enough. Like the model knows so much that they can interpret and extrapolate the knowledge, and that’s enough exploration to advance the intelligence. And I think maybe just one more minute on that, I have time. So the last, the universal list of mass is kind of funky, right? That which is, almost 130 years ago. And so the model now actually can know all the subjects of mass. So it’s a universal list, right? So it can maybe generate a lot of new knowledge or new ideas in that way. And the other most important thing for a human to get new knowledge is interacting with the environment. And this can be very efficient for computers. If there is a perfect simulator, for example, there’s for board games for goal, you can just simulate on the computer so you can get infinite data by using compute. And this is a fundamental blocker. If you cannot simulate all the same to real, the gap is really big, right? People are building this called work models for video or for the whole world for physics and so on, which is really hard still. And so that’s why the body AI is harder. You don’t have a perfect simulator of the physical world we are in. So I’m still open to music though, because if we can build a model that can reduce the search space like AlphaFold, we can build a positive flywheel. You have more efficient search and then you can generate more data, and then you have even more efficient search and so on, right? So that positive flywheel can lead us to super intelligence. And this relation model can do this very efficiently already when you have a new knowledge. Said sure. So I want to repeat the question. The gist of the question is when humans discover new mass, a lot of them are serendipity, right? There’s no particular goal first. I don’t agree with that. Preis, a lot of the pure mathematician. They still have a goal to solve a lot of the mass we invented in the process of solving a conjecture, right? But there are definitely a lot of serenity into those. I have a perfect answer for that in a few slides. Okay? The other thing is the open question: is embodiment necessary for a GI? It may or may not be depending on how you define it. If you define a GI to be the virtually economically valuable task, maybe not. But the thing is, humans can behave like the volume for AI burnout, right? For example, if the AI wants to do this physics experiment, they don’t have the environment to do that. But maybe humans can do the experiment for the AI and give the result to AI. And so that can also form this positive feedback loop. We do want to avoid a future that the AI views. We just become the embodiment of AI, which is not a good future to live in. The other thought, the current thought is: can RL generate new ideas? That’s a recent paper saying that current first learning really incentivizes reasoning capability beyond the base model. The base model is a pre-trained model, right? So they did an analysis. Here’s a brief summary: This is a number of samples. The Y is the passi K; basically, you passi one, meaning you do one rollout, and the rollout is correct. Pass 1 million means you do a million rollouts, and one of them is correct. What they find out is that after your RL, the passi one increases significantly. But past that, a million is the same. That basically means if you generate a million possible results, one of them is correct, even from the base model. So the RL model didn’t generate anything new. I don’t agree with that finding because this is on a particular view of open-source models and so on. I do believe RL can generate new ideas. The other key thing is exploration. This leads to the question: can we do more effective exploration in RL? We saw that humans have curiosity, and that leads to exploration. We believe exploration is definitely needed. How to do that is an open problem. This goes back to inspiration and exploration, and maybe that’s enough. I am personally very optimistic because if you look at how humans advanced science and technology, how much of it is purely genius ideas? It’s a very tiny amount, right? Most of our work in science is based on the interpretation of the exploration of previous works, right? The model can do that incredibly well compared to humans. I think that’s maybe just enough. I view the question in the end that we’re becoming running all the time. There’s really encouraging evidence from Alpha Evolve to validate the idea. If you haven’t read that paper, it’s a fascinating paper that just came out a month ago. They demonstrated the power in context learning and guided exploration. Remember, when I was an undergrad, I was learning how to do matrix manipulation more efficiently, right? The state-of-the-art algorithm was from 1969. Wait, how many years ago? A lot of years ago—50 years ago. Yeah, it’s 50 years ago. It’s crazy, right? And then you use AI to improve on that, basically. Oh, this is a state of art. Can you improve upon that? And step by step, they actually discovered a better algorithm than that. This is very inspiring. Basically, AI can solve the problem we didn’t know how to solve for like 50 years, in some sense, right? Out of the 50 open problems, I think 20% of them, they improve the state of art, especially for this MP heart comparator problems. AI is much, much better. I really believe, I studied pure math, and the reason for that is in the freshman year of college, I learned Galois theory. I attended a seminar by Professor Hu in my university, and I became fascinated by Galois theory—just so beautiful, amazing. That led me to pure math. My dream was, oh, I want to solve a conjecture, something like the Riemann hypothesis, right? A couple of years in graduate school, I realized I’m not suited for that. I couldn’t do that. So, I gave up, and now my new dream is we can build AI to solve the Riemann hypothesis, and I’m much more confident about that than myself solving the hypothesis. The second goal is we want to vastly improve data efficiency learning. If you think about the current AI, the data efficiency is still very bad compared to humans. For human learning, if you teach a person a new board game, it probably requires several minutes or maybe a few thousand tokens to learn to play the game pretty well. But the current models couldn’t do that. It requires maybe ten to a hundred times more tokens to teach a model something really well. But think about this: why is that the case? This is just personal thinking and discussions with some friends. Humans do not learn by predicting next tokens. We do sometimes try to predict what you’re going to do or what your next move is, but we do not predict exactly what your token is. The reason is that distinguishing this is really important because predicting your token means there are a lot of random structures in the token that the model is trying to predict. So basically, the model is wasting a lot of computational resources and parameters predicting something random. As humans, we predict on a much higher level, at a more abstract level. This way, we know the essence of the problem rather than trying to predict the structure of the tokens. Think about the idea; there are a million different ways to say the same idea. If you are predicting the next tokens, you are trying to capture these random structures in there. I have no idea how to solve this. What is the next paradigm? This is the open question: how do we vastly improve data efficiency? If you can solve that, I think that could be what the next paradigm for AI will be. The other thing is to make open problems like misreasoning more manageable. Did I answer your question through these slides? How should I—let’s go back to that. Your question was, mathematicians discover things a lot of time by serendipity, right? Models are not—this goes back to this slide, the funky one, right? I agree a lot; some of the discoveries are by chance, not by any utility or anything. If you look, go back to this one from Sir—I don’t know what I may seem. To the word, but to myself, I seem to have been only like a boy playing on the seashore, and the wording myself now and then finding a smoother or a prettier shell than the ordinary. And well, the greatest ocean, the truth, lay discovered before me, right? This is very poetic, poet thing that I really love that code. And to your point, meaning some of the science and technology will discover by serendipity, right? But AI can improve. It’s basically a search problem. There are a vast space of signs we need to search, and then we get a pebble or we get a shelf, right? But the good thing about AI is that it is going to reduce that search space so much that serendipity becomes common things, right? So, I’m going structures, and then you know, use that for… I think that’s a really good point. We can talk about that after the talk a little bit more. I’m really optimistic, and I also see progress in directions like that. And I don’t believe anything special about human serendipity. Compared to machine serendipity, let’s continue. The other open question is, what is the next scaling paradigm, right? We talked about some of that. If you look at paradigm one, paradigm two, we scaled the depths of the layers and we scaled the test compute RL. Maybe you want to scale the number of two self-play contexts and memory. The model still doesn’t have infinite memory yet, which is really important, right? Because for a really good friend, you share a common memory that is almost infinite. And, lifelong learning is really, really important. I’m running low on time, so I’m going to be a little bit faster. The other safety concerns are three kinds of safety. One is that the model may generate unsafe content. This is very similar to traditional trust and safety work. The second one is that maybe bad actors can leverage the model to do bad things, right? We want to prevent that from happening. The third one is more serious or exponential, meaning the model itself becomes bad and misaligned. There is a lot of research going into this. I think that’s a really, really important problem we’re tackling. In the third part of the talk, I want to dream a little bit and see onto the future what we envision. I envision that the future can be, and some of them maybe in the near future, some of them maybe five, ten years down the road, maybe more than ten years. And maybe we’ll see. So one quote I really like from Sam is he always says, “the days are long” because everyone felt that, right? I had a long day sometimes, but “the decades are short.” And I also feel that I cannot remember; I graduated more than ten years ago. This is very scary in many ways. Most people tend to underestimate the impact of AI; they’re going to overestimate the short-term impact or the short-term development of AI. But I think one of the reasons I want to give this talk is that I think most people also underestimate the medium to long-term impact of AI. It’s going to fundamentally change everything. So let’s see, what’s the future you like? I have a hypothesis: when we have a generalizable human prayer, we have unbounded compute RL, and we have a good environment, that equals a superintelligence. If you look at how humans learn, that’s exactly how humans evolve, right? And we think the model can evolve even faster than that. So let’s see, the first area is, I think agents are going to be everywhere. And you will see much more reliable and capable agent in the next year or so, and they become a real reality. And I think to a large extent, it’s just basically an execution problem. There’s not a lot of research open problem to make agents work. AI for science is something we want to discuss before, but it’s something I’m really, really excited about. If you think about science, it is a search problem, right? It’s a massive search space. Once we have enough data to boost and drive a model, then you can make the search much, much more trackable. I think in our most subjects in science, we’re actually in that environment or in that stage; we accumulate enough updated human data. Let’s use AI to make the search more trackable. I really hope that you know about move 37 of the alpha goal moment. People know about this or should I explain it a little bit? Yeah, good. There’s a quote from Max, who is, I think, the chief scientist from the Isomorphic Lab saying, “I really agree with that by the way. Doing drug design without AI is like doing science without mass.” I think AI becomes the new mass for science in the next decade also. If you’re not using AI in your research, please consider doing that. I believe every university should have a huge capability to either develop their own model, our own clusters, or use cloud resources. It’s gonna be science, and it’s gonna be a lot. I’m really fascinated by AlphaFold and a lot of material science and AlphaFold we talk about, right? Our friends working in the AlphaFold areas tell me a lot of the drugs are already actually in clinical trials. I always thought that’s kind of in the far future, but over the last year, it has progressed so much. One particular thing I think there are in clinical trials is how to neutralize the poison from snake bites. Those are really easy problems. You just need to cover a certain part of the protein. Then from the protein structure, you go to amino acid sequences, right? However, the problem before was that from a structure to amino acid sequence, there are millions of possibilities. This is a gigantic search space. You cannot do that in the wet lab because you don’t have enough graduate students or enough resources. But graduate students are really good. AlphaFold has a solution; from the protein structure to the amino acid sequence, that’s a high recall, low precision mapping, right? AlphaFold is the universal high precision mapping. Basically, you map to the protein structure space and compare it to the protein structure that you want. Then you can narrow that down to maybe a hundred amino sequences in that way. Exactly. You have enough graduate students to do the experiment. So, that’s how those drugs go to clinical trials. Again, I think this is one more demonstration that AI can reduce the scientific search space a lot. If you’re doing any search in your discovery, use AI in math, physics, biology, chemistry—everywhere. This is the protein structure if you’re interested, let’s go. I also have a conjecture on AI for science. I really hope in the future that in addition to using graduate students, we have a lot of machines and robots in our labs that can automate much of the experimentation. Then you will have a really fast feedback loop. So the model common is hypothesis, although the amino acids you need to do, right? And then the machines just do it and then come back and tell the AI, “Oh, this is this work that doesn’t.” And the AI evolves. Keep continuing training on that, and that feedback loop is going to be super valuable. The other thing is what I hope is Alpha 40 is a specialized model, right? What I hope is similar to the, you know, from Alpha Go to ours. What if we can develop a generalizable model that covers all subjects, similar to LMS, rather than building a specialized model in science? I think enterprise R&D is going to be fundamentally changed by AI. There are two things there. One is AI can fundamentally boost the R&D productivity, for example, of the coding agent. The other thing I’m even more excited about is maybe AI will break the ceiling for a lot of R&D frontiers, similar to how AI accelerates science. AI product education is something I’m super passionate about as well. If you look at society, I think one fundamental thing is that people have access to different quality of education, which leads to further division in society. There are two things AI can help with: Lowering the barrier: AI can make topics more accessible. Because it’s synthetic data, if you look at a random article about a subject, you may get intimidated. It’s not written in a friendly way. But AI can lower that barrier a lot. Becoming a personal tutor: AI can be personalized. There are studies showing that having a personal tutor can significantly improve learning efficiency, sometimes doubling or quadrupling it. The other thing is raising the ceiling. What we call 10x learners can do this all the time. Now, over the weekend, I just run research on different topics that I’m interested in and read the report, and suddenly become not an expert but rather than a stranger to a field, you become entry-level knowledgeable, which is amazing. The hypothesis we have is that maybe in a few years, rather than spending five years on a PhD, you could accumulate five PhDs or 10 PhDs in that time. If you really want to learn, AI for healthcare is super critical as well. I think AI is already better than most healthcare providers that people have access to. We have numerous examples of how ChatGPT can save lives, and they are real cases across different domains. I’m also really excited about the idea that if AI can have the holistic context of a patient’s medical and health history, it can predict and enable a lot more preventive healthcare in many ways. Embodied AI is something that may be a little further away, but once that’s true, it’s going to have a tremendous impact on society as well. It’s not as mature because we don’t even know how to efficiently tokenize actions. There are suggestions, but they’re not super good yet in leveraging the existing massive amount of video data for generalization. The key thing about building AI is we do not have a lot of data. If you look at how LLM started, I did a Twitter post about the following, and then it got really popular; I actually deleted that post. The following is, at that moment, I felt over all the years we have been preparing for AI. Why I’m saying that we invented the printing press to record our intelligence and then we invented inter-computers, and then imagine the internet. All those things leading to accumulation, all the data we need to build AI and all the compute capabilities we build lead to the capability we want to train AI. In some way, we have been preparing for this; for solving human beings has been preparing the moment for AGI for so many years. This is a little bit emotional; that’s why I deleted. But if you look at robotics, we do not have the internet equivalent of robotics. The only way maybe everyone has a recorder from their eyes is how things are happening in the real world. We do not have the amount of data for robotics to be super effective yet; even if we have that amount of data, we don’t have good ways, a really simple, efficient way to learn that yet. So those are the two open problems for robotics. A lot of the demos are really cool for robotics, but the demo-to-product gap is pretty high. In case you don’t know, here is my dream come true case: the universe. One of the reasons we never discover aliens might be that we communicate in different modalities, like we never discover each other, or we live in different dimensions, and so on. Another reason is that most of the universe is not inhabited by humans. It’s really hard for us to go out, but that’s not true for silicon intelligence. What if we build embodied intelligence that can explore the universe for us? I think that’s going to be super cool in the future. That’s the end of my talk, and thanks for your attention. I’m happy to take any questions. This is a code I got from Rich T, and this applies to a lifelong agent in RL and also applies to humans: Never stop learning. Oh, hi. Thank you for the great talk. I have two questions. I think one is, what is Moore’s Law for LLM and where are we on the curve? That’s part one. The second is, I think I’m a skeptic of AGI because what we are doing with all the neural networks and the matrix multiplication is a very narrow part of the human mind, which is just human reasoning. But human reasoning is a very narrow part of the human experience. I think we are hardly touching on the human experience, which also consists of emotions, and we’re very early in the process of that. There is another piece we are completely missing: what is consciousness and self-consciousness? I think NYU, some people are studying that, but without any of that, we will never get anywhere close to AGI or SI. To answer the first question, I think Moore’s Law can mean many things. One is more Moore’s Law for IM can be interpreted as the scaling law. As we discussed, scaling law is an intrinsic property of the data. We really want to make better data, higher quality, more data, and faster. In terms of a lot of consciousness about AI and so on, I agree, and I don’t have a really good definition of consciousness. There’s one paper coded by risk Sutton. Again, I think Saturn does saw all the doubt I had. But it mentioned that maybe consciousness is coming from embodiment, like we are conscious because we have a body. Like this is my feet, my arm, and this is my head. And so, my heart, and I don’t know if that’s true or not because I don’t really understand what’s consciousness; it is kind of a little bit abstract to me. But I agree there are definitely a lot of open problems, and I’m optimistic. We have, we’re gonna take two more questions. Okay. In the discussion, Luca is our department chair. I wanna make sure we get to, “Thank you for a great talk.” Two words that came up a lot in your talk are: Efficiency often related to data efficiency, if I understood correctly. Scalability. And then at some point in the talk, you had a sentence about the human brain, and I don’t know if it was provocative or not, but you said, “after all maybe it’s not so special.” My understanding is that the human brain does often not always do a lot of amazing things while staying within 20 watts. So my question to you is how would you talk about scalability and efficiency if you consider also the energy aspect? Yeah, I agree. I think human brains are just so much more efficient in terms of learning than the silicon computers. Right now, the two aspects of this are: There are enough data in a lot of the subjects already, even though the current LMS are not as data efficient as humans; they can learn really well. I’m not super worried about the energy efficiency that much yet because I’m really confident we can learn infinite models with energy and compute. If anything helps, human beings are really good at generating energies and compute over the long run, so we wouldn’t be compute-bounded anyway. I do agree on one thing: I don’t have really good thinking along this; humans just learn so much more efficiently. We see a tiny fraction of the data the machine sees. To me, that’s kind of the either of the module architecture or the loss function. We talk about the next token prediction loss; that’s not how human beings do it. We have a higher level abstraction in learning things. That part is a solving problem. I don’t know how to address that. Anyone who can address that is going to win the next Nobel Prize. Oh, thank you. My name is Nick Gu. I’m a computational neuroscientist and a neuro AI researcher. I should have two questions, but I’ll ask one. There was a paper published by IPO probably in the past few days, saying that the reasoning for complex tasks is where these models actually break at high complexity. Yeah, I accidentally read a paper from 2010 where they were talking about human reasoning models and it comes to mind that humans do not always use pure, logical reasoning. I’m sure the audience can think about the last time you made a decision with purely logical reasoning. I’m curious to hear your perspective on specific AI developments in terms of the reasoning models and how it would be. Sorry, I’m a bit nervous about how I am able to share this. Yeah, so thanks for the question first. You expressed it well. I know nothing about the brain, and I didn’t read that paper. I do see a lot of discussion on Twitter about that paper. And then there is a new paper coming out, a paper called Human Brain Really Reason, right? I think it depends— all depends on our opinion on reasoning. It is more about whether humans can really reason. Like, does the human brain think that’s called reasoning according to that paper’s definition as well? I think that’s an open question. We don’t know. Yeah, we do one for it. Reasoning is the way that you plan and backtrack all those competencies. Humans do. I think currently the model can already do that pretty well. So I tend to not argue about more abstract and hypothetical things, but in terms of utility, they’re actually mostly there in many ways. Last question, DHA. Hey, thank you so much for joining us. I have a pretty simple question. My name’s DHA Ka. Do you believe there’s a general algorithm for discovery and invention? So there’s a spectrum of discovery and invention, right? And there’s interpolation and random exploration basically. If you look at AlphaGo, they explored some randomness in there. Because they can always have a win or lose situation, the randomness is okay; then you can learn from the randomly good or bad things. I don’t know a really good algorithm. My belief is still maybe it’s something in between, and that can take us already super far. The ceiling is already super high in terms of how humans— you know, serendipity works. We do not discover things randomly. That’s hard to understand. I don’t know that. Thank you so much. Thanks, everybody, for coming. So do keep us updated, subscribe to our engineering, Twitter accounts or LinkedIn so we can have more and better lectures in the next semester as well. Thanks again for coming. Bye. window.tocIndex = { "index": [ { "index_sentences": "So a few of my team members, although weekend, have been cooking a really good model and released last Saturday, on this Saturday actually.", "section_level": 1, "section_title": "Introduction" }, { "index_sentences": "Now let's skip the A and let's start the talk.", "section_level": 1, "section_title": "History of AI: The Two Cities" }, { "index_sentences": "So it all started around 1948-1950.", "section_level": 2, "section_title": "Early Beginnings and Machine Synonymy" }, { "index_sentences": "So let's go to self-supervised learning first, about 13 years ago.", "section_level": 2, "section_title": "Self-Supervised Learning (SSL)" }, { "index_sentences": "There's, so basically that's the first large-scale deep learning model, like using GPUs and a lot of data that achieves astonishing error rates on ImageNet.", "section_level": 3, "section_title": "Key Milestones in SSL" }, { "index_sentences": "There's, so basically that's the first large-scale deep learning model, like using GPUs and a lot of data that achieves astonishing error rates on ImageNet.", "section_level": 4, "section_title": "AlexNet" }, { "index_sentences": "Then in 2013, there's a really fun funding from Google called word2vec.", "section_level": 4, "section_title": "word2vec and Embeddings" }, { "index_sentences": "In 2015, there's a really good method for optimization called Adam.", "section_level": 4, "section_title": "Adam Optimizer" }, { "index_sentences": "In 2015, another really good paper came out titled ResNet.", "section_level": 4, "section_title": "ResNet and Deep Network Training" }, { "index_sentences": "Back then, there were a lot of thoughts about deep learning.", "section_level": 3, "section_title": "Overcoming Early Doubts" }, { "index_sentences": "One of them is different, like deep, this is a non-convex optimization.", "section_level": 4, "section_title": "Non-convex Optimization & Local Minima" }, { "index_sentences": "Thought number two. So if you are from traditional statistical analysis, right? You have a number of parameters larger than the number of data points you have, that's a disaster.", "section_level": 4, "section_title": "Over-parameterization & Overfitting" }, { "index_sentences": "Then there's the sequence to sequence model, learning and attention.", "section_level": 3, "section_title": "Sequence Models and Attention" }, { "index_sentences": "And then that's where the transformer came in.", "section_level": 3, "section_title": "Transformers: The Most Important Paper" }, { "index_sentences": "So there's 2018, that’s the GPT-1, 2019, the GPT-2, 2020, the GPT-3, and we discovered that those models are really generalizable.", "section_level": 3, "section_title": "Generative AI and Generalizability" }, { "index_sentences": "So there's 2018, that’s the GPT-1, 2019, the GPT-2, 2020, the GPT-3, and we discovered that those models are really generalizable.", "section_level": 4, "section_title": "GPT Series and GenAI Definition" }, { "index_sentences": "Before we could build models by anything almost if you had data, right?", "section_level": 4, "section_title": "Specialized vs Generalizable Models" }, { "index_sentences": "And then in 2020, there's a really famous paper on the scaling law.", "section_level": 3, "section_title": "The Scaling Law" }, { "index_sentences": "So basically, the X-axis is the log, the lower, the better.", "section_level": 4, "section_title": "Observations and Prediction" }, { "index_sentences": "And then let's go back to the point we started in the beginning. This is the bitter lesson.", "section_level": 4, "section_title": "Bitter Lesson and Compute Scaling" }, { "index_sentences": "Then there are a lot of things that make you wonder again: where is scaling coming from? It's just observation, right?", "section_level": 3, "section_title": "Understanding Scaling and Emergent Abilities" }, { "index_sentences": "I think a lot of people, including myself, do not have a lot of conviction. Now I do, but the thing is that compression of prediction leads to a really understanding and intelligence, and we don't know, right?", "section_level": 3, "section_title": "Learning as Compression of Prediction" }, { "index_sentences": "And then let's go to the second, the reinforced learning part.", "section_level": 2, "section_title": "Reinforcement Learning (RL)" }, { "index_sentences": "This, the whole thing, the deep reinforcement learning, started in 2015.", "section_level": 3, "section_title": "Deep RL Milestones" }, { "index_sentences": "There's a DQ in the DPU network that can play 2000 Atari games really, really well, much better than humans.", "section_level": 4, "section_title": "DQN" }, { "index_sentences": "And, 2016, I think that's the biggest event maybe in the last decade about AI.", "section_level": 4, "section_title": "AlphaGo / AlphaZero" }, { "index_sentences": "Then in 2019, people moved to digital video games like StarCraft and became the Grandmaster there.", "section_level": 4, "section_title": "Game Superintelligence" }, { "index_sentences": "Then there's another branch at Berkeley and OpenAI, a group of people including John Schulman, who addressed the instability of the DQN algorithm.", "section_level": 3, "section_title": "Addressing RL Instability: PPO" }, { "index_sentences": "The question is though, why did this impressive reinforcement learning treatment have limited direct impact on productivity in everyday life?", "section_level": 3, "section_title": "Limited Impact on Productivity" }, { "index_sentences": "Things changed when we combined pre-trained reinforcement learning; I call this \"part-time wine,\" a pre-trained and low-computer RL.", "section_level": 3, "section_title": "Combining Pre-trained Models and RL" }, { "index_sentences": "In 2022, there’s InstructGPT, which is basically training language models to follow instructions to be useful.", "section_level": 4, "section_title": "InstructGPT and ChatGPT" }, { "index_sentences": "This can save people then literally save people's lives.", "section_level": 4, "section_title": "Real-world Impact" }, { "index_sentences": "But then you ask yourself, where is the general ability coming from? The majority of the general ability is still coming from the pre-training.", "section_level": 4, "section_title": "Generalizability Source" }, { "index_sentences": "We want to scale the test compute, and if you look at the also 2025, there is an open-source version from deep.", "section_level": 3, "section_title": "Paradigm Two: Scaling Test Compute" }, { "index_sentences": "The caveat for there is on the public, you know, the papers publish; a lot of them only work well with verifiable rewards, meaning it's a mess problem or it's a coding problem.", "section_level": 4, "section_title": "Verifiable Rewards Limitation" }, { "index_sentences": "And so the core idea, I think behind that, that paper is high quality.", "section_level": 4, "section_title": "Generating More Data with Compute" }, { "index_sentences": "The other thing is just anecdotal about why the chain also works, right?", "section_level": 4, "section_title": "Adaptive Compute: Chain of Thought" }, { "index_sentences": "This is my personal journey.", "section_level": 1, "section_title": "Personal Journey and Overcoming Skepticism" }, { "index_sentences": "Initially, I'm very optimistic about the AGI, because there are a lot of factors that non-com maximization.", "section_level": 2, "section_title": "Initial Doubts" }, { "index_sentences": "Why I'm more a G field, right?", "section_level": 2, "section_title": "Overcoming Skepticism" }, { "index_sentences": "One is perhaps, brain cells aren't uniquely special.", "section_level": 3, "section_title": "Brains Not Uniquely Special" }, { "index_sentences": "The other thing is maybe scale is the thing that matters more fundamentally rather than how complicated the structure is.", "section_level": 3, "section_title": "Scale Matters More" }, { "index_sentences": "Okay, come back to risk reset, right? Similar, Turing had the same idea 75 years ago.", "section_level": 3, "section_title": "Turing's Infant Brain Idea" }, { "index_sentences": "The other thing that made me a GIP over the last two, three years is you constantly observe, impossibilities become realities.", "section_level": 3, "section_title": "Impossibilities Becoming Realities" }, { "index_sentences": "So even though there are a lot of open problems and lots of reasons to be concerned, there are also a lot of reasons to be optimistic.", "section_level": 2, "section_title": "Overall Optimism" }, { "index_sentences": "I'm checking how I'm doing on time. Good. So let's talk about open problems.", "section_level": 1, "section_title": "Open Problems in AI" }, { "index_sentences": "I want to make a few definitions to keep the discussion grounded a little bit because everyone has different definitions of AGI and AI and so on.", "section_level": 2, "section_title": "Definitions" }, { "index_sentences": "A lot of people say on the internet, \"scaling law failed.\"", "section_level": 2, "section_title": "Data Failure and Bottleneck" }, { "index_sentences": "The other direction is actually to improve the data efficiency of learning.", "section_level": 2, "section_title": "Improving Data Efficiency and Discovery" }, { "index_sentences": "And then we ask ourselves, how does human discover new knowledge?", "section_level": 3, "section_title": "Human Discovery: Curiosity, Exploration, and Serendipity" }, { "index_sentences": "We are curious. So we always explore new things, and so we want to really teach the model to be curious.", "section_level": 4, "section_title": "Curiosity and Exploration" }, { "index_sentences": "My personal belief is maybe interpretation and exploration is enough.", "section_level": 4, "section_title": "Interpretation and Extrapolation Hypothesis" }, { "index_sentences": "And the other most important thing for a human to get new knowledge is interacting with the environment.", "section_level": 4, "section_title": "Interaction with Environment / Simulation" }, { "index_sentences": "But the good thing about AI is that it is going to reduce that search space so much that serendipity becomes common things, right?", "section_level": 4, "section_title": "Serendipity and Search Space Reduction" }, { "index_sentences": "The other thing is the open question: is embodiment necessary for a GI?", "section_level": 2, "section_title": "Is Embodiment Necessary?" }, { "index_sentences": "The other thought, the current thought is: can RL generate new ideas?", "section_level": 2, "section_title": "Can RL Generate New Ideas?" }, { "index_sentences": "The second goal is we want to vastly improve data efficiency learning.", "section_level": 2, "section_title": "Vastly Improving Data Efficiency (Learning)" }, { "index_sentences": "If you think about the current AI, the data efficiency is still very bad compared to humans.", "section_level": 3, "section_title": "Comparison to Humans" }, { "index_sentences": "Humans do not learn by predicting next tokens.", "section_level": 3, "section_title": "Next Token Prediction Limitation" }, { "index_sentences": "What is the next paradigm? This is the open question: how do we vastly improve data efficiency?", "section_level": 3, "section_title": "Open Problem for the Next Paradigm" }, { "index_sentences": "The other thing is to make open problems like misreasoning more manageable.", "section_level": 2, "section_title": "Making Open Problems More Manageable" }, { "index_sentences": "The other open question is, what is the next scaling paradigm, right?", "section_level": 2, "section_title": "Next Scaling Paradigm" }, { "index_sentences": "The other safety concerns are three kinds of safety.", "section_level": 2, "section_title": "Safety Concerns" }, { "index_sentences": "In the third part of the talk, I want to dream a little bit and see onto the future what we envision.", "section_level": 1, "section_title": "Dreaming into the Future" }, { "index_sentences": "I envision that the future can be, and some of them maybe in the near future, some of them maybe five, ten years down the road, maybe more than ten years.", "section_level": 2, "section_title": "Vision and Perspective" }, { "index_sentences": "I have a hypothesis: when we have a generalizable human prayer, we have unbounded compute RL, and we have a good environment, that equals a superintelligence.", "section_level": 2, "section_title": "Hypothesis for Superintelligence" }, { "index_sentences": "I think agents are going to be everywhere.", "section_level": 2, "section_title": "Agents Everywhere" }, { "index_sentences": "AI for science is something we want to discuss before, but it's something I'm really, really excited about.", "section_level": 2, "section_title": "AI for Science" }, { "index_sentences": "If you think about science, it is a search problem, right?", "section_level": 3, "section_title": "Science as Search Problem" }, { "index_sentences": "Doing drug design without AI is like doing science without mass.”", "section_level": 3, "section_title": "AI as the New Math" }, { "index_sentences": "Our friends working in the AlphaFold areas tell me a lot of the drugs are already actually in clinical trials.", "section_level": 3, "section_title": "Drug Discovery Examples" }, { "index_sentences": "I really hope in the future that in addition to using graduate students, we have a lot of machines and robots in our labs that can automate much of the experimentation.", "section_level": 3, "section_title": "Automating Experimentation" }, { "index_sentences": "What if we can develop a generalizable model that covers all subjects, similar to LMS, rather than building a specialized model in science?", "section_level": 3, "section_title": "Generalizable Science Model" }, { "index_sentences": "I think enterprise R&D is going to be fundamentally changed by AI.", "section_level": 2, "section_title": "Enterprise R&D" }, { "index_sentences": "AI product education is something I'm super passionate about as well.", "section_level": 2, "section_title": "AI Product Education" }, { "index_sentences": "AI for healthcare is super critical as well.", "section_level": 2, "section_title": "AI for Healthcare" }, { "index_sentences": "Embodied AI is something that may be a little further away, but once that's true, it’s going to have a tremendous impact on society as well.", "section_level": 2, "section_title": "Embodied AI" }, { "index_sentences": "In case you don't know, here is my dream come true case: the universe.", "section_level": 2, "section_title": "Universe Exploration" }, { "index_sentences": "That’s the end of my talk, and thanks for your attention.", "section_level": 1, "section_title": "Conclusion and Q&A" }, { "index_sentences": "This is a code I got from Rich T, and this applies to a lifelong agent in RL and also applies to humans: Never stop learning.", "section_level": 2, "section_title": "Conclusion" }, { "index_sentences": "I'm happy to take any questions.", "section_level": 2, "section_title": "Questions and Answers" }, { "index_sentences": "what is Moore's Law for LLM and where are we on the curve?", "section_level": 3, "section_title": "Q1: Moore's Law for LLM and AGI Skepticism" }, { "index_sentences": "Thank you for a great talk. Two words that came up a lot in your talk are: Efficiency often related to data efficiency, if I understood correctly.", "section_level": 3, "section_title": "Q2: Scalability, Efficiency, and Energy" }, { "index_sentences": "There was a paper published by IPO probably in the past few days, saying that the reasoning for complex tasks is where these models actually break at high complexity.", "section_level": 3, "section_title": "Q3: AI Reasoning Models and Human Reasoning" }, { "index_sentences": "Do you believe there's a general algorithm for discovery and invention?", "section_level": 3, "section_title": "Q4: Algorithm for Discovery and Invention" }, { "index_sentences": "Thank you so much. Thanks, everybody, for coming.", "section_level": 3, "section_title": "Closing Remarks" } ] }; window.faq = { "qas": [ { "answer": "The speaker's optimism increased because navigating high-dimensional space is difficult to get stuck in a local minimum, the local minimums encountered are often flat and near the global minimum, and deep learning models tend to learn patterns first before noise, even with more parameters than data points, due to how stochastic gradient descent works.", "index_of_source": "I was talking to a statistics professor yesterday, and I realized a lot of the things I learned in graduate school about statistics had wrong intuitions.", "question": "Why is the speaker optimistic about AGI despite potential issues like non-convex optimization and overfitting that troubled him initially?" }, { "answer": "The bitter lesson over the last 70 years is that algorithms that effectively leverage compute ultimately outperform those that rely heavily on human engineering inductive biases, as seen across image, NLP, and other domains.", "index_of_source": "The bitter lesson over the last 70 years is important.", "question": "What is the \"bitter lesson\" in AI development over the last 70 years?" }, { "answer": "One hypothesis is that scaling reflects the structure of the data distribution, which often follows a power law. Discovering rarer, higher-intelligence structures in the data requires significantly more compute, causing capabilities to appear suddenly when sufficient scaling is achieved, rather than smoothly.", "index_of_source": "One is that scaling actually reflects the structure of the data distribution.", "question": "What is one hypothesis for why the scaling law holds and seemingly emergent abilities appear suddenly in AI models?" }, { "answer": "From an information theoretical perspective, intelligence can be seen as the ability to reduce unpredictability (surprise). By predicting the next word, LLMs are compressing patterns to become less surprised, which is a form of intelligence. From cognitive science, the human brain constantly compresses sensory input and universe information (like physics laws), so learning is compression.", "index_of_source": "If you can completely predict something, there's no new information to you, right?", "question": "How does predicting the next word in language models relate to understanding and intelligence according to the speaker?" }, { "answer": "Despite impressive performance, reinforcement learning achievements in games initially had limited direct economic impact because they were very specialized systems (task-specific superintelligence) designed for environments that were not economically valuable in a generalizable way.", "index_of_source": "The question is though, why did this impressive reinforcement learning treatment have limited direct impact on productivity in everyday life?", "question": "Why did impressive reinforcement learning achievements in games (like AlphaGo or OpenAI Five) initially have limited direct impact on productivity in everyday life?" }, { "answer": "To fundamentally improve AI capabilities, the main bottleneck is data. This means needing more, better, or higher-intelligence data for models to learn effectively through the scaling law.", "index_of_source": "To fundamentally improve the capabilities, what we need to fundamentally improve is we need more or better or higher intelligence data so that the model can learn it through the scaling law.", "question": "What is the main bottleneck preventing fundamental improvement in AI capabilities today, according to the speaker?" }, { "answer": "The speaker believes current AI models are significantly less data-efficient than humans because humans learn at a higher, more abstract level than merely predicting the next token. Predicting next tokens involves capturing a lot of random or low-level structure, wasting computational resources compared to human learning which focuses on the essence of the problem.", "index_of_source": "If you think about the current AI, the data efficiency is still very bad compared to humans.", "question": "Why does the speaker believe that current AI models are significantly less data-efficient compared to human learning?" }, { "answer": "One major reason for the speaker's increased optimism about AGI over the last few years is constantly observing that things previously considered impossible are becoming realities every couple of months, leading to doubt about the validity of perceived impossibilities.", "index_of_source": "The other thing that made me a GIP over the last two, three years is you constantly observe, impossibilities become realities.", "question": "What is one major reason the speaker's optimism about achieving AGI increased over the last two to three years?" } ] };

2025/6/29
articleCard.readMore

Mamdani, Trump and the End of the Old Politics

Mamdani, Trump and the End of the Old Politics The Democratic primary that just wrapped up in New York was a collision between two very different candidates on almost every level. Ideologically, outsider versus insider, name recognition. But it was also a collision in a way that I think matters for much beyond New York City politics of two very different theories of attention. Andrew Cuomo ran a campaign that was based on a tried-and-true strategy of buying attention. He had this gigantic super PAC with tens of millions of dollars purchasing all the advertising money he can buy, absolutely dominating airwaves with negative ads about Zoran Mamdani. In his own words, “Zoran Mamdani wants to defund the police.” Zoran Mamdani is a 33-year-old dangerously inexperienced legislator who’s passed just three bills. Zoran Mamdani, a risk New York can’t afford. Paid for by Fix the City. And then you had Mamdani, who was running a campaign on a very different theory of attention—a theory of viral attention. A campaign built on these vertical videos that if you opened Instagram, if you opened TikTok, and you were in any way connected to his ideas or to New York City, this was all you saw. So what’s your take? “I should be the mayor. New York is suffering from a crisis, and it’s called halal-flation.” Did you know that Andrew Cuomo gutted the pensions of hundreds of thousands of New Yorkers? Mr. Cuomo, and furthermore, the name is Mamdani. M-A-M-D-A-N-I. You should learn how to say it. Attention works differently now. This is one of the core political theses of this entire podcast. It is laced through so many of these episodes. You just watched these two incredibly different attentional strategies collide, and Cuomo got flattened. He got flattened. It was not close. There are things you cannot learn about how to win elections in other places from an off-year June Democratic primary in New York City using ranked-choice voting. But there are things you can learn about how attention works right now, and that’s in large part the subject of this conversation. Now, I’m not a New Yorker, but I want somebody who is a New Yorker, who has deep roots here, and who really understands political attention. So I asked my friend Chris Hayes, an MSNBC anchor and the author of a phenomenal book on attention to politics, The Siren’s Call, to join me. As always, my email is reclineshow at nytimes.com. Chris Hayes, welcome back to the show. “It’s great to be back.” So, Zoran Mamdani won the primary. He sure did. You just wrote a book about political attention, and this was one of the most attentionally remarkable and innovative campaigns I’ve seen. So I want to hear the Siren’s Call analysis of the Zoran Mamdani campaign. So the first thing I would say about him is he genuinely came from nowhere. I live in New York City and spend between 16 and 20 hours a day reading about and thinking about politics. And, like, I knew there was a Democratic Socialist Assemblyman named Zoran Mamdani. Right. I didn’t even know he was running for mayor until he popped up in my Instagram feed or TikTok. Right. So at one level, just to level set here, this is someone who had zero attention on him, who went from having zero attention to monopolizing attention in the race. I think the way he did it was through viral videos. It’s the first time I’ve seen a Democratic candidate be totally native to the medium of our time, which is short vertical video in the algorithmic feed. I want to play one of them here. This is one of the first times he came across my radar, which was this video he did right after the 2024 election. Hillside Avenue in Queens and Fordham Road in the Bronx are two areas that saw the biggest shift towards Trump in last week's election. Even more residents didn't vote at all. Most of these people are working families. They're working one to two, three jobs. And rent is expensive. Foods are going up. Utility bills are up. And that's your hope, to see a little bit more of an affordable life? Absolutely. You know, Gaza, who should I vote? Either side will go ahead, send bombs from here to kill my brothers and sisters. You know, we have a mayor's race coming up next year. And if there was a candidate talking about freezing the rent, making buses free, making universal child care a reality, are those things that you'd support? Absolutely. He'd have my vote all day. We need child care that is affordable. Buses should be free. The hike in the MetroCards is totally unaffordable. My name is Zoran Mamdani. I’m going to be running for mayor next year. Wow. Yes. Yes, sir. And I’m going to be running on that platform. Thank you. I’m going to vote for you. Your energy is… Thank you. Thank you. What struck me about that video when I saw it was so many politicians do communication in terms of what they are telling you. A lot of what was fascinating about Mamdani’s campaign was he turned the act of listening into a form of broadcasting. That’s exactly what I found so striking about it. When I first saw the video, I didn’t know he… until you get to the end when he’s like, “I’m running for mayor.” I was like, oh. There’s two things about it. One is, the whole point is he’s listening to people. And two, that is a very recognizable trope of this form of video. The guy on the street, like the infamous Hakitua girl, is because there’s a guy walking around Broadway in Nashville, sticking microphones in people’s faces. This is an established genre. So he’s taking this established genre that has its own kind of features and is familiar. And then he’s doing this really innovative thing. I, as the politician, am not going to speak at you. I’m just going to put mics in people’s faces and ask them questions. It’s incredibly effective. He is the first politician I have seen be native to the thing that is after what I think we think of as social media. Yes. So, there are a lot of politicians. Donald Trump is one of them, Bernie Sanders is another, who, in a way, they were very dominant on Twitter, on Facebook, on a kind of mostly text-based, high-engagement, social-sharing era of media. And the thing that’s come after it with TikTok on Instagram, you see it now more on X, too, is much more algorithmic, right? You can come out of nowhere much easier and very visual, vertical video, not primarily text-based. Zoran was not dominant as a figure in text on X. No. It was videos. It was visuals. It was beautiful. The graphic design in that campaign was beautiful. Yeah, there’s a great New York Magazine piece about this. And always in a suit, right? So, a highly recognizable outfit. I mean, he was very visual; there was an incredibly consistent visual grammar. Totally. Right? There were very certain filters on most of his videos. And then when he would do, like, videos about more intense subjects like ICE, they would take those filters off. Yes. Or make a starker one, right? His, I mean, his mother is, right, like, an amazing filmmaker. Exactly. His sense of film and visual grammar was very, very, very strong. The last time I think I saw something like it would be Howard Dean with Meetup back in 2004 or Barack Obama with Facebook in 2008, right? Or Trump on Twitter. Trump was truly native to what Twitter is. Yes, you’re right. That’s a great point. Conflict. I’m thinking of Democratic candidates. But, yes, Donald Trump and Twitter in 2015 and the way that that, his performance on Twitter became the way that people, a lot of people came to know him, right, as a politician. One point I want to make here that I think is important, I think we both agree on, is that with all these discussions, there’s stuff that’s new and there’s stuff that’s timeless, right? The guy is very charismatic. He is very politically talented. That would be true if he was running in the 1950s. It would be true. You know, if he was doing whistle-stop tours, like, the guy can talk. He is a very talented communicator. So, I don’t want to overstate the degree to which the medium is determinative. You could make short-form videos, and they wouldn’t work as well unless you… he’s got Riz, like he just does. The thing that’s so wild about it, though, is that there’s a perfect pairing between that charisma, that way of communicating, with the form that he used, and then the fact that the algorithmic social media means a thing can blow up. And I don’t think you can even talk about the Mamdani one without also, like, what his foil was. Andrew Cuomo and Zoran Mamdani were perfect foils for each other. You could not have scripted it better. And Cuomo had this gigantic super PAC behind him. There was this real sense, I mean, correctly so, from any sort of normal rules of politics, that how is Mamdani or anyone else going to climb uphill against the amount of attentional artillery that that super PAC could and would buy? And we know that they were just absolutely dominating the airwaves 24-7, basically. I cannot overstate to people outside the New York viewing area how insane the same ad. You know when I saw this ad was, I saw this ad one time. I mean, I saw it like 17 times in this one experience. Right, yes, yeah. Because I was at a bar and they had a TV on. Exact same. One of the things that struck me the whole way through on the Andrew Cuomo campaign was how old its understanding of communication was. And the idea, at some point, I would watch people talking about Cuomo as a juggernaut. Intentionally, in my world, Mamdani was a juggernaut. He didn’t exist. Cuomo didn’t exist. In fact, I think this—he was hiding from it, by the way, too. But he didn’t exist. Well, that’s another thing we can get to is the sort of what Mamdani was doing on social media through things he was creating. And then there was what he was doing on other media outlets, which was also the opposite of Cuomo. Yes, very much. So, on the first point, to take a step back, people really have to understand that for probably, I’d say, the last 40 years, there’s this formula for how—and I think it’s true for both parties, but I know Democratic politics better. You raise a lot of money and then you spend it on TV buys. That’s what a campaign is: raise a lot of money, spend it on TV buys. That is how they choose candidates. Is, can you raise the money so that you can do the TV buys? The DSCC and the DCCC, who recruit congressional candidates, the senatorial candidates, one of the main things they are testing is, can you raise the money? Yes. And what are you doing to raise the money? You are buying attention. What you’re doing is buying attention through 30-second ads that are going to run on the local news in the three weeks before the election. Yes. That is 90% of the campaign. The last 10% is, yes, you go to events and you shake hands. I mean, maybe it’s 80%. I’m sort of overstating a little bit. But you saw Cuomo just run this play, which was limit media availabilities, only pick your spots, be confident that this enormous carpet bombing is going to happen late down the stretch. And it totally backfired and didn’t work. I really want to hold on this for a minute because you cannot buy attention now the way you once could. Exactly right. You can only earn it. This goes back to the conversation we had right after the 2024 election because, I mean, that was also a period for all that Donald Trump really did have a lot of money behind him in that election. Kamala Harris had more. Yeah. She raised a ton of money. They spent a ton of money. And they absolutely did not dominate attention. You were almost watching between Cuomo and Mamdani an almost pitch-perfect version of the old attentional strategy versus a pitch-perfect version of the most modern native attentional strategy collide. I do think the underlying product here matters. Cuomo was just a bad product. He was a scandal-ridden, high-negatives, very widely disliked former governor who had to resign in disgrace, running against this sort of fresh-faced figure. But it also was a real collision of these strategies in a way that I do think people should watch. If I’m the DSCC or the CCCC, I would start thinking not about who do I think can raise money, but who can raise attention. Themselves by being out there on all these platforms and actually creating things that are native to the places they’re running in, which will be different if you’re an Ohio Senate candidate, or a Wisconsin Senate candidate than if you’re a New York City mayor or primary candidate. But Wisconsin, Ohio, Missouri, and all these places like Kansas, they have their own things that people care about and their own cultures. And they also, just to be clear, how else are people getting information now? I mean, look, above a certain age and among certain demographics, people still sort of consume the news as the news in whatever form that takes. More and more voters, and particularly voters who are in that outer concentric circle of political or news interest, that Democrats lost by 15 points in 2024, that Democrats have struggled to win, that you have to win if you’re going to win Ohio. Those folks, how else are they going to know about you? They’re not, if they’re not watching the evening news when you’re buying your ad points and they’re not watching network news and they’re not watching linear cable. Literally, how do they find out about you? They’re going to find out about you from their phones. So, how do you get to them? I mean, you really have to think through this. How will this person know that I’m running, what my face is, what I look like, what I stand for? How will they know? If you don’t have a theory for that, that’s other than “well, we bought a bunch of points on TV,” you’re cooked. It’s not going to work. We did this show a couple of months ago about the tension. It was after the election, and that particular show got very wide distribution among Democratic politicians. I’m sure you heard this too. And then so some of them would come to talk to me later, and they were trying to do video. I have just thought a lot since then about why their videos are so bad. Members of the Senate Democrats, and for that matter, the House Democrats, they have a lot of money in their campaign committees. They have a lot of money for communications. They could hire very, very good people. And it’s actually not the case that you can’t make an argument about, you know, the big, beautiful bill or something go viral. I know you can because I do it. You know you can because you do it. I just look at what all of their content looks like. I think, does nobody there have a sense of what they like to watch? Because definitely they don’t like to watch this. But the absence of taste among people who are, in theory, skilled political communicators is weird to me. Okay, I’m going to, here’s a structural answer to that question, which I don’t hold me to, but here’s a hypothesis. Democratic Party politics are really complicated politics of multiracial, multiethnic, multilingual coalitions. I think often the things that success in Democratic politics selects for is skill at managing these coalitional tensions, which is a really difficult thing to do. Hakeem Jeffries is very good at that. Nancy Pelosi is the best at it. No one, and I think including Nancy Pelosi, would be like, “I want to listen to a Nancy Pelosi podcast.” Nancy Pelosi is not a great public communicator. She is a legendary, all-time great manager of coalitional tension. I think the coalitional politics of Democratic politics select for people who are very skilled at managing these very different, difficult coalitional issues. That is a different skill than public communication to the normies. Okay, but let me push on this a little bit. I think you’re right about Hakeem Jeffries here. A Chuck Schumer, right? Absolutely. But you think about a Cory Booker. Yeah, he is quite skilled. You think about a Chris Murphy. Yeah. There are high-level… Why can’t they do… Yeah, they are out there. And Chris Murphy walks across Connecticut every year. Yeah, he does that too. Cory Booker did the 25-hour filibuster. Right. Or not quite filibuster, but long speech. There is a dimension where I know they want to communicate. I know they want what they’re saying to break through. They are willing to say things. I mean, Chris Murphy has been very out there on the level of alarm he is raising. They’re good podcast guests, right? If you were to rank Senate Democrats on how good they are on a podcast, Murphy and Booker would be high up there. Yeah, definitely. But I guess the thing I am saying is that the amount of agita I have heard Democrats express about the lack of a liberal Joe Rogan, whatever it might be, as opposed to understanding attention as not something other people gift to you, but something you earn yourself or you look for as a skill in other people, or you have some other kind of filmmaker coach you in. It’s just, the gap is so much wider than it seems like it needs to be at this point. And like watching all these people just get flattened by someone like Momdani, it really speaks to it. Yeah. I mean, part of the question here, though, right, is about being native to new forms. Yeah. Like I have made a few TikTok videos and like they’re not that good. Yes. And I think. I mean, yeah, I’ve not seen your TikTok. But I think I’m a pretty skilled public communicator. Like this is what I do for a living. It’s what I’ve done for a long time. There are these weird, you know, we talked about sort of grammar, or like there are these sort of differences of different mediums, formats, visual grammars in different times that I also think here’s actually a key thing. I think you have to be a consumer to be a producer. And I think this is a huge gap. I really think this is a real problem. Now, if I started to get serious about making TikTok videos where I talk to the camera, having watched a lot more, I would be better now. If I practiced, I’d get better. But the sort of textural sense Momdani has for the format, you can’t just read some packet or just jump in from nowhere. That seems like a thing where you should be looking for certain kinds of talent. Yeah, that I agree with. Right. There’s a reality that a lot of people who run for office are news anchors. Yeah. Mike Pence had been a talk radio host. Yep. Carrie Blake, right, had been a news anchor, right? Like a lot of these people have experience in front of a camera. And I just think you’re going to start if both parties were smart, they would be looking for people who have attentional skill. So one thing we saw here is that, yes, Momdani was trying to make this election about affordability, about material concerns. But Cuomo won the precincts where median income was under $50,000. What did you make of the somewhat strange structure of the coalitions? I don’t really have a good theory on it yet. The one piece of election analysis that has stuck out the most to me is this triangle that breaks down precincts by their degree of racial integration. Have you seen this triangle? It’s so fascinating. So basically, it breaks down precincts by how white they are, how black they are, or how other they are. This is by census. So these are not the racial categories that I would use to describe you. But basically, what it finds is that the precincts that are basically all black and then the precincts that are all white were Cuomo precincts. And the more mixed a neighborhood was in its racial makeup, the better Momdani did, which I find to be a fascinating result. Now, that might just be a proxy for the— Yeah, it might cross-correlate. Between the income stuff you’re talking about? I mean, I think I understand. My mom and I were talking about this because she was, my mom was talking about the Bronx, and the Bronx was like a Cuomo borough, which is sort of ironic because if you go back to the whole opening bid of Momdani, which is like, “I’m here in the Bronx, in Fordham Road, in this place that swung, I’m talking to people, I’m going to address your concerns.” And then he ran up the numbers in the DSA precincts, but he couldn’t have won unless he made it outside those perimeters. I think, look, I think name recognition is part of it. I think “the devil you know” or familiarity matters to voters often on the kind of periphery of an electorate in a Democratic primary, but I don’t have a good theory of why it was the case. Like if it was, there are other patchworks that I could sort of theorize better than those. What do you think? I don’t know either. I mean, I think you can come up with a couple of arguments. One is that maybe that’s cross-correlating something that’s just informational. Those voters were less attached to the discourse, not telling the algorithm they wanted to see a bunch of Sauron Momdani videos. They sort of know who Andrew Cuomo is. And they’re more mobilized by interest groups that used to be more powerful, but that were largely, like the interest groups largely signed out with Cuomo, the unions, churches, right? Cuomo did a lot of his campaigning among black churches. So you might be seeing something that has to do with almost machine politics and mobilization politics, which Cuomo was leaning on very heavily. There’s also a crime and disorder question here, right? So if you’re a voter making $35,000 a year, living in NYCHA housing, you are much more exposed to crime and disorder than, you know, a voter in Williamsburg making $137,000. Adams won running against crime and disorder, running up the totals among, you know, working-class voters. So we know that that politics is powerful. I have the sort of view that Momdani could only have won in a time when crime had actually gone down quite a lot, as it has. Because if this really was a big crime and disorder election, I think that that would have been a big problem for him. And he wasn’t well trusted on those issues. Another is that this is a consistent thing we see in the data with left-wing candidates. So I think you could just say this is something we’ve seen happening a lot. I mean, Donald Trump also won voters under $50,000. So there are different things happening as you move up the income scale where people are voting much more expressively, even though Momdani tried desperately hard to run the most materialist campaign possible. But politics is very expressive. There’s not a bad thing about it; it’s just a reality. And I voted against my material interest in this mayoral election. As did I. So everyone gets to do that. This podcast is supported by RIT. Who asks, who does your laundry? Simplify your life by having RIT do it for you. With one-touch in-app scheduling, pickup and delivery are effortless. Your clothes come back fresh, folded, and ready to wear. Handled by laundry experts who get every detail right. RIT makes laundry and dry cleaning the easiest part of your week. Sign up at RIT.com and save $20 on your first order. So I think you can cut politicians into these two categories. They’re the politicians for whom you can identify a policy that stands for them immediately: “Build the wall” for Donald Trump. “Medicare for all” for Bernie Sanders. “The Green New Deal” for AOC. Momdani had like four or five, right? There was freezer rent, there were free buses, there was free daycare, it was publicly owned grocery stores. All these are actual policies, and they’re worth talking about, but what they are is memetic. Yeah, totally. So Hillary Clinton running against Bernie Sanders had 70 policies. But none that actually defined her. Kamala Harris, I cannot give you the policy that stands for Kamala Harris. The same is true for Brad Lander and a bunch of the other people in this campaign. Which is not to say they didn’t have them. They had them. They had. Brad Lander had a depth of policy on his campaign website in this mayoral race that I only associate with presidential campaigns. It was so detailed, and a lot of them are great, right? Yes, Brad Lander was my choice in the campaign. But I said this when I wrote this piece about him, that there are politicians who communicate about policy. And there are politicians who use policy to communicate. One problem with a lot of establishment politicians is they communicate about policy. The people who thrive right now on the attentional networks use policy to communicate. You can lament that what modern media is doing is flattening policy down to this sort of bumper sticker level of memetic communication. And I kind of do lament it. But it’s also true. Like abundance has been a big deal. But it’s the word. And then there’s all the stuff behind it. And that’s a much more complicated set of conversations. But it cuts through. But if you don’t have the memetic tip of the spear, yes. I mean, there’s a question here that I think is interesting in terms of replicability—like how much that ability is structurally producible and how much is just like telling someone to dunk a basketball. You know what I mean? Like certain people have talents for things. Yes. Right? There is a question here to me about how much it comes down to talent. People have instincts and knacks for this. But you’re absolutely correct about this. And I think to go back to that video, like there is this kind of one plus one equals two thing happening there. He goes up to Fordham Road in the Bronx area. I know well. It’s like right by where my mom grew up. In fact, I was just having lunch around there for Father’s Day. He asked people, and they’re like, “groceries cost too much.” And then at the end, it’s like, “we’re going to try public grocery stores.” Now, to be clear, the grocery business runs at margins of like one to three percent. It’s not private profit that’s making the price of groceries more. I’m not convinced that the solution is going to solve the problem, particularly in this case, which I think is sort of the most dubious. But it’s also like, I don’t know, worth trying. And it also is an attempt to address people’s concerns. I’ve had a lot of conversations with people about publicly owned grocery stores. I basically understand this modest pilot of like five stores that he has proposed as getting caught trying on something. Yeah, right. I do think this gets to something very real. Are the only policies that can become a medic in this way, these sort of huge sweeping conflicts at their heart, they make people not like them at the same time they make people like them—build the wall, Medicare for all, ongoing rent freeze. Can policy be memetic? Can it be communicative? And be good? I don’t just mean be good because I’m not like, I think it would be great. Like if you can pay for free daycare, terrific, right? I think we should have free daycare. So I don’t want to just create a good, bad division here. Like all good policy is complicated. And, you know, that’s not my belief. But there is a way in which to survive memetic products have to be simple. Memes are simple. The thing behind the meme might be complicated and good or bad or whatever. But for something to get energy, I think it has to be easily rememberable. I think it has to be big. Yep. It has to activate something people care about. And it probably has to be controversial. Medicare for all dominated. People forget this now. Every 2020 Democratic primary debate, I remember, was just a lot of Medicare for all debate. Anybody who knew anything about what kind of Congress that Democrat was going to be facing, no matter who won the primary, knew we were not going to get Medicare for all. Faz Shakir, Bernie Sanders campaign manager, was on my show earlier this year or maybe late last year, saying like we would have gotten as close as we can get. But we basically would have expanded the age range of Medicare. Right? Right. Everybody knew it. But the reason that it could dominate so much was it unleashed controversial energy. Yep. There was a debate. Would you abolish all private health insurance? Were you willing to raise taxes on middle-class Americans to fund this? It was intentionally salient because conflict is intentionally salient. Exactly. A lot of policy is built for compromise. Yeah. Right. But memes are not built for compromise. We actually, I think we have a good tangible example in recent history in exactly this context from the mayor that Zoran Mandani says was the best mayor of his life; that got the New York Times very mad at him, which was Bill de Blasio’s universal pre-K. As a non-New Yorker, Bill de Blasio sure seemed like a perfectly good mayor to me. My kids are in 3K. I’m a de Blasio supporter. So let’s talk about universal pre-K for a second. De Blasio’s problems are not bad. Universal pre-K did have that memetic energy. It’s simple and straightforward. Every kid in the city has to go to kindergarten. We’re going to make a new grade below it. This is informed by real empirical work that’s been done, and we’re going to have a tax structure that funds it and makes it happen. It was controversial at the time. There were lots of people who said this was a bad idea. “You’re going to put local daycares out of business.” I mean, there was conflictual energy around it. And then they delivered it. I sent my first kid to it; it was year two maybe that it was up and running. I walked into this school that had been leased by the Department of Education that had formerly, I think, been a big Catholic school. They were like, “This is like one of the biggest pre-Ks in the whole city.” It was like 20 classes. I was like, “This is the most extraordinary accomplishment I’ve ever, like, I can’t believe you guys stood this thing up and that my kid’s going here for free and comes out every day.” So, that’s an example. I just want to give an example of everything that you said. It was memetic policy. It cut through. It identified Bill de Blasio. It was one of the hugest things they got into power. They actually did it. It actually worked. That is an example of all those things happening. And yet it didn’t stop everybody from turning on Bill de Blasio. Right. Because then it’s like, “What have you done for me lately?” Also, if you don’t have a, here’s the thing about that promise, I will say. If you don’t have a kid that age, great. It’s highly salient to me. I have a three-year-old. Yes, yeah. For me, I was like, “This is awesome.” I see a lot of people like on Twitter celebrating Mamdani’s win. And I think Mamdani’s win is exciting. But I’ve said this before, the downside for him was not that he loses a primary. Like, the bad outcome is that he wins and fails at governing. He cannot get the tax increases he needs from Albany. He does not control the tax increases he needs for this agenda. And Kathy Hochul has said, “No.” She has already made a no, like, raising taxes like this pledge. And she’s not going to break it. So he’s not going to have the money he needs. An extended rent freeze—I know people who do nonprofit housing. Same. And there are people who are ideologically aligned with Mamdani, and they do not think this is a good idea. Yeah. I know people in nonprofit housing who feel the same way. The reason is, over, you know, you do it for one year. Okay, fine. But over an extended period of time, you will reduce the incentive to build that housing. You will reduce the incentive to care for that housing. He’s like, Mamdani will say, “Oh, you have these other programs. You can apply for relief.” All that stuff is complicated. And you make a market less profitable to be in, and fewer people will be in it. A lot of the things, like free daycare, he probably just can’t pay for. So if you set up these expectations… Yeah, that’s the big question. And then you don’t meet them. Is it okay because your supporters know you tried? Or is it a kind of like a structural thing where you have set yourself up for failure? I think it’s the most important question in some ways. I mean, one thing I would say is I like experimentation and new ideas. So when he was asked about the public groceries, I think it’s in the Bulwark podcast. And he says, “We’ll try it. And if it doesn’t work, c’est la vie.” Yeah. And like, I love that answer. Politicians never give that answer. They never give that answer. Like, “Let’s try.” You know, the person who really most embodied that spirit is FDR. If you go back and you read about, you know, the first hundred days, they’re just trying a lot. Like, we now think about FDR as this colossus who remade the relationship between the citizen and the federal government, right? A lot of that stuff did not work. It fully failed. A lot of the interventions failed. They did a lot of clunky stuff. Now, it’s a totally different time. He had this enormous mandate. It was a crisis. But I will say that I like the idea of experimentation. I like the idea of these ideas coming from outside of what the consensus around sensible policy is. But the test for it is, can you deliver? One thing that struck me a lot about Mamdani was his ability to listen. To sense a zeitgeist, but also to listen to voters, right? The relentless focus on affordability. That was an act of listening. Totally. And then being able to respond to it. And it’s been one of my views for a while. It’s actually the introduction of my book: “We have moved into an era of politics that is going to be all about affordability.” Housing inflation. Cost of child care inflation. Cost of health care inflation. It’s actually moderated in some ways, but it’s still quite bad. Educational pricing, right? For four-year colleges, that kind of thing. That had been building for decades. That is not a thing that happened in 2022 and 2023. That had been building for decades. And now, you know, things are, like, they kind of rise and, like, they’re an issue. And then, like, they’re actually intolerable. Yeah, right. And so, future politicians were going to have to develop a set of ideas and a way of talking about bringing costs down, not just bringing subsidies up. Whether Mamdani’s particular policies will work to do that was the focus. That struck me as a politician native to this era of concerns. I mean, think about the rent freeze, right? He wasn’t saying, “We’re going to give rent rebates through a tax filing where you file a tax and we’ll give you $150 back.” It was like, no, we’re just going to cap the price. The concern is whether or not, from a policy perspective, my concern with Mamdani. Mamdani talks a very, very good game. I think he gets that you need housing supply. But his plans are all public housing, which is fine, but that’s much harder. When he talks about market rate housing, he sort of is like, “I really believe in market rate housing as long as it accords to our sustainability, union, and affordability needs.” Right, yeah. And it’s like, when you need a lot of housing, adding a lot of conditions to that housing is going to both, like, raise the price. So I really think there’s a question about whether or not he can deliver affordability if he’s not able to increase supply. I would feel better about a rent freeze that was paired with an incredible explosion of building. If what was happening was like, we were freezing rents and there were cranes everywhere. Right, yeah. Okay, fine. Because maybe in three years, we have a lot of housing coming online. But if you, at this level of supply creation, freeze rent for an extended period of time, you might begin to, like, constrict supply down the road and create a bigger problem for the future. There are some levers we could pull on this. Housing is a particularly tough one because it takes time to build houses, and we make it hard to build houses. I’m very skeptical that Mamdani can make free daycare happen. I don’t think he’s got the money to do it. There’s more infrastructure that would be needed than was required even for a $3K initiative. Yeah. But you could conceptually do free daycare. You could definitely do it nationally. There are ways to approach some of these things, but I think this is what politics economically is going to be about for an extended period. I think one wrinkle to the housing question, which I think is a really important thing to always keep coming back to when you discuss in your book, is that “one person’s price is another person’s income.” There is a real, genuine material conflict in New York City between renters and homeowners. It’s not false consciousness. It’s not a distraction. It’s not culture war nonsense. Like, if you own a home and most of your wealth is in your home, you want to see that wealth go up. If you are trying to enter the housing market or are a renter, rising house prices are bad for you. You will not be excited about Mamdani or anyone coming in saying, “We’re going to build a ton of public housing next to you.” Like, that’s the other thing that’s very difficult about public housing and affordable housing. All these homeowners who want their high home prices do not want that down the block from them. That material fight, you know, has been won by homeowners in California, who have been beating down people trying to buy homes and renters for decades now to a degree that’s truly catastrophic. I think it’s fair to say I do worry that the structural nature of public opinion now is negative in a way that makes even good governance not resonate with people, if that makes sense, or the structural limitations on governing. One of the two. It’s just very hard because of how many things contribute to a working-class person who lives on Fordham Road feeling squeezed in every direction. Can Zora Mamdani unilaterally make it so they don’t feel that way? It’s hard to say. Can they feel that I got a mayor who’s trying to make my life better? Yes. So translating this kind of communication from campaign to governance, not that many people have had to do it, but Obama had to do it. And I think I would say he failed to do that. I think the sense is that he was an amazing, amazing, amazing campaigner. And then given the reality of incremental victory, he was never sort of able to narrativize that in a way that could ease the disappointment a lot of people felt. I think that’s in some ways why the liberalism he represented after him, for at least some time, had a hard time because he had raised hopes so high for a lot of people. And then it’s like, “eh, you know,” I mean, things did change. I’m a big fan of Barack Obama. The Affordable Care Act is a huge and ongoing achievement. But how do you narrativize the difference between people’s hopes for your campaign and what they got? Donald Trump is interesting because he comes after Obama. He also makes huge, sweeping, wild promises. They never built the wall. Obama never did, right? They never built the wall. They don’t build the wall. But Donald Trump has this way of communicating throughout his entire presidency. I mean, he loses re-election, right? So it doesn’t work exactly. But that he is, it’s like somehow he’s a president, but he’s not responsible for what happens. No, he’s at war with his own government. He’s like the deep state. So there was a narrative that Donald Trump maintained as president that allowed him to explain away the difference between what he attempted and what he achieved. And now Trump is president again, and he has much more control over the government. So he’s not, it’s not as much of a deep state narrative this time. Although he has spent the last 24 hours railing against the intelligence apparatus. Yes, exactly. Like it’s very classic. So yes, because they say that the Iranian strikes only set it back by a couple of months, and he’s saying it’s false. So there’s one, is like, can you use it as a form of power? But then can you use it if you’re not being able to get it done, right? Can you narrativize the grimy, gritty, just reality of governing in a way that maintains the faith people have in you, even as you’re not being able to deliver to them what you promised? It’s, I think, there’s a few things I’d say about that. One, I think mayor is different than president in a lot of ways, partly because it is much more retail. And you can get a long way by showing up a lot. I mean, Eric Adams actually does that pretty well. And, you know, I thought the—there’s a club opening. What’s that? He’s going to be there. You know, and this is, you know, this is Chuck Schumer’s legendary talent, not as mayor, as senator, but before that as congressman. There is a little bit of a trap that is difficult to avoid, which is like, it will be more difficult to govern than it is to campaign always. Andrew Cuomo’s father quite famously said, “we campaign in poetry and we govern in prose.” I think that part of the way, I guess, that you escape that trap is talented political communication. I mean, I really do. Like, I think you have to do a good job. Like, you can’t be a total failure as a mayor, right? Like, the city has to feel like there’s tangible improvements in people’s lives, but that alone won’t be enough. You basically need both. You know, I thought the Mamdani video to close out the campaign where he walks the length of Manhattan and he’s just like talking to people, dabbing people up, eating a slice of pizza, drinking water. Like, you have to keep doing that, I think, to be an effective mayor. And I think that does actually allow you to narrativize. Yeah. Because it’s like, I’m out here in the streets and I’m talking to people and I’m hearing what you’re saying about what you’re trying to do and I’m communicating to you about what we’re trying to do. The getting caught trying, I think, is sort of the key part of that. I think this is something you’re seeing with Donald Trump right now, which is he actually has an instinct for how to turn policy that isn’t affecting that many people into something that is intentionally salient. Yeah. Which is to make it a performance. Yeah. He performs everything. Including war. Including war, the deportations, the sending people to foreign prisons and having Christine Noem, like, pose at them in her flak jacket. That there’s a way that he feels to me— I mean, he’s a genuine attentional innovator. Say what you will about Donald Trump. Yeah. And that he is trying to make much more of policy into a public performance. I mean, there is a reason. I mean, Dr. Phil is embedded with the ICE teams. Dr. Phil is embedded with the ICE teams. His cabinet is full of people from TV, be they reality TV stars from one period, like Sean Duffy, all the way over to the Secretary of Defense, Pete Hegseth, who’s a weekend cable news host. So there is this way in which I think Trump has been trying to sort of square this. Most people will not feel the effect of most of his policies. But what if he can turn those policies into programming? Yes. But here’s the irony, right? He’s at 10 points underwater and all the stuff is pulling at exactly what you would predict from thermostatic public opinion. From the use of the bully pulpit, David Shore had a thing the other day about one of the most consistent counterintuitive findings: when a president talks about something, its negatives go up. The sort of negative bully pulpit. Now, the question to me is, and this feels very unresolved because of how sui generis Trump is and how sui generis his trajectory has been, does it net out as a positive? The question of attentional domination, he does it better than anyone. He is a genuine innovator and a weird genius for attention at a pathological and feral level that is not replicable. But the constant show, the constant conflict, like his negatives are high. He lost re-election. He stuck around. He won. He almost immediately started to tank in the polls. He’s a very polarizing figure. It works at some level. There’s some power to it. But how much does it work still remains unclear to me. I think that’s right. What it works to do is set narrative. That is its own dimension of power. It is a kind of power that he exerts in a way few presidents do over culture. And I’d say this is true for Mamdani. Mamdani as a discourse object. Trump is a discourse object, right? It’s not like Zoran Mamdani is the only person to have recently won a Democratic primary anywhere in the country. In Jersey, Mikey Sherrill. Just one news house member. Just won the primary for governor. Sherrill, I think, is an incredibly impressive politician. A former Navy helicopter pilot, right? I find her very, very, very charismatic. Yeah, she’s very good. More on the moderate side of things. There was not a debate. Does every Democrat need to reckon with the victory of Sherrill in the way that right now there’s a discourse of how does every Democrat and possibly every politician, possibly every human being, need to reckon with what we just saw in this June Democratic primary in New York City? The governor, the former governor of North Carolina, Roy Cooper, who served two terms in a state that Trump has won every time he’s been on the ballot there, left with, I think, a 55-56 percent approval rating. No one’s like, “We need to find the next Roy Cooper.” That guy is an insanely effective politician in very difficult terrain and has none of these attentionally salient qualities. We talked about this last time, which is like high risk, high reward, high volatility stuff. There are tradeoffs here. I guess this is where the question you were asking a minute ago feels like it bites to me, which is: does this kind of attentional dominance net out as a positive? It can clearly win. It can clearly win primaries. It clearly can help you exert a cultural and narrative force. And an ideological force, I would say, which is important. Like, above and beyond what you would be able to do, right? AOC is not the only Democrat who has knocked off another Democrat in a primary. She’s not the only Democrat to win a House seat. She is incredibly salient as a national politician because of her ability to drive attention. On the other hand, I recently was talking to a bunch of various people in the sort of New Democrats caucus, which is the more moderate House Democrats caucus. One thing that struck me just talking to them is that a couple of them are very talented communicators, but most of them communicate in their bearing and the way they are is not flashy, aggressive ideological projects. It’s a kind of, like, “This person might coach your little league team.” Yeah, yeah. You know, and so these things work and don’t work in different places. I don’t think we have a good way of answering the question of when it is valuable to drive this kind of attention and when it is not. So here’s what I would say. I think one place where it matters is presidential politics. Yes. I think presidential politics, like, there’s just no question that it matters at that level. You need someone who is an insanely skilled communicator with an incredible appetite and instinct for attention—the kind of person who wants to go do three-hour podcast interviews. Yes. I think if you have a person who’s not that, you’re really in trouble. The other thing that I think is worth considering is the valence of incumbent versus challenger, where I actually think this sort of is interesting to think about. I think this kind of attentional dominance works better as a challenger than an incumbent. Sure. For exactly the reason we’re talking about, right? So, we’re seeing right now Donald Trump recreate some of the thermostatic public opinion on immigration that he had in the first term. This was part of what drove Democrats to adopt a line on immigration that was to the left of what their previous line had been, partly along the lines of how public opinion had changed in recoiling in horror at what Donald Trump was doing on immigration. So my point being here is that there are more upsides to downsides of the challenger for this high volatility, high risk, high reward attentional trade than there are for the incumbent. I also think there’s a dimension here where they work. This is very, very, very valuable in primaries. Everything we were saying a minute ago about policy that becomes memetic is policy that unlocks a lot of attention, usually through controversy, where some people really like it and other people really hate it. What you’re hoping to do when you unleash that kind of conflict energy is that there are more people who really like the thing than really hate it. The trade that you often see some of these candidates make is they are unleashing energy in the primary that might hurt them in the general. It is often made observation about Donald Trump that he seems to underperform in the general. He is incredibly dominant at the primary level. But Trump and candidates like him who are less talented than he is, MAGA candidates, tend to underperform in the general. The view is that another, you know, I think a lot of people believe, and I’m one of them, that if Republicans had run Marco Rubio in 2016, they would have won by more. I actually think that’s true in 2024 also; if they run Nikki Haley, if they run probably even Ron DeSantis, they would have won by more. Like, the conditions were there for that. Trump creates a lot of negative attention on him in general elections. New York is weird in a lot of ways. But one is that the expectation is if you have won the Democratic primary, you have won. The fact that that is not a complete expectation with Mamdani speaks to the way that there’s at least a belief that he will generate counter-mobilization against him at a higher rate than, like, a Brad Lander would, than some of these other candidates. But it’ll probably be okay for him in New York City because, again, it’s so dominated by Democrats. This sort of thing raises the question of how do you stand out in a primary campaign in a non-representative electorate that agrees with you much more than the general electorate does. But then if you’ve done that, what do you do with these positions you’ve taken, partly if you’re dealing with a general electorate that is not all the way to your side? So I always think, like, just to finish this one example on it, in Ohio, when J.D. Vance ran for Senate, Mike DeWine, who’s like an intentionally not very skilled, kind of more older-school Republican, he was governor. He won his re-election campaign that year by like 20-ish points. Vance underperformed in the Senate race. I mean, he won, but it was by 6, 7, 8 points. It was not an amazing performance, in part because he had taken very, very MAGA positions. Now, has it worked out for J.D. Vance? Yeah. But not in the sense that J.D. Vance overperforms with general election audiences. This is where it gets complex; it’s a really uncertain trade. I think to add one wrinkle here that I think is interesting and slightly weedsy but worthwhile is that, you know, New York City has ranked choice voting. The ranked choice voting allows voters to rank five different candidates. That created some interesting incentives that are a little different in this race, which I actually think worked against part of what you’re saying there, which is like being the biggest bomb thrower is the most distinguishing. However, the way ranked choice voting works is you don’t want to alienate other people’s supporters because you want them to rank you second or third or fourth. One of the things I thought was very interesting about how Mamdani navigated this, and I think huge props here to Brad Lander, who came in third in the sort of first round of voting, was that there were all these cross endorsements and this sort of coalition building. So it wasn’t just bomb throwing; there’s a kind of politics you see, particularly in Republican primaries, where it’s like the rest of these people are sellouts and I’m the truest MAGA. There kind of wasn’t that. Mamdani wasn’t running against the Democratic establishment. There wasn’t this kind of sentiment you see among the left flank of the Democratic Party regarding these corporate sellouts. There was not very much of that directed at Cuomo, but he cross-endorsed other candidates as well. I think the reason that’s salient for the general is that, yes, it’s in a primary, but it’s also coalition building. Yes, I think that coalition building actually ends up being extremely important in general. By the way, New York City had five straight terms of a Republican mayor. Let’s not forget that. Yes, the idea that the expectation is that the Democrat wins is a fairly recent vintage. Giuliani won twice, and Bloomberg served three terms. That was 20 years in a row for a Republican mayor. I think this is an especially terrible time to be a refugee. This is an especially terrible time to be a refugee. Because after fleeing war, violence, and persecution, it seems like the world’s capacity for compassion is running short. As the numbers of refugees globally reach catastrophic levels, UNHCR, the UN Refugee Agency, needs your support. In just 72 hours, your donation will deliver shelter, protection, and food to refugees when they need it most. Your compassion today will make a life-saving difference. Go to unrefugees.org/donation to make your gift. I think some of his people will not like hearing me say this. I read Mamdani as a left pluralist, not a left populist. Yeah, I agree. People, I think, have very shifty definitions of populism. But in its classic definition, what actually makes somebody a populist politician is not that they believe in redistribution or that the working man is getting screwed a bit. It’s that they believe the system is built around true people and then the small conspiratorial enemies of the people who are keeping everybody else down. If you can just break through them and have your villains and destroy your villains, you can sort of hit the more utopic politics you’re looking for. I have seen many right populists and left populists. What struck me often about Mamdani’s affect was that it felt a bit like a TikTok affect. Because TikTok—I mean, people forget this—but it had its whole thing, and it doesn’t really work this way anymore. For a very long time, they were really pushing it to be a positive platform. Mamdani always seemed much more motivated by his sympathies than his resentments. In contrast, Cuomo felt much more motivated by his resentments than his sympathies. This also played into the RCV (Ranked Choice Voting) dynamic you’re discussing. I think it would have been natural to assume that these other more establishment, long-serving New York politicians would be likelier to cross-endorse and work with the frontrunner, the former governor. Right, who could both, in theory, give them more because he was likelier to be elected for most of the campaign. But also, someone they would have known better because he’s been in New York politics forever. To me, this was both politically meaningful and substantively meaningful because it undercut the central argument of Cuomo’s candidacy. They all hated him—not all, Jessica Ramos endorsed him, but they largely disliked him. Brad Lander really clearly dislikes Cuomo, and so do a lot of them. They did not want Cuomo ranked. This created an interesting space where the dynamics were not what you would have thought in a local insurgent versus Democratic establishment race. There’s this validation role that ends up happening from that. If you’re hearing that the guy’s this terrifying, scary figure who’s an extremist, but then the other candidates in the field are cross-endorsing with him and appearing with him, it makes it much harder for that to land. I think, again, to Mamdani’s credit, I agree with you that he does not have a kind of—I think it’s well said that he’s animated by his synthesis as opposed to his resentments. His affect is welcoming and pluralistic and also not, “they’re out to get me.” He really just does not portray that at all, which I think can be a real problem for a certain form of left populist politics. It’s a rigged system, it’s all rigged, the fix is in. He got $25 million dropped on his head by super PAC money. Bloomberg wrote a $5 million check two weeks later—there was a little bit of a rigged game against him. But he did not let that—again, if you look at that Walking the Length of Manhattan video, that’s—the effect there is welcoming and inclusive at all times. But this is where I don’t want to over-McLuhan everything, and say the medium is always a message and everybody’s shaped by their mediums, because obviously a lot of people on TikTok are in vertical video who are not, like, Zoran Mamdani or don’t even follow what I’m talking about. But I believe—I believe this strongly—that the rise of populist right and, to a lesser extent, populist left politics all across the world, all at the same time, I believe the single strongest force there was not just immigration. And it wasn’t—I mean, you can really look at this in the data. It was not economics. Right. I think it was the rise of these central communication platforms of politics being high-conflict, high-engagement, compressed-text platforms. And I think those platforms, in a way that we do not have incredibly good even language for, are somewhat illiberal in their design. By that, I mean that they are structured in a way that makes the fundamental temperament of liberalism hard to do. They’re not well-suited for deliberation. They’re not well-suited for tolerance. Right? They’re not well-suited for on the one hand, on the other hand. Right? The things that make deliberative, liberal democracy kind of function, those habits of mind, the way you hear when, like, Barack Obama—Barrack Obama’s not good at Twitter. He’s just not. Twitter’s bad. No, he’s not. Right? It’s terrible. Because they’re about groups. They’re about engagement, like, within and then against other groups. They’re about, like, drawing these lines very, very carefully. And I think they just create, by nature, a more populist form of politics, or at least they create a communicative structure of politics where it is easier for outsider populist politicians to thrive. The thing coming after it, which I don’t know if it will hold this way, but this kind of vertical—like, when you look at TikTok, when you look at Instagram reels, again, it’s not that no content is high-conflict political content, but most of it just isn’t. It’s much more like day-in-the-life stuff. It’s very highly visual. And you just kind of saw that a little bit in this campaign. I think there was something in the grammar of Mamdani that was so inflected by that era. I mean, he’s, like, really our first Vine politician. Like, people forget all this, but I think there was something there. His grammar was not Twitter’s grammar. Fun, kind of goofy, kind of like, yeah. His grammar was TikTok’s grammar. Yeah, I think that’s a really interesting point. I mean, I’m sort of thinking this through. So, I think I agree that social media, as constituted over the last decade, is structurally illiberal. I think I agree with that. Relentlessly, algorithmically competitive attention markets are going to drive towards the parts of us as ourselves that are the furthest from deliberation. Yes. Right? So, like, I have a whole chapter in the book about Lincoln-Douglas debates and, like, how different that is. Not that, you know, not that that should be the model for everything. So, I agree with that. I think I’m sort of thinking through this idea of the visual grammar and kind of, like, affect of the vertical video as being less conflict populist in its nature, which I think is a really interesting idea. I mean, one thought I had, and you just said that about Barack Obama’s bad at Twitter, is that it was funny. I watched the whole Mamdani speech, and I was like, it’s fine. He’s not great at giving a speech. Like, Barack Obama was great at giving a speech. That is not his message. There are great one-minute clips in his speeches, though. There are great one-minute clips in his speeches. But, like, his vertical video performance is a 10 of 10. His speech performance was not a 10 of 10 to me. And I think that speaks to something about the nature of that. And I think you’re right that, like, I guess the one, here’s the one counterpoint I would say. It seems to me like there are ways in which those algorithms over time, and partly this has to do with the weird black box of the algorithm, right? They do start to get more and more conflict-embracing because the clapback video and the posting of the comment of someone said something, and then you, you, like, respond to the comment, and it’s up there in a window. Like, stitching became this thing that really generates conflict. Like, here’s this, like, dumb, clueless person saying this thing, and I come in and I stitch and talk about how stupid they are. So, I do think there’s still that incentive, but I think you’re right that, overall, the vibes, directionally, in vertical video right now, are more positive than the vibes of, say, the cesspool that is X. Well, it’s also, the other thing here, just reality, is it’s more capacious. I mean, the fundamental reality of the Twitter text box is a little less true now, but it still is basically true, is that it’s a compression mechanism. And the move towards languid podcasting, where we’re just sitting here vibing for two hours, or longer, right? I was amazed; I knew this was out there, but on the Abundance Tower, I went and did some of these podcasts, like Mike Friedman and others, where it’s like, you really do three to four hours. But even in this, you can put up six-minute videos. I mean, I have videos that go out on TikTok that are six, 12 minutes. Actually, a lot can be in there. It is compressed compared to the Lincoln-Douglas debates, but it is a lot less compressed than what the original Instagram box allowed you. Yeah. Than what the dominant, for a very long time, Twitter box allowed you, than what a Facebook post offered. And then, what MomDine was doing a ton of was podcasting. Right? And then getting clipped from that. And then it gets clipped, but it does come in the context of these sort of much longer conversations that create a different vibe between people. You know, I actually find it very hard to maintain—I’ve had many people into this show because they are such harsh critics of me. And I find that they find it very hard to maintain the criticism when you’re in a sort of extended social dynamic. “It’s devious of you.” Well, it’s not—it’s actually sometimes a problem. Sometimes I have to, like, cue them. “Remember you hate, right? Like, we’re here to talk about this.” But these things, you just really see when you do that, like, how much mediums shape us all. It’s much harder to be a jerk to somebody’s face than it is under these dynamics. And so, it’s not that all vertical video is going to be sunny. Right. But it just is going to be different in ways that I’m not even sure we’re quite ready to understand in politics. Yes, I totally go with that. And I also think that, like, you know, this is—I’m just sort of spitballing here. So, I can hear already in my head the academics who study this being, “you’re totally wrong.” But let me just throw this out. We’ve got the kind of semi-apocryphal story of the 1960 debate with Nixon and Kennedy and how people who listened thought Nixon won and people that watched thought Kennedy won. And, like—if you go watch that debate, Nixon just does not look that bad to me. No, I agree. I’ve done this a few times, and Nixon looks totally fine. No, the reason I say apocryphal is that I’m not even sure it’s true. It’s sort of become this kind of mythos about how this works. And it’s capturing the central sort of McLuhan insight about how much the medium structure says. There was this kind of—there’s a sort of preliterate politics in America when you have a very small percentage of voters who can actually read. Then you have the beginnings of radio politics, and people know about the fireside chat. Television is totally transformative to American politics. The first wave of Internet politics that lasts for a very long time is written politics. It’s the politics of text. I mean, all the stuff that’s happening with blogs, when we came up, and Facebook posts and all this stuff. We are now moving—like, we’re going through this transformation where everything will be video. I mean, at least for the foreseeable future, who knows? These trends change on a dime. I think it’s interesting to consider what that does. The media strategy, too. Okay, I’m recruiting candidates. People that can get attention. Those are going to be scarier propositions. Because part of attention is sometimes conflict, provocation, views that are not boring, that jump out at you, and interviews and talking to a lot of people where you might say something that is a quote-unquote gaffe or that people don’t like or offend certain people. The institutional orientation of the Democratic Party is like, “yeah, no.” And I think there’s a great example of this with Mamdani down the stretch. Talk about his media. He went everywhere. He said yes to everything. He gave an interview to a Pakistani news channel in Urdu. Have you seen this? No. At some level, I was like, why are you doing this? This is down the stretch. This is like in the last week. But it’s like, right, maybe that gets back to Urdu-speaking New Yorkers who share the clip. Like, you know, he then also goes on mainstream. He goes on alternative. He goes on subway takes. And then he does the bulwark. Now, the bulwark is like sort of a, you know, centrist, center-right anti-Trump network. Center-left. I’m at this point. Okay, fine. It’s center-left at this point. It’s in the big— I love the boy. Tim Miller’s great, but— Yeah, it’s in the big Democratic—it’s in the anti-Trump tent very strongly. It’s strongly in the anti-Trump tent, but it is founded by people who used to be Republicans and whose feelings about, let’s say, Israel, tend more towards the right of the Democratic coalition. They ask him this question about this phrase, “globalize the intifada.” This is a very popular phrase at protests on the left. Maybe some people say that phrase with good intent, but there are certainly some people who are saying that phrase with violent intent. So I wonder what you think about that. He gives an answer that starts off with, I thought, a very long and good thing about Jewish safety and the Jewish folks that he’s talked to in New York City. Just a few weeks ago, I had a conversation with a Jewish man in Williamsburg who told me that he—the same door he would keep unlocked for decades is one that he now locks out of a fear of what could happen in his own neighborhood. Then he basically says, look, you know, intifada is Arabic for struggle, and that, in fact, the word is used in the Holocaust Museum website to mean struggle. The very word has been used by the Holocaust Museum when translating the Warsaw Ghetto Uprising into Arabic because it’s a word that means struggle. As a Muslim man who grew up post-9/11, I’m all too familiar with the way in which Arabic words can be twisted, can be distorted, can be used to justify any kind of meaning. I think that’s where it leaves me with a sense that what we need to do is focus on keeping Jewish New Yorkers safe. The question of the permissibility of language is something that I haven’t ventured into. The headline that comes out from it—I don’t think it was a great answer, to be very clear—is, “refuses to condemn globalized intifada.” So I thought to myself, I’m like, oh, okay, so now we’re seeing the cost, right? Like, we’ve seen the benefit, he’s been everywhere, but going everywhere means you might have a news cycle where you say something like that. I think it’s pretty striking that he won anyway. Because I do think the old way of thinking is like, say no to 10 things if it means that you never have the news cycle about globalized intifada. Him embracing the strategy he did meant that he had a news cycle in a city with a million Jewish voters. People’s views on this can be very strong. That was all about him refusing to condemn globalizing intifada. A kind of nightmare scenario if you’re a political staffer on that campaign—a genuine nightmare scenario. That didn’t have the effect that I think a lot of people would have. That really implies the politics of that are not what people think they are. I will say, I will only speak for myself on this. My priors on Andrew Cuomo, I was not like an incredible fan of the governorship from afar back when he was being talked about as a presidential candidate. Then everything that happened that led to his resignation struck me as really kind of upsetting. But I was sort of, you know, I’m open to people’s redemption. I think you have to be open to redemption. Two things about that campaign: One was that the number of people, even people who endorsed Cuomo, who talked to me about his cruelty or his tendency for revenge. That’s an amazing sentence to utter. I had somebody tell me he was a sociopath and then endorse him a couple of days later. That was like one line that I just couldn’t get over, right? Somebody who, this is the way they have treated people in public life. That’s a bar I want candidates to be above. But the other thing that actually closed it, that made for me that I would not rank him, was the way he used Israel in the campaign. I’m a Jewish person. I have very, very deep feelings about what is happening in Israel and Gaza. I found it so cynical, so repulsive, just such a vicious way to weaponize, I thought, both sort of Mamdani’s ethnicity. What’s happening in Gaza is a horror. People should be horrified. The whole thing just struck me as gross. I knew a lot of people for whom it read that way. The thing in the debates, where they got into a fight over visiting Israel: Mr. Mamdani, what’s the first country you’re going to visit? I would stay in New York City. My plans are to address New Yorkers across the five boroughs and focus on that. Mr. Mamdani, can I just jump in? Would you visit Israel as mayor? I’ve said in a UJA questionnaire that I believe that you need not travel to Israel to stand up for Jewish New Yorkers. That is what I will be doing as the mayor. I’ll be standing up for Jewish New Yorkers and I’ll be meeting them wherever they are across the five boroughs, whether that’s in their synagogues and temples or at their homes or at the subway platform. Because ultimately, we need to focus on delivering on their concerns. And just yes or no, do you believe in a Jewish state of Israel? I believe Israel has the right to exist, not as a Jewish state, but as a state with equal rights. He won’t say it has a right to exist as a Jewish state. And his answer was no, he won’t visit Israel. I said that. That’s what he was trying to say. No, no, no. It was such an obvious political game. Yeah. It was cynical. It was. Yeah, it was deathlessly cynical. Yes. And I have to say, I mean, it was also comical at a certain level. Like my formative years were spent at Shabbat dinners at my friends’ houses and going to bar mitzvahs and being in this milieu of Jewish New York. And it’s incredibly precious to me. I feel profound gratitude and affection for that. And, you know, my wife’s half Jewish. I’m not discussing my credentials, but it’s close to me. I’m not Jewish, but it’s a culture that I love deeply and feel bound to. And so, yeah, I found it deathlessly cynical, deathlessly cynical. The other thing that complicated this, and this is an interesting angle of this whole thing, is that Andrew Cuomo, like me, is a paizan from New York. The guy’s not Jewish. Yeah, Brad Lander, who cross-endorsed Momdani, is Jewish and very devoted to questions around Israel and justice. He’s also the highest-ranking Jewish official in New York City. A lot of the things that happened in this campaign happened on like a literal level and a metaphorical, symbolic level at the same time. One thing that I thought about that moment when Momdani didn’t condemn globalizing Nevada was it had this quality of, this is what he believes. He is not going to sell out a politics and a community to whom he either belongs or has very, very deep sympathy for why they feel the way they do. And with Cuomo, like, I’m not saying he does not have beliefs about Israel, but it felt like the OPPO researchers had come to him with a packet, and he was now going to use what was in the packet. A lot of things are not— We could talk about the popularity of different ideas. Some things are also just communicating what kind of person you are. But also, I’ve been very interested by the way that Israel and Gaza have become highly kind of symbolic, like, attentional in both directions, right? There is the Gaza’s genocide direction. And also, the people who have made themselves aggressively into, like, moderates, anti-leftist moderates. You see this a bit with Cuomo, but you see it with Richie Torres, right? You see it with John Fetterman, is, like, the strongest and most consistent fight they pick is on Israel. It’s, like, now weirdly the ideological delineator. Israel’s become the culture war, I think, within the Democratic Party. And if you want to really send a strong signal, like, I’m just struck by how many of the signals sent for people who do not have a lot of power over, you know, American policy towards Israel are sent on this issue. And I think there’s also an added dimension to that, which is that there’s just enormous estrangement between the establishment of the party and the base of the party. That’s right. I saw the polling on the Iran strikes were, like, 85 percent of Democrats opposed and I think 13 percent approved. Yeah. Now, if you looked at Democratic legislators’ responses, you would not think that those were the numbers. Donald Trump really exploited a huge gap between the elites in the party and the establishment on immigration and trade and the base of the party to tremendous effect. There is something like that in the Democratic Party right now on the issue of Israel. There are just poll after poll after poll. I think this has to do with a bunch of complicated factors, although I think the driving factor has been the war in Gaza since October 2023. And I think you really saw it play out in this race. I mean, New York City is the most Jewish city in the country and the most Jewish city in the world, one of the most Jewish cities in the world. Outside Tel Aviv. Outside Tel Aviv. It’s the second, you know, highest number of Jewish citizens. It’s also like that number fails to represent how Jewish the city is in terms of its cultural milieu and, like, the fabric of New York, right? And I think it’s shocking to a lot of people, and even to me, I have to say, that someone with his politics on this conflict just won the Democratic primary. And did it without shifting from that. Like, he used to support defunding the police and now I think Bo says he does it and actually doesn’t. He does not want to defund the police as mayor. He held his line here. He is an anti-Zionist, I think, and is now still. Right. He said, like, “Israel should not be a Jewish state.” Yeah. I mean, I think that—I feel a little weird about this conversation because I really—it’s thorny for a million reasons, but it’s also, I respect the views of people that are closest to it and I am not the closest to it. So I’m always kind of trying to check that in me. So it’s weird for me to be like, “it’s bad for the Jews.” I’m not a Jew. I think the way this is developing within the Democratic Party is kind of dangerous. Yeah. I think the idea of, like, this is a signifier of the rich elites who control everything behind closed doors, which is both an anti-Semitic trope and something that touches on something close to being true about how money flows in Democratic politics is like a really combustible mix. I think that’s right. But I’d say two other things about it being a signifier. One is it—it’s a signifier in two directions, right? It’s a signifier in one direction of being willing to stick to your beliefs that I think a lot of people in the base feel that even Democrats who actually agree with them will not say on Gaza and how bad and horrifying that has been. We’ll not quite say it or sugarcoat it or will not vote with it. And so there is something both—again, I believe the belief is authentic to Mamdani, but also—it’s expressive. Yeah. Showing that you will stand up to that kind of pressure. Yep. In the other direction, it’s showing that you will not be cowed if you’re Richie Torres’s, you’re Fetterman’s. It’s showing you’ll not be cowed by a different thing in the party. Yes, exactly. The woke mob. Like the woke mob. Yeah. Right? So it’s become a kind of declaration of independence from that. I will just say on the point you just made about how saying something true can veer close to saying something anti-Semitic. One thing I have just appreciated about Mamdani, and I appreciate about the Mamdani-Lander Alliance, I’m a Jewish person. It is very important that it is possible and understood to be possible, that you can be anti-Zionist without being anti-Semitic. And I’m not anti-Zionist in that way. I’m like a kind of two-state solution person who doesn’t really believe that that is possible, and I’m not sure what I think is plausible at this point. But putting my own politics aside, I very fundamentally believe Mamdani is anti-Zionist and not anti-Semitic, and he did a very, very, very good job, in my view. In answers of making that clear, Lander acted as a very important cross-validator for him. But in a world where Israel is going to be as brutal as it has been in Gaza, and is going to play much more of a role of like a regional hegemon militarily, which is what it has stepped into, and people are going to have very, very strong opinions, including very, very strong negative opinions on what it means for there to be roughly 7 million Palestinians who do not have equal rights and are under Israeli control. It is very, very, very important that you just have to be able to be against what the Israeli state has become and not anti-Semitic. I think it is an incredibly dangerous game that pro-Zionist people have played trying to conflate those things. Because if you tell people enough that to oppose Israel is to be anti-Semitic, at some point they’re going to say, “Well, then I guess I’m anti-Semitic.” “Well, I guess I’m anti-Semitic.” Yeah, that’s the fear. I think that the taboo around anti-Semitism, which is born of the worst atrocities in human history, is like a wildly important taboo that is breaking down everywhere we look. Let’s be clear. Like that taboo is disintegrating. Yes. And it’s disintegrating for a lot of people, and it’s terrifying that it’s disintegrating. And I, you know, the one thing I’ll say again, and this is me, offering advice that no one asked for from the position of just, you know, the Catholic boy from the Bronx who now lives in Brooklyn. But, like, I think there’s tangible, concrete things that Mamdani can do. He should be going to Borough Park, and he should be going to Ocean Parkway, and he should be talking to folks there and being like, “We’re not going to agree on Israel.” Let’s just say that from the beginning. I want you to feel safe and heard. I want your communities to thrive. I want the city to work for you. Let’s talk about how we make that happen. And I think they’re tangible. Like, there’s huge security concerns. Huge. Yeah, if you heard him on Colbert, I thought he did a very beautiful job walking that line. Yeah, I agree. You know, I remember the words of Mayor Koch, who said, “If you agree with me on 9 out of 12 issues, vote for me. 12 out of 12, see a psychiatrist.” And I had an older Jewish woman come up to me at B’nai Jeshur in a synagogue many months ago after a Democratic Club forum, and she whispered in my ear, “I disagree with you on one issue.” I’m pretty sure you know which one it is, and I agree with you on the others. I’m going to be ranking you on my ballot. And I say this because I know there are many New Yorkers with whom I have a disagreement about the Israeli government’s policies. And also, there are many who understand that that’s a disagreement still rooted in shared humanity. Because the conclusions I’ve come to, they are the conclusions of Israeli historians like Amos Goldberg. They are echoing the words of an Israeli prime minister, Ehud al-Mair, who said just recently, “what we are doing in Gaza is a war of devastation.” It is cruel. It is indiscriminate. It is limitless. It is criminal killing of civilians. These are the conclusions I’ve come to. Stephen, please. And by the way, I think that is a good place to end. Also, our final question: what are three books you’d recommend to the audience? This is an oldie but a goodie, The Name of the Rose by Umberto Eco, which is the most recent novel I’ve read. It was one of these things that I started, put down for months, and then took back up. You know how you do that with novels where you’re like, “I sort of remember where we are,” but the book is incredible. The second one is an incredible book that is not out yet, that I am able to read an advanced reader copy of. It’s by Rob Malley and Hussein Agha. It’s called Tomorrow’s Yesterday. Just got recommended in the last episode, too. It’s really something else. Partly, it’s beautifully written. It’s two people that have genuinely, incredibly distinct perspectives on the Israeli-Palestinian conflict and who have been in the room at a bunch of times. So that is a great book. And the last book is a history of the Cultural Revolution called Mao’s Last Revolution by Michael Schoenhaus and Roderick McFarquhar. I don’t know why I suddenly was seized with an interest in reading about the Cultural Revolution, except that I was looking to escape to a political environment that was, like, more dire and toxic than our own. You were reading? So I was like, for some reason, scrambled to that. I read that book; it’s amazing, although, I mean, my God, sort of suffocating in some ways to be inside that universe. And then there are, like, a few whiffs of familiarity that are unnerving. Chris Hayes, always such a pleasure, man. Thank you. Loved it. This episode of The Ezra Klein Show is produced by Roland Hu and Jack McCordick. Fact-checking by Michelle Harris with Kate Sinclair and Mary Marge Locker. Our senior engineer is Jeff Geld, with additional mixing by Amin Sahota and Isaac Jones. Our executive producer is Claire Gordon. The show’s production team also includes Marie Cassione, Elias Isquith, Marina King, Annie Galvin, Jan Koble, and Chris Stinlin. We have original music by Pat McCusker, audience strategy by Christina Samulewski, and Shannon Busta. The director of New York Times Opinion Audio is Annie Rose Strasser. This is an especially terrible time to be a refugee. Because after fleeing war, violence, and persecution, it seems like the world’s capacity for compassion is running short. And as the numbers of refugees globally reach catastrophic levels, UNHCR, the UN Refugee Agency, needs your support. In just 72 hours, your donation will deliver shelter, protection, and food to refugees when they need it most. Your compassion today will make a life-saving difference. Go to unrefugees.org/donation to make your gift. window.tocIndex = { "index": [ { "index_sentences": "Can you guess where a large part of your Uber fare goes?", "section_level": 1, "section_title": "Introduction" }, { "index_sentences": "Out of control insurance costs.", "section_level": 2, "section_title": "Uber Fare / Insurance Costs" }, { "index_sentences": "The Democratic primary that just wrapped up in New York was a collision between two very different candidates on almost every level.", "section_level": 1, "section_title": "The New York Democratic Primary" }, { "index_sentences": "Andrew Cuomo ran a campaign that was based on a tried-and-true strategy of buying attention.", "section_level": 2, "section_title": "Cuomo's Campaign Strategy (Buying Attention)" }, { "index_sentences": "And then you had Mamdani, who was running a campaign on a very different theory of attention—a theory of viral attention.", "section_level": 2, "section_title": "Mamdani's Campaign Strategy (Viral Attention)" }, { "index_sentences": "Attention works differently now.", "section_level": 2, "section_title": "Attention Works Differently Now (Collision and Outcome)" }, { "index_sentences": "Now, I'm not a New Yorker, but I want somebody who is a New Yorker, who has deep roots here, and who really understands political attention.", "section_level": 1, "section_title": "Interview with Chris Hayes" }, { "index_sentences": "So I asked my friend Chris Hayes, an MSNBC anchor and the author of a phenomenal book on attention to politics, The Siren's Call, to join me.", "section_level": 2, "section_title": "Introducing the Guest" }, { "index_sentences": "Chris Hayes, welcome back to the show.", "section_level": 2, "section_title": "Start of Interview" }, { "index_sentences": "So, Zoran Mamdani won the primary.", "section_level": 2, "section_title": "Siren's Call Analysis of Mamdani" }, { "index_sentences": "I want to play one of them here.", "section_level": 3, "section_title": "Mamdani Video Example" }, { "index_sentences": "What struck me about that video when I saw it was so many politicians do communication in terms of what they are telling you.", "section_level": 3, "section_title": "Listening as Broadcasting" }, { "index_sentences": "He is the first politician I have seen be native to the thing that is after what I think we think of as social media.", "section_level": 3, "section_title": "Native Communication Styles" }, { "index_sentences": "One point I want to make here that I think is important, I think we both agree on, is that with all these discussions, there's stuff that's new and there's stuff that's timeless, right?", "section_level": 3, "section_title": "New Mediums vs. Timeless Talent" }, { "index_sentences": "And I don't think you can even talk about the Mamdani one without also, like, what his foil was.", "section_level": 3, "section_title": "The Cuomo-Mamdani Contrast" }, { "index_sentences": "So, on the first point, to take a step back, people really have to understand that for probably, I'd say, the last 40 years, there's this formula for how—and I think it's true for both parties, but I know Democratic politics better.", "section_level": 3, "section_title": "Traditional Campaign Strategy (TV Buys)" }, { "index_sentences": "I really want to hold on this for a minute because you cannot buy attention now the way you once could.", "section_level": 3, "section_title": "Earning Attention in the Modern Era" }, { "index_sentences": "If I'm the DSCC or the CCCC, I would start thinking not about who do I think can raise money, but who can raise attention.", "section_level": 3, "section_title": "Party Strategy Implications" }, { "index_sentences": "I have just thought a lot since then about why their videos are so bad.", "section_level": 2, "section_title": "Why Politicians' Videos Are Bad" }, { "index_sentences": "Okay, I'm going to, here's a structural answer to that question, which I don't hold me to, but here's a hypothesis.", "section_level": 3, "section_title": "Structural Reasons (Coalition Management)" }, { "index_sentences": "Okay, but let me push on this a little bit.", "section_level": 3, "section_title": "Counterexamples and Remaining Gap" }, { "index_sentences": "I think you have to be a consumer to be a producer.", "section_level": 3, "section_title": "The Need to Consume Media" }, { "index_sentences": "That seems like a thing where you should be looking for certain kinds of talent.", "section_level": 3, "section_title": "Importance of Talent and Instinct" }, { "index_sentences": "So one thing we saw here is that, yes, Momdani was trying to make this election about affordability, about material concerns.", "section_level": 1, "section_title": "Election Analysis Beyond Attention" }, { "index_sentences": "What did you make of the somewhat strange structure of the coalitions?", "section_level": 2, "section_title": "Primary Election Coalitions" }, { "index_sentences": "I don't really have a good theory on it yet.", "section_level": 3, "section_title": "Theorizing on Coalition Makeup" }, { "index_sentences": "So I think you can cut politicians into these two categories.", "section_level": 2, "section_title": "Policy as Memetic Communication" }, { "index_sentences": "You can lament that what modern media is doing is flattening policy down to this sort of bumper sticker level of memetic communication.", "section_level": 3, "section_title": "Why Memes are Simple and Effective" }, { "index_sentences": "We actually, I think we have a good tangible example in recent history in exactly this context from the mayor that Zoran Mandani says was the best mayor of his life; that got the New York Times very mad at him, which was Bill de Blasio's universal pre-K.", "section_level": 3, "section_title": "Universal Pre-K: A Successful Memetic Policy" }, { "index_sentences": "And yet it didn't stop everybody from turning on Bill de Blasio.", "section_level": 1, "section_title": "From Campaign to Governance" }, { "index_sentences": "The downside for him was not that he loses a primary.", "section_level": 2, "section_title": "The Challenge of Delivering on Promises" }, { "index_sentences": "I mean, one thing I would say is I like experimentation and new ideas.", "section_level": 2, "section_title": "Experimentation in Governance" }, { "index_sentences": "One thing that struck me a lot about Mamdani was his ability to listen.", "section_level": 2, "section_title": "Affordability in Modern Politics" }, { "index_sentences": "The concern is whether or not, from a policy perspective, my concern with Mamdani.", "section_level": 3, "section_title": "Housing Policy Challenges (Supply vs. Price Control)" }, { "index_sentences": "I think one wrinkle to the housing question, which I think is a really important thing to always keep coming back to when you discuss in your book, is that \"one person's price is another person's income.\"", "section_level": 3, "section_title": "Renters vs. Homeowners Conflict" }, { "index_sentences": "I do worry that the structural nature of public opinion now is negative in a way that makes even good governance not resonate with people, if that makes sense, or the structural limitations on governing.", "section_level": 2, "section_title": "Governing Challenges in a Negative Public Opinion Climate" }, { "index_sentences": "So translating this kind of communication from campaign to governance, not that many people have had to do it, but Obama had to do it.", "section_level": 2, "section_title": "Campaign vs. Governance Communication" }, { "index_sentences": "I think this is something you're seeing with Donald Trump right now, which is he actually has an instinct for how to turn policy that isn't affecting that many people into something that is intentionally salient.", "section_level": 3, "section_title": "Trump and Policy as Performance" }, { "index_sentences": "From the use of the bully pulpit, David Shore had a thing the other day about one of the most consistent counterintuitive findings: when a president talks about something, its negatives go up.", "section_level": 3, "section_title": "Attentional Innovation and Polarization" }, { "index_sentences": "What it works to do is set narrative. That is its own dimension of power.", "section_level": 1, "section_title": "The Trade-offs of Attentional Dominance" }, { "index_sentences": "I think one place where it matters is presidential politics.", "section_level": 2, "section_title": "When Attentional Dominance Works" }, { "index_sentences": "The trade that you often see some of these candidates make is they are unleashing energy in the primary that might hurt them in the general.", "section_level": 3, "section_title": "Tradeoffs: Primary vs. General Election" }, { "index_sentences": "I think to add one wrinkle here that I think is interesting and slightly weedsy but worthwhile is that, you know, New York City has ranked choice voting.", "section_level": 2, "section_title": "Ranked Choice Voting and Coalition Building" }, { "index_sentences": "I think some of his people will not like hearing me say this.", "section_level": 2, "section_title": "Mamdani as Left Pluralist" }, { "index_sentences": "What struck me often about Mamdani's affect was that it felt a bit like a TikTok affect.", "section_level": 2, "section_title": "Mamdani's Affect and Media Structure" }, { "index_sentences": "But I believe—I believe this strongly—that the rise of populist right and, to a lesser extent, populist left politics all across the world, all at the same time, I believe the single strongest force there was not just immigration.", "section_level": 3, "section_title": "Social Media (Illiberal) vs. Vertical Video" }, { "index_sentences": "I mean, the fundamental reality of the Twitter text box is a little less true now, but it still is basically true, is that it's a compression mechanism.", "section_level": 3, "section_title": "The Evolution of Media Formats" }, { "index_sentences": "Because part of attention is sometimes conflict, provocation, views that are not boring, that jump out at you, and interviews and talking to a lot of people where you might say something that is a quote-unquote gaffe or that people don't like or offend certain people.", "section_level": 2, "section_title": "The Risk of the \"Go Everywhere\" Strategy" }, { "index_sentences": "He gave an interview to a Pakistani news channel in Urdu.", "section_level": 3, "section_title": "The \"Globalize the Intifada\" News Cycle" }, { "index_sentences": "The headline that comes out from it—I don't think it was a great answer, to be very clear—is, \"refuses to condemn globalized intifada.\"", "section_level": 3, "section_title": "Mamdani Won Anyway: Old Rules Didn't Apply" }, { "index_sentences": "I will say, I will only speak for myself on this.", "section_level": 1, "section_title": "The Israel Issue" }, { "index_sentences": "My priors on Andrew Cuomo, I was not like an incredible fan of the governorship from afar back when he was being talked about as a presidential candidate.", "section_level": 2, "section_title": "Cuomo's Cynicism vs. Mamdani's Stance" }, { "index_sentences": "But also, I've been very interested by the way that Israel and Gaza have become highly kind of symbolic, like, attentional in both directions, right?", "section_level": 2, "section_title": "Israel as Culture War in the Democratic Party" }, { "index_sentences": "And I think there's also an added dimension to that, which is that there's just enormous estrangement between the establishment of the party and the base of the party.", "section_level": 2, "section_title": "Estrangement on Israel: Elites vs. Base" }, { "index_sentences": "But in a world where Israel is going to be as brutal as it has been in Gaza, and is going to play much more of a role of like a regional hegemon militarily, which is what it has stepped into, and people are going to have very, very strong opinions, including very, very strong negative opinions on what it means for there to be roughly 7 million Palestinians who do not have equal rights and are under Israeli control.", "section_level": 2, "section_title": "Anti-Zionism vs. Anti-Semitism" }, { "index_sentences": "I will just say on the point you just made about how saying something true can veer close to saying something anti-Semitic.", "section_level": 2, "section_title": "Mamdani's Approach to Jewish New Yorkers" }, { "index_sentences": "And I, you know, the one thing I'll say again, and this is me, offering advice that no one asked for from the position of just, you know, the Catholic boy from the Bronx who now lives in Brooklyn.", "section_level": 3, "section_title": "Addressing Concerns and Shared Humanity" }, { "index_sentences": "And by the way, I think that is a good place to end.", "section_level": 1, "section_title": "Conclusion and Outro" }, { "index_sentences": "Also, our final question: what are three books you'd recommend to the audience?", "section_level": 2, "section_title": "Book Recommendations" }, { "index_sentences": "This episode of The Ezra Klein Show is produced by Roland Hu and Jack McCordick.", "section_level": 2, "section_title": "Production Credits" }, { "index_sentences": "This is an especially terrible time to be a refugee.", "section_level": 2, "section_title": "UNHCR Sponsorship Message" } ] }; window.faq = { "qas": [ { "answer": "Mamdani won by employing a strategy of \"viral attention,\" utilizing short vertical videos on platforms like Instagram and TikTok to bypass traditional, bought advertising.", "index_of_source": "And then you had Mamdani, who was running a campaign on a very different theory of attention—a theory of viral attention.", "question": "How did Zoran Mamdani win the Democratic primary against Andrew Cuomo, given Cuomo's financial advantage?" }, { "answer": "The text hypothesizes that success in Democratic politics often selects for skills in managing complex multiracial, multiethnic, multilingual coalitions, which is a different skill set than effective public communication to a broader audience (\"the normies\") through modern media.", "index_of_source": "Okay, I'm going to, here's a structural answer to that question, which I don't hold me to, but here's a hypothesis.", "question": "What does the text suggest is the primary reason traditional Democratic politicians struggle with creating effective online video content?" }, { "answer": "An interesting observation from election analysis was that Andrew Cuomo won precincts that were basically all Black or all White, while Mamdani performed better in neighborhoods that were more racially mixed.", "index_of_source": "What did you make of the somewhat strange structure of the coalitions? I don't really have a good theory on it yet.", "question": "What unintuitive observation was made about the coalition structure that supported Zoran Mamdani?" }, { "answer": "Memetic policies are simple, easily rememberable, big, activate issues people care about, and often controversial, allowing them to cut through the noise and identify the politician, even if the detailed policy behind them is complicated.", "index_of_source": "Momdani had like four or five, right? There was freezer rent, there were free buses, there was free daycare, it was publicly owned grocery stores.", "question": "The text discusses Zoran Mamdani's policy proposals as being \"memetic\". What does this mean in the context of his campaign?" }, { "answer": "The text questions the feasibility of delivering on these promises, suggesting Mamdani might not have the necessary funds (especially needing tax increases from state government) or that policies like extended rent freezes could potentially discourage building and maintaining housing supply in the long run.", "index_of_source": "I'm very skeptical that Mamdani can make free daycare happen.", "question": "What potential conflict is highlighted regarding Mamdani's major policy promises like free daycare or rent freezes?" }, { "answer": "Mamdani acknowledged the term's use in protests but emphasized the safety of Jewish New Yorkers and explained \"intifada\" means \"struggle,\" used historically by the Holocaust Museum. Despite a headline suggesting he refused to condemn it, he won the primary, implying the political impact was less damaging than traditionally expected.", "index_of_source": "He gives an answer that starts off with, I thought, a very long and good thing about Jewish safety and the Jewish folks that he's talked to in New York City.", "question": "How did Zoran Mamdani navigate the controversial question about \"globalize the intifada\" and what was the perceived political outcome?" }, { "answer": "The hypothesis is that older platforms (Twitter, Facebook) are structurally illiberal, favoring conflict, engagement, and compressed text, while newer vertical video platforms, despite some overlap with conflict, are often more visual, capacious, and directed towards 'day-in-the-life' content, potentially fostering a different, less populist affect.", "index_of_source": "I believe this strongly—that the rise of populist right and, to a lesser extent, populist left politics all across the world, all at the same time, I believe the single strongest force there was not just immigration.", "question": "What is the proposed difference in how political communication functions on older social media platforms like Twitter versus newer vertical video platforms like TikTok?" }, { "answer": "Instead of buying attention through traditional means like TV ads, politicians must now \"earn it\" by being present on various platforms and creating content that is native to those platforms and the places they represent, which can be different for local versus national candidates.", "index_of_source": "You really want to hold on this for a minute because you cannot buy attention now the way you once could.", "question": "The text argues that attention cannot be bought the way it once could. How does it suggest politicians must now gain attention?" }, { "answer": "The major challenge is translating campaign promises and communication strategies into successful governance, especially given the reality of incremental change and structural limitations, which can lead to disappointment among supporters if expectations set during the campaign are not met.", "index_of_source": "So translating this kind of communication from campaign to governance, not that many people have had to do it, but Obama had to do it.", "question": "What challenge does the text identify for politicians who campaigned on large, ambitious promises once they enter office?" }, { "answer": "RCV incentivizes candidates not just to be the most distinguishing \"bomb thrower,\" but also to avoid alienating other candidates' supporters, encouraging cross-endorsements and coalition building to secure second, third, or fourth rankings.", "index_of_source": "The ranked choice voting allows voters to rank five different candidates.", "question": "How does Ranked Choice Voting in New York City primaries potentially change the strategic incentives for candidates like Mamdani?" } ] };

2025/6/28
articleCard.readMore

1. Introduction: Freeman’s Top Five Tips for Studying the Revolution

1. Introduction: Freeman’s Top Five Tips for Studying the Revolution Prof: Now, I’m looking out at all of these faces and I’m assuming that many of you have probably arrived here with some preconceived notions about the American Revolution. I’m assuming that at least some of you are sitting there and in the back of your mind you’re thinking—Declaration of Independence, a bunch of battles, George Washington, a little bit of Paul Revere thrown in—and all of those things are going to appear in the course, but obviously the real American Revolution is a lot more complex than that. It’s more than a string of names and documents and battles, and as a matter of fact, in many ways, the American Revolution wasn’t just a war. If you went back to the mid-eighteenth century, went back to the period of the Revolution or maybe just after it, and you asked people how they understood what was happening, many of them would tell you that the war was actually only a minor part of the American Revolution. Some would tell you the war actually wasn’t the American Revolution at all and you’ll see the— I should mention that the syllabus is finally up online so it’s there for you, but you will see when you look at the syllabus that at the very start of it there are two quotes and I want to read them here because they make this point really well. So the first quote is from a letter by John Adams and he’s writing to Thomas Jefferson in 1815 and he’s heard about an attempt to write the history of the American Revolution so this is what Adams has to say about that. “As to the history of the Revolution, my ideas may be peculiar, perhaps singular, but what do we mean by the Revolution? The war? That was no part of the Revolution.” There is the moment where you go “Huh?” “It was only an effect and consequence of it. The Revolution was in the minds of the people, and this was effected from 1760 to 1775, in the course of fifteen years before a drop of blood was drawn at Lexington.” Okay. So there we have John Adams saying that the war was actually no part of the Revolution. It’s a pretty famous quote but it’s a pretty interesting statement. Now I want to mention here, and it’s very early in the course for me to have worked you in to liking John Adams and I’m going to talk more about him in a few minutes, but I will mention here since I’ve just read that quote if partway through the semester you decide you’re just dying to read dead people’s mail, which is basically what historians do for a living, a great volume to read is actually the letters that Jefferson and Adams sent back and forth to each other over the course of their lives. They’ve all been pulled together into one volume and the best part of that volume is the end section, the letters in which these guys were writing to each other in their old age. So you have these two Founder figures, former presidents, and they’re just basically letting it rip in these letters. They’re talking about everything. They’re talking about all the things actually you probably wouldn’t talk about normally: religion, politics, who they hate, who they like, what they thought of the Revolution, what they thought of their own presidency, what they thought of the other guy’s presidency, the top ten Founder funerals. Actually, there’s a little section, although I think it’s the top three Founder funerals, but it’s a weird, really interesting range of stuff and it’s just these two people really excited about the fact that they’ve retired and all they need to do now is write to each other and really get to know each other better. So it’s a great volume. It’s edited by Lester Cappon. The last name is C-a-p-p-o-n if you’re interested. Okay. So that quote I just read you is actually from that series of letters, Adams saying that the war was no part of the Revolution. Adams does say, “Well, maybe my ideas are a little bit peculiar,” but he’s not the only one spouting that kind of thought. So here is Benjamin Rush, who I guess in a way you could say was doctor to the stars. He was actually this renowned doctor from the revolutionary and early national period, and he had a lot of high-placed political friends. So here’s Benjamin Rush writing in 1787: “There is nothing more common than to confound the terms of the American Revolution with those of the late American war. The American war is over, but this is far from being the case with the American Revolution. On the contrary, nothing but the first act of this great drama is closed.” So there you have Benjamin Rush saying that, boy, this is a common problem. A lot of people mix up the American Revolution with the American war, and they’re not just one and the same thing. The war is over. The Revolution goes on, and Rush is saying this even as late as 1787. It’s four years after the treaty that ended the war. We’re heading into Constitution territory, and to Rush, the Revolution is continuing. So what do these people mean? Well, in part, they are expressing part of what this class is going to be exploring. They’re basically suggesting that the American Revolution represented an enormous change of mindset as loyal British colonists—right?—long-standing loyal British colonists, were transformed gradually into angry revolutionaries and ultimately into Americans. Like John Adams suggests, the beginnings of this transformation predate the actual fighting, and like Benjamin Rush suggests, it doesn’t just come to a close when you sign a peace treaty. So when you look at things from this broad view, the Revolution actually becomes the beginning of a period in which the American nation was really inventing itself, and this is a really dramatic kind of invention. You have—In a sense, we’re just little pipsqueaks at this point, and so you have these little pipsqueaks and they are actually saying, “Okay. We reject monarchy. We’re going to turn towards a democratic republic.” They’re saying, “Yeah. Well, we know the power’s been at the imperial center forever. We’re going to turn our backs on that and pull power into what’s basically the margins of the British empire.” They’re turning away from an assumption that the few are in power and they’re saying, “Well, what if we try putting the many in power?” Those are pretty dramatic changes, and they aren’t of course the only changes. People—Colonists began to think about themselves differently. It’s really easy to underestimate the degree to which individual colonies at that time were really like little independent nation-state colonies. They were not united in any sense of the term. There wasn’t any tradition of colonies being able to communicate between each other. It was actually in some ways easier to communicate with the mother country than to get some kind of news up and down the Atlantic seaboard. Colonists often knew more about the mother country than they knew about people from other colonies. When you look at correspondence from this period, people often refer—Northerners will refer to Southerners as though they’re people from a strange, alien country who have weird accents. It’s hard to know what they are saying; they dress so strangely. It’s amazing to think about the differences, the degree to which colonies really stood alone in this time period. And this idea, that there really is pretty much no reason to assume that these colonies would have been… Able to join together, that’s pretty much going to be in the first two or three lectures of the course. What we talk about is we try and get a sense of who these colonists are, and how they’re ending up moving their way into a revolution. So this scattered group of independent colonists gradually came together to form one united nation, not the goal but the outcome. Given everything that I just said, you can see why this idea that there might be a united nation is actually a pretty big surprise. You can see why a lot of people assumed that it could never work. You can actually also assume why a lot of people might not even have liked it as an idea, and you can even see why after the Constitution goes into effect and the government is getting under way, even then people were really just not sure this thing was going to work. They really—They referred to it as an experiment, which is really how they viewed it. And it’s amazing when you look at letters from the 1790s you’ll see these little throwaway comments like “If this government lasts more than five years, here’s what I think we should do.” Okay, there—It’s a completely weird mindset and it’s not something that we would assume is there, but this is pretty much a high-stakes experiment. So this class is going to explore this big shift in mindset, and the war will be at the center of this shift, and it’s going to do this from a participant’s point of view. It’s going to really grapple with how things made sense at the time to the people who were there. And I’m going to go more into that in a minute or two. I want to talk for just a second about how the course is organized and just for a minute or two about some of the readings for the course. The course is partly chronological and partly thematic so we do proceed along, we follow the narrative of events of how things evolved, all those nasty acts, people protesting, have a war, try to figure out what to do after the war. We do follow that sort of trajectory, but we’re also going to once in a while step back and look at the big picture, so that we’re not just following events; we’re going to be always putting events in context. And the readings for the course go in that same direction. We’re going to read Gordon Wood’s Radicalism of the American Revolution, which is a really great overview of this time period and also presents an argument, obviously as you could tell from the title, that the Revolution was really radical. Some people agree with that and some disagree, and actually one of the discussion sections is geared around discussing that very idea, and by the end of the course you’ll probably have some pretty strong ideas not necessarily agreeing with mine but, based on what you’ve read and what I’ve said and what you’ve thought, you’ll probably have some strong ideas about how radical was the American Revolution. We’re going to be reading Robert Gross’s The Minutemen and Their World, which you can hear is right along the lines of what I was just saying. It really gives you a sense of what it was like at the time for people who ended up doing things like fighting at Lexington and Concord. We’re going to read Bernard Bailyn’s Faces of Revolution, which includes an array of chapters on different people who played a major role in the Revolution as well as chapters on the ideals and the ideology or basically the logic of American independence, and Bailyn is really well known as sort of—He wrote this amazing book on the ideology of the American Revolution, and what you’re going to be reading; he basically took a big, meaty chunk from that book, the part that everybody really focuses on, and put it in this book. Faces of Revolution. So we will be reading that as well as part of the readings for the course. We’re also going to be reading Ray Raphael’s A People’s History of the American Revolution, which does just what the title would suggest. It looks at how different kinds of people: Native Americans, average rebels, African Americans, Loyalists, and women, experienced the Revolution. In addition to reading historical scholarship, we’re going to be reading some of the literature of the period. We’re going to be reading Thomas Paine’s Common Sense, which I love. How many of you have read Common Sense before? A good number of you, not—yeah, some of you. I love Common Sense. I think it’s an amazing piece of writing, and I think when you read it for this course you’ll get a sense of why it had such a huge influence at the time. We’re going to read some essays from The Federalist written by Alexander Hamilton, James Madison, and John Jay, but we are not going to read them as—You may have read them before. You may have encountered The Federalist essays as the grand source of authority on the Constitution. Right? How could it not be that when you have Founder-type guys talking about the Constitution and they were the guys who were at the convention? Well, the fact of the matter is The Federalist essays weren’t intended to be an objective document. They’re actually really subjective, and we’re going to look at them in this course as what they were written to be, which is a really big commercial advertisement for this new experimental Constitution. They were actually trying to sell people on an idea, and because of that, as we’ll see when we read that for this course, there are things in there that maybe are a little bit exaggerated and things in there that maybe aren’t talked about in great detail, and one or two things that probably aren’t really true, but it was in a good cause. Right? These guys are saying, “I really think this Constitution is the way to go. Let me say something that’s going to really calm you so that we can go ahead with this experiment.” And of course, we’re going to be reading the Declaration of Independence. We’re going to be reading the Constitution. We’re going to be reading a lot of documents and letters and other kinds of assorted items to really give us a sense of the period, and at one point I’m even going to bring in a newspaper from the period so that we can actually look at it and get a sense of how people are getting news of the war at that moment. A lot of these documents we’re going to pull from a book called Major Problems in the Era of the American Revolution, and it’s a nice collection of primary documents and essays about sort of related themes. It always makes me laugh when I say that title because it’s part of a series of books and the books are Major Problems in the Revolution, Major Problems in the Early National Period, Major Problems in the Civil War, so basically all of American history appears to be a major problem, which—It kind of gives me pause, but despite that, it’s a nice collection of things, and we’ll be using that for the class. So that gives you a sense of how the course is going to flow and what these readings are going to do, which brings me to the portion of the lecture that I’m going to call Freeman’s Top Five Tips for Studying the American Revolution, and I want to explain before I launch into them what the heck I mean. Basically, when I was preparing this lecture and thinking to myself what do I want you to know about right at the outset. before we even start talking about the Revolution itself. And I ended up with a list of things that, as I talk about them here, may seem obvious, but the more I talk about them, I think the less obvious they’ll appear. They’re actually really important to consider in a course that deals with something like America’s founding. There’s a lot wrapped up in that. Just that phrase. Just think about the phrase—right?—the ‘Founding Fathers,’ the ‘Founding Period.’ You can see the capital letters. You don’t even need to see it in writing. In your mind, it’s always capitalized. We assume a lot of things about this time period, and it’s sort of an iconic period when you think about American history. To us, a lot of the people and events of this time period and the documents of this time period are kind of what America is all about, which is understandable. But to think about the founding period as historians, we need to think differently. We need to be aware of all of those assumptions, all of that cultural baggage that we bring when we’re looking at something like the American Revolution. We need to be aware of them, and then we need to get past them so that we can really begin to understand the people and events of the Revolution for what they were. And that’s how I got to Freeman’s Top Five Tips for Studying the American Revolution; five things that you should bear in mind when studying this period—five things, obviously, that will be useful to remember throughout the course—basically all of them aimed at just shaking the assumptions right out of us. The first tip is actually really related to that point. The first tip is: Avoid the dreaded Revolutionary War fact bubble. What I mean by that is you’re going to be sitting here, and over the course of the semester, you’re going to hear a lot of familiar names and events: the Boston Tea Party, George Washington, the greatest hits of the Revolution—the things you know and love and learned in high school. They’re all going to be here, and hearing all of these beloved greatest hits, you may be tempted to sort of sit back in your seat and drift along with the happy, familiar events. “Aah, the story of the American Revolution; I love the story of the American Revolution.” Well, I love the story of the American Revolution, but there’s a different story of the American Revolution besides all these names, facts, and dates that you probably have arrived here with in your head. It’s a really good dramatic story, but it’s not a string of facts, so thus the fact bubble. It’s not a fact bubble. The Revolution obviously is a lot more than that, and you need to sort of almost be aware of the fact and then allow yourself to step back and look at the big picture. And John Adams and Benjamin Rush and others like them would have been the first to tell you the facts, in a sense, are the least of it. So that’s—Tip number one is don’t get lost in the dreaded Revolutionary War fact bubble, which I have to say makes me think of the first time that I taught this course. I was actually a brand new professor, and I had just come to Yale, and it was my first course and it was my first lecture in my first course and I’m [sound cuts out]. It actually was in Connecticut Hall, which, for those of you who don’t know, dates back to the period when this course is talking about and was Nathan Hale’s—essentially his dorm. So there I am. I’m a brand new professor to Yale, and I’m teaching a course about the Revolution, and it’s in a building that dates to the Revolution. So I’m having sort of a “wow” Yale moment as it is, and I’m off, I’m giving my lectures. and I’m really excited. I give about three of them and someone raises their hand after about three lectures and they have kind of a puzzled expression on their face. I said, “Yes?” And he says, “Excuse me, Professor Freeman. What are we supposed to be memorizing? Where are the facts and dates?” [laughs] So as a new professor, my first impulse was: Darn! I forgot the facts and dates. [laughter] I got it wrong. [laughs] But actually, the fact of the matter is, they’re not the star of the show. Certainly, dates are not the star of the show. There are dates you’re going to have to remember, so don’t think Easy Street; there’s not a date I have to know. There will be some dates, but this isn’t a story about dates. It’s obviously something a lot more interesting and a lot broader than that. Okay. Avoid the fact bubble. Tip number two: Think about the meaning of words. Now on the one hand, this may seem really obvious, and you may be sitting here thinking, oh, great, this is going to be a semester of Freeman saying: “What does revolution mean? What does war mean?” which would be a really, really, really long semester, and that’s actually—there might even be a point where I’ll say, “What does revolution mean?” I even kind of, sort of said it already, but that’s not what I mean when I say think about the meaning of words. What I really mean here is be careful what you assume about words because what seems obvious in meaning to you now probably meant something really different in 1776 or 1787, and I want to look at one example because it’s a really striking one, and that’s the word “democracy.” Okay. So sitting here in this room, by our standards, democracy is a good thing. Right? Democracy is a good thing. Every once in a while, as a professor, you say something and then you think with horror about how it’s going to look in your notes. So you’ll have all these notes, and then it will say “democracy is a good thing,” [laughs]—a really sophisticated class we’re teaching here at Yale. But to us, it’s good, and to people in the founding generation, not so much. They weren’t so sure about it. To them, the word “democracy” signaled a kind of government in which every single person participated personally, not a government based on representation. We’re talking mass politics, in the minds of most people in the founding generation just the definition of what chaos was. So just listen to a sentence in one of the last letters written by Alexander Hamilton in 1804, the night before his duel with Aaron Burr. So he’s sort of speaking to posterity in case he should die, and this is what he writes in this letter: “I will here express but one sentiment, which is that the dismemberment of our empire”—I love the fact that America’s an empire in 1804—“will be a clear sacrifice of great positive advantages without any counterbalancing good, administering no relief to our real disease, which is democracy.” Okay. Our real disease is democracy, Alexander Hamilton. Now admittedly, Hamilton might not be the shining example of the point I’m trying to make here because he’s not exactly Mr. Democracy, so you wouldn’t really expect him to be clapping his hands for it. But now listen to Thomas Jefferson, who maybe you would brand Mr. Democracy. So Jefferson in 1816 is chatting away with someone in a letter about what America’s trying to do and whether America’s actually achieving it, and he says, Actually, democracy is pretty impractical. He can imagine it in a town, but outside of one town, it just won’t work; again, really clear. Their sense of what that word means is really different from our sense of what that word means. Now Jefferson immediately goes on to add that democratical—a democratical but representative government is a good thing. Right? A democracy, not so much, but democratical, which is a great—In the eighteenth century, they were always adding “ical” on to the end of things, which could end perfectly happily with a “c.” [laughs] This is a great eighteenth-century-sounding word, “democratical.” A democratical representative government is a good thing, but democracy not. So the moral of this story is don’t fall into what I call ‘democraspeak.’ Don’t write papers where you toss around terms like: “democracy,” “liberty,” “freedom,” without really thinking about what you mean and what they meant. As Americans, we’re used to tossing those words around, but to early Americans, if you think about it, to early American slaveholders, words like “liberty,” and other such words, have a much more complicated meaning. So tip number two: Think about the meaning of words. Which brings us to tip number three: Remember that Founders were people. Now, as I was writing this, I thought, oh, that’s another one of those things I don’t want to see in people’s notes: [laughs] Democracy is good, Founders are people. [laughs] Such a highfalutin’ course I’m teaching here. Again, it sounds really obvious, but what I really mean here is we tend to forget this pretty simple fact. We forget that the Founders were people. We assume that they were these all-knowing demigods who were sort of calmly walking their way through the creation of a new model nation. We kind of deify them. We put them up on this sort of—aah—founder mountaintop of American history, and it’s—really it’s easy to do. Sometimes just listening to their words or reading their words would inspire you to want to do that. Here is a random phrase. I thought, what could I write here that would be sort of inspiring Founder talk, and this was the one—two sentences that I came up with just because they always stick in my mind because they sound kind of amazing. This is actually Thomas Paine, Common Sense. In the middle of it, he writes, “We have it in our power to begin the world over again. The birthday of a new world is at hand.” Okay. That’s really—that’s inspiring stuff. That’s fine writing, but that’s inspiring talk and it’s supposed to be obvious because Paine’s trying to convince people that independence is a really good idea. But these kinds of words, this sort of glorious rhetoric shouldn’t block out the simple fact that the Founders were people. They were regular human beings. They were well-educated, they were thoughtful, they were sometimes well-meaning, they were sometimes hard-working, maybe sometimes not so much. They were people aiming high; they were people who did feel responsible to posterity, but still, they were people. And to me, this is one of the really exciting things about history generally and about this time period specifically. We’re talking about people trying to figure things out. We’re talking about the most basic things about America—right?—its existence, [laughs] that it is a nation, that it has a constitution, and we’re looking at people trying to figure out how all that stuff is going to be created and how it’s going to work. These are people who are scared. These are people who make dumb mistakes on occasion. They’re figuring it out as they go. The history of this period is a history of decisions of various kinds, and these were hard decisions and they were being made by people who did not know the answers. They’re making it up as they go along. I think that’s just fascinating when you read their correspondence and get a sense of how much they’re really in the dark. Actually, when you read letters from the period, a lot of them say the same thing right when the government launches. I think George Washington, James Madison, and a Pennsylvania senator almost say the exact same thing. It’s almost like they went into a room and said, “So how will we express being scared at this moment? Aah, here’s a good sentence.” They all say basically, “I feel like I’m walking on hollow ground. I feel like the ground’s going to break beneath my feet.” They’ve just launched this Constitution and they all are sort of standing there on the national stage thinking, what if this all explodes? Do we actually know what we’re doing? That’s a really fascinating part of this period to me. Now of course sometimes people did try to figure out the answers in a wonderful sort of Enlightenment way, and my favorite example of this is James Madison. He prepared himself ultimately for the Constitutional Convention by studying all governments across all time. [laughs] Can you imagine? “Well, I’d better go study all of government ever [laughter] to get ready for this convention,” but he does. It’s a great sort of Enlightenment thing to do. He’s thinking, “I will now discover the eternal pattern of politics,” and he’s thinking if I can do that, then I can reach for the best, I can avoid the worst, and whatever we’re going to do when we make a new government maybe we’re going to actually do something better than what’s come before. So there’s a logic to it; as ambitious as it is, there’s a logic. And he was serious about it, so it wasn’t even just, “I wonder what happens if I read a lot about government.” He actually was serious about it. He even made a kind of a little chart in which he listed the governments and then listed pros and cons. You know, like what did I think of Sparta? He’s sort of [laughter] amazing, across all time, so I love the fact that he did that Founder-like thing to do, but there you see a person, a really intellectually ambitious person, trying to figure things out like how do we know what to make? [laughs] What’s this constitution supposed to look like? How are we going to figure that out? Okay. I’ll just study all of government and let you know. So when we talk about the American Revolution, we’re talking about people, and this course takes this idea really seriously. Part of what we’re going to be doing over the course of the semester is looking at the Revolution from the vantage point of participants, trying to see how people at the time understood the events unfolding around them. How did the colonists understand themselves as British subjects? How did they feel about the British empire, about the King, or about Parliament? How did they come to put American-ness in the foreground? How did rebelling against their own country make sense? And that’s something I think we also tend to forget about the Revolution. We think of it as our war; it’s us against them and them equals the British, but of course we were the British. So it’s something that you don’t think about, but people at the time, certainly in the 1790s, People referred to this as “the Civil War” because that’s what it was. It’s just not the way we happen to think about it because in our mind, we’ve already traveled down the road and we’re already “us,” but in their mind, it was in a sense brother against brother; it was us against us. So we’re going to try to keep that sort of thing in mind as we explore how the events of this time made sense at that time. And I will say here we are not going to forget about the British, so we’re not going to have a patriot-centric course. The British have a logic to what they’re doing, whether they’re making policy or whether they’re fighting battles, and we will definitely look at the logic of what the British are doing as well as what the colonists are doing. Now I will admit right here up front that I will be offering you a sampling of really arrogant British quotes about crude colonists, and I’m doing that partly because there’s just so many good, really arrogant quotes about American colonists that I can’t resist sprinkling them through my lectures, and I can’t resist that so much that I have two here that I just randomly added in because I have a reason to, and so I will. But it’s just—it gives you a sense of how at least some of the British were thinking and looking at really these little upstarts on the margins of empire. So random arrogant British quote number one, and this is from a customs official: “American colonists are a most rude, depraved, degenerate race, and it is a mortification to us that they speak English and can trace themselves from that stock.” Wow. Just—Even English is a problem. That’s a statement. Arrogant quote number two, and this one I picked just because it’s about George Washington, and it’s hard to imagine people saying something arrogant about George Washington, who seems to be Mr. Symbol of Authority. But here’s another British official who met with George Washington and then wrote back home about what he thought. And he says, “Somehow, and I can’t imagine how, he’s learned the basics of how to behave in a court society.” It’s like, ‘ooh.’ That must have been a really fun interview, too, if that’s the attitude this guy had. So there’ll be a sprinkling of that because at dramatic moments in the war, there’s always someone who steps forward and offers that point of view. That said, we are going to do justice to the British side of the dispute, to the logic behind the policy that they were making, because it’s not as though they make a bunch of dumb policies that make absolutely no sense and we’re righteously outraged and then there’s a war. Their policies made sense to them. They didn’t happen to always make sense to the colonists, but they made sense to them, and the same thing goes with British battle plans which, looking back in the long view of time when we start talking about them, might seem a little goofy, but there’s actually a lot of logic for what they were trying to do when they were attacking the colonies. Oh, and I will also mention—I guess I probably don’t have time to talk about it now. When I was preparing this lecture, casting around, trying to figure out what will I put in the lecture, I don’t know how I came across it, but I discovered the Battle of New Haven. Did anyone know about the Battle of New Haven? Because I did not know. And I know there’s hostilities around—and I don’t know if there’s Yale lore that of course you’re all sitting there saying, “Well, of course we all know about the Battle of New Haven,” but I— I’m the only one who doesn’t know about the Battle of New Haven, but it’s actually a good story. I’m not going to tell it now. I’m going to leave you in suspense. It will appear when we start getting into the fighting of the Revolution, but let me just give you the sneak preview, which is it does involve the president of Yale College with a gun in his hand running to fight the British, so it’s [laughter] a Yale moment that we have, so I will talk about the Battle of New Haven. Okay. So we aren’t going to be looking at a story of good guys versus bad guys. We will be reconstructing opposing points of view, trying to figure out how those points of view made sense and then obviously we’ll be able to step back and say, “What happens when you put those two opposing points of view in contact with each other?” So tip three: Founders are people. Which brings us to tip four: We’re not just talking about Founders. The Revolution was not just a quiet conversation between a bunch of guys wearing wigs and knee britches. Right? We sort of have this image that the Revolution is guys in short pants with wigs in a room doing this. No. [laughter] That’s the entire founding I think in our minds sort of. I’ll say that right. So that’s not the Revolution. Right? That’s not what happened. We’re talking about a revolution, a popular uprising by vast numbers of colonists fought on American ground by Americans of all kinds, and it meant different things to different kinds of people. This is not to say that the Founders aren’t important, far from it, and as you will gather very quickly in this course, I love these guys. Right? I love talking about these guys, I love writing about these guys, so I’m certainly not saying, “Who cares about Founders?” But what I am saying is that they’re not the only ones who mattered. They didn’t have their own revolution while everybody else watched. We’re talking about a popular revolution grounded on the ideas and actions of people throughout many different levels of society. Now somewhat conversely, this brings us to John Adams. As I promised at the beginning, John Adams is coming and here’s John Adams. You’ll be hearing from him more than once this semester and actually you already heard from him once so I can promise that that’s true. This isn’t because I think that John Adams is the most important figure from the period. It’s not because I think that he’s always right. In fact, the reason I quote him a lot is he’s a brilliant, blunt, really direct commentator with—and this is all-important; you almost need a drum roll here—he has a sense of humor. John Adams has a sense of humor. It’s not every day that you find a Founder with a sense of humor. laughs I can vouch. There aren’t a lot of chuckling Founders. Certainly on paper there’s not a lot of chuckling going on among the Founders. Probably in person there was, but on paper not a lot of them commit humor to paper and John Adams does. He’s even self-deprecating sometimes, which—Nobody wants to be self-deprecating on paper when they know that they’re going to be a Founder, but John Adams sometimes is, and I’m going to offer one little, tiny dumb example, the first thing that popped into my mind when I thought, ‘well, what am I going to say to show John Adams’ sense of humor?’ So this is actually from that same series of letters in their old age when they’re writing back and forth to each other. So Adams is writing to Jefferson and he signs his letter with this: “John Adams in the eighty-ninth year of his age, too fat to last much longer,” [laughter] which is not typical Founder talk. [laughs] George Washington is not signing his letters that way, but Adams commits that sort of stuff to paper. What that means is not only is he blunt, direct, and intelligent, but he also even gets to be humorous as well. So in Adams, we have this sort of cantankerous, sometimes bemused, more often irritated, occasionally self-aware, sometimes really not, stubborn, book-steeped, event-experiencing, action-taking tour guide. He’s not going to be there all the time. There are going to be long stretches of the course where we don’t find John Adams, but he’s definitely going to make repeat appearances. In the middle of the course, he’ll even get to tell a really good story, which he actually basically wrote down and said, “Let me tell you this story.” So that’s—As a historian, what’s better than historical characters doing exactly what you want them to do? “Here’s a cool story on the Revolution. You can quote it in your lecture courses.” [laughter] Oh, and this actually makes me think of another John Adams-related thing before I get to tip number five, which I will get to. Partly I’m curious about this and partly I want to mention something. How many of you saw the HBO mini-series on John Adams? Okay, a goodly number of you. I mention that for a specific reason. Now, I will say I of course watched it, and the period when it was airing was a really interesting period for me because as an eighteenth-century historian, this may be the only time I’ve ever been culturally relevant to popular culture. [laughter] I was like: it’s my moment. Right? People are coming up to me and saying, “I’ve got a question about John Adams.” Wow. [laughter] This is great. So it was an interesting moment when people actually were thinking about John Adams, and I will say also I watched it with a few historians, and we were prepared to throw popcorn at the screen. We ended up pretty much liking it, and we were surprised. About halfway through, we all looked at each other and said, “It’s actually pretty good.” So I don’t—of course, there are always things that any kind of TV or movie production about history gets wrong, so I won’t say that there’s nothing wrong in it. However, there is one thing that is wrong, and I’m going to mention it because if you have pictures in your mind from the mini-series as you sit here in this course, it could be a bad thing to think that they are accurate. What I’m talking about is actually the—I think it’s the first episode. It is of course the first episode, and that’s the episode where the Revolution is beginning and you see people milling about sort of with fists. Right? That represents the Revolution. You see the beginning of the Revolution. The bizarre thing about the way that they depict it is, apparently according to the producers of this mini-series, if there was something happening in the early stages of the Revolution, John Adams apparently was there. Boy, they’re shooting at Lexington and Concord. Adams races across the countryside [laughter] to get to Lexington and Concord. Boston Massacre—John Adams staring at the—[laughter] Well, the idea is really that John Adams somehow is never off his horse, riding around Massachusetts trying to be an eyewitness to every [laughter] historical event. Now I understand that probably the people who made this Thought: how the heck are we going to communicate Boston Massacre, Lexington and Concord? This is a film about Adams, and we can’t say, “Put Adams over here while we now turn to random people on a field shooting.” laughter So I understand narrative-wise why they needed to do this, but Adams was not at every historical event laughs in the Revolutionary War. He was at many, and he definitely had an insider’s view of the Boston Massacre, but he was not everywhere in Massachusetts. Okay. That oddly enough brings me to tip number five in the Freeman Guide, and tip number five is: remember contingency. Again, an obvious thing but something we don’t think about. People at the time didn’t know what was going to happen, so Adams could not race to places where he didn’t know the things that happened yet: “Something might be happening at Lexington.” People didn’t know what was going to happen. Think for a moment about all of the things that we assume about the Revolution. We assume that the colonists were right and that the British were wrong. We assume that a Revolution was inevitable. We assume that there was broad agreement at any one time about what should be done. Right? Of course we need to declare independence. Of course the colonists are going to win the war. Of course there should be a national union. Those are all the sorts of things that I think we do assume, and that’s a lot of assumptions; that’s a lot of “of courses,” but in fact it’s important to remember that people didn’t know what was going to happen. You really need to allow for contingency because literally what they assumed was: anything can happen. Anything can happen. Again, one of the things that I love about this time period is that the emotions are so heightened. If you’re in an atmosphere where everything’s up in the air and you’re in the middle of a revolution or you’re trying to create a government and you literally don’t know what’s coming next and anything can happen, maybe I’ll get hanged by the king, maybe I’ll get shot going home, maybe America will hate the Constitution so much they will throw rocks at my head. I mean, I don’t know what they were thinking—maybe the Constitution will last four days and then collapse. Whatever they’re thinking, the fact is because they literally think anything can happen, anything could fall apart at any second, the emotions are really raised and it’s why a lot of the rhetoric in this period is so extreme. It’s not that these guys are trying to be dramatic. They actually are dramatic; they’re feeling that this is a dramatic kind of a moment, and I don’t think you get that sense. I don’t think you get that idea unless you remind yourself about contingency, about the fact that there are no predetermined outcomes and that anything can happen. I think particularly when you’re studying a revolution, it’s really important to remember contingency, and we will discover what contingency means in this time period over the course of the semester. And I will end there. I will see many of you perhaps on Thursday. I will probably know next week better about the reality of when we’ll be meeting for discussion sections. window.tocIndex = { "index": [ { "index_sentences": "Now, I'm looking out at all of these faces and I'm assuming that many of you have probably arrived here with some preconceived notions about the American Revolution.", "section_level": 1, "section_title": "Rethinking the American Revolution" }, { "index_sentences": "So the first quote is from a letter by John Adams and he's writing to Thomas Jefferson in 1815 and he's heard about an attempt to write the history of the American Revolution so this is what Adams has to say about that.", "section_level": 2, "section_title": "Defining the Revolution: Views of Adams and Rush" }, { "index_sentences": "Now I want to mention here, and it's very early in the course for me to have worked you in to liking John Adams and I'm going to talk more about him in a few minutes, but I will mention here since I've just read that quote if partway through the semester you decide you're just dying to read dead people's mail, which is basically what historians do for a living, a great volume to read is actually the letters that Jefferson and Adams sent back and forth to each other over the course of their lives.", "section_level": 3, "section_title": "Digression: The Adams-Jefferson Letters" }, { "index_sentences": "Well, in part, they are expressing part of what this class is going to be exploring.", "section_level": 1, "section_title": "The Course's Approach: Mindset Shift and Nation Building" }, { "index_sentences": "You have—In a sense, we're just little pipsqueaks at this point, and so you have these little pipsqueaks and they are actually saying, \"Okay.\"", "section_level": 2, "section_title": "Dramatic Changes and the American Experiment" }, { "index_sentences": "I want to talk for just a second about how the course is organized and just for a minute or two about some of the readings for the course.", "section_level": 1, "section_title": "Course Organization and Readings" }, { "index_sentences": "The course is partly chronological and partly thematic so we do proceed along, we follow the narrative of events of how things evolved, all those nasty acts, people protesting, have a war, try to figure out what to do after the war.", "section_level": 2, "section_title": "Key Texts and Documents" }, { "index_sentences": "So that gives you a sense of how the course is going to flow and what these readings are going to do, which brings me to the portion of the lecture that I'm going to call Freeman's Top Five Tips for Studying the American Revolution, and I want to explain before I launch into them what the heck I mean.", "section_level": 1, "section_title": "Freeman's Top Five Tips for Studying the American Revolution" }, { "index_sentences": "The first tip is: Avoid the dreaded Revolutionary War fact bubble.", "section_level": 2, "section_title": "Tip 1: Avoid the Revolutionary War Fact Bubble" }, { "index_sentences": "I have to say makes me think of the first time that I taught this course.", "section_level": 3, "section_title": "Anecdote: Students Asking for Facts and Dates" }, { "index_sentences": "Now on the one hand, this may seem really obvious, and you may be sitting here thinking, oh, great, this is going to be a semester of Freeman saying: \"What does revolution mean?\"", "section_level": 2, "section_title": "Tip 2: Think About the Meaning of Words" }, { "index_sentences": "What I really mean here is be careful what you assume about words because what seems obvious in meaning to you now probably meant something really different in 1776 or 1787, and I want to look at one example because it's a really striking one, and that's the word \"democracy.\"", "section_level": 3, "section_title": "Example: The Meaning of \"Democracy\"" }, { "index_sentences": "Now, as I was writing this, I thought, oh, that's another one of those things I don't want to see in people's notes: [laughs] Democracy is good, Founders are people.", "section_level": 2, "section_title": "Tip 3: Remember That Founders Were People" }, { "index_sentences": "The history of this period is a history of decisions of various kinds, and these were hard decisions and they were being made by people who did not know the answers.", "section_level": 3, "section_title": "Founders as Fallible Decision-Makers" }, { "index_sentences": "Now of course sometimes people did try to figure out the answers in a wonderful sort of Enlightenment way, and my favorite example of this is James Madison.", "section_level": 4, "section_title": "Example: James Madison Studying Governments" }, { "index_sentences": "Part of what we're going to be doing over the course of the semester is looking at the Revolution from the vantage point of participants, trying to see how people at the time understood the events unfolding around them.", "section_level": 3, "section_title": "The Participant's Perspective: A Civil War" }, { "index_sentences": "And I will say here we are not going to forget about the British, so we're not going to have a patriot-centric course.", "section_level": 3, "section_title": "Considering the British Perspective" }, { "index_sentences": "Now I will admit right here up front that I will be offering you a sampling of really arrogant British quotes about crude colonists, and I'm doing that partly because there's just so many good, really arrogant quotes about American colonists that I can't resist sprinkling them through my lectures, and I can't resist that so much that I have two here that I just randomly added in because I have a reason to, and so I will.", "section_level": 4, "section_title": "Examples of Arrogant British Views on Colonists" }, { "index_sentences": "When I was preparing this lecture, casting around, trying to figure out what will I put in the lecture, I don't know how I came across it, but I discovered the Battle of New Haven.", "section_level": 4, "section_title": "Digression: The Battle of New Haven" }, { "index_sentences": "The Revolution was not just a quiet conversation between a bunch of guys wearing wigs and knee britches.", "section_level": 2, "section_title": "Tip 4: The Revolution Was a Popular Uprising" }, { "index_sentences": "As I promised at the beginning, John Adams is coming and here's John Adams.", "section_level": 3, "section_title": "Reconsidering John Adams: Humor and Personality" }, { "index_sentences": "Oh, and this actually makes me think of another John Adams-related thing before I get to tip number five, which I will get to.", "section_level": 3, "section_title": "Digression: The HBO John Adams Mini-Series" }, { "index_sentences": "That oddly enough brings me to tip number five in the Freeman Guide, and tip number five is: remember contingency.", "section_level": 2, "section_title": "Tip 5: Remember Contingency" }, { "index_sentences": "People at the time didn't know what was going to happen, so Adams could not race to places where he didn't know the things that happened yet: \"Something might be happening at Lexington.\"", "section_level": 3, "section_title": "Contingency and Heightened Emotions" }, { "index_sentences": "I will probably know next week better about the reality of when we'll be meeting for discussion sections.", "section_level": 1, "section_title": "Conclusion" } ] }; window.faq = { "qas": [ { "answer": "According to John Adams, the Revolution was \"in the minds of the people\" and happened from 1760 to 1775, before any fighting began at Lexington.", "index_of_source": "As to the history of the Revolution, my ideas may be peculiar, perhaps singular, but what do we mean by the Revolution?", "question": "According to John Adams, what was the true nature of the American Revolution, and when did it primarily occur?" }, { "answer": "Benjamin Rush stated that it was common to confuse the two, asserting that the war was over in 1787, but the Revolution was far from finished, with only \"the first act of this great drama\" closed.", "index_of_source": "There is nothing more common than to confound the terms of the American Revolution with those of the late American war.", "question": "How did Benjamin Rush describe the relationship between the American Revolution and the American War?" }, { "answer": "The formation of a united nation was not a guaranteed outcome but rather a \"big surprise,\" as the individual colonies initially functioned more like independent nation-states with poor communication and significant differences between them.", "index_of_source": "Given everything that I just said, you can see why this idea that there might be a united nation is actually a pretty big surprise.", "question": "What is one counter-intuitive aspect about the formation of the United States during the Revolutionary era, as described in the text?" }, { "answer": "The American Revolution represented a transformation of loyal British colonists into angry revolutionaries and ultimately into Americans, involving dramatic changes like rejecting monarchy for a democratic republic and shifting power from the imperial center to the margins.", "index_of_source": "They're basically suggesting that the American Revolution represented an enormous change of mindset as loyal British colonists—right?—long-standing loyal British colonists, were transformed gradually into angry revolutionaries and ultimately into Americans.", "question": "What significant shift in mindset is the American Revolution said to represent, according to the speaker?" }, { "answer": "To the founding generation, \"democracy\" typically meant a government where every person participated directly, not through representation, which they often viewed as chaotic or even a \"disease\" (as Alexander Hamilton put it).", "index_of_source": "To them, the word \"democracy\" signaled a kind of government in which every single person participated personally, not a government based on representation.", "question": "What did the term \"democracy\" generally mean to people in the founding generation, and how did they view it?" }, { "answer": "It means remembering that the Founders were regular human beings, not \"all-knowing demigods,\" who were often scared, made mistakes, and were figuring things out as they went, such as their expressed fears about the new Constitution possibly failing.", "index_of_source": "These are people who are scared.", "question": "What does the speaker mean by the tip \"Remember that Founders were people,\" and what is an example of this?" }, { "answer": "While the Founders were important, the Revolution was fought by vast numbers of diverse colonists on American ground, meaning it involved people from many different levels of society and meant different things to different groups, not just a conversation among a few elite men.", "index_of_source": "The Revolution was not just a quiet conversation between a bunch of guys wearing wigs and knee britches.", "question": "Why does the course emphasize that the American Revolution was a \"popular uprising\" and not just about the Founders?" }, { "answer": "The speaker quotes John Adams frequently because he was a \"brilliant, blunt, really direct commentator\" who possessed a \"sense of humor,\" which is rare among the Founders in their written communication.", "index_of_source": "In fact, the reason I quote him a lot is he's a brilliant, blunt, really direct commentator with—and this is all-important; you almost need a drum roll here—he has a sense of humor.", "question": "Why does the speaker often quote John Adams, even if not always agreeing with him?" }, { "answer": "Remembering contingency means recognizing that people at the time did not know what was going to happen; there were no predetermined outcomes, and literally anything could have happened, which explains the heightened emotions and extreme rhetoric of the period.", "index_of_source": "People at the time didn't know what was going to happen.", "question": "What does the speaker mean by the tip \"remember contingency\" when studying the Revolution?" }, { "answer": "While generally good, the mini-series incorrectly depicts John Adams as being physically present at nearly every major early event like the Boston Massacre or the fighting at Lexington and Concord, which was not historically accurate.", "index_of_source": "The bizarre thing about the way that they depict it is, apparently according to the producers of this mini-series, if there was something happening in the early stages of the Revolution, John Adams apparently was there.", "question": "Why does the speaker caution against relying on depictions like the HBO mini-series on John Adams for historical accuracy?" } ] };

2025/6/26
articleCard.readMore

Session 1: LLM Scaling and the Role of Synthetic Data

Session 1: LLM Scaling and the Role of Synthetic Data My talk will be a lot more on the synthetic data side of things. The topic of the workshop is actually really close to my heart because, in many ways, a lot of my machine learning work has been on robustness. And actually, the reason why I’ve been studying language models and the sort of scaling behaviors of language models is because of their connection to robustness. So, to give a little bit of context, I’ve been just kind of shocked at how good language models are. I’m a natural language processing and statistical machine learning researcher. When GPT-3 came out, I think many of us, both in machine learning and NLP, were kind of shocked at just how versatile it was. If you sort of remember the days of BERT and fine-tuning models for specific tasks, perhaps you remember models that you would fine-tune on really specific tasks and then they would sort of be brittle and break once you deployed them outside their context. A lot of the large pre-trained models that we have today don’t really seem to have these same behaviors, and that was just kind of a really eye-opening moment for me because I’d spent many years working on algorithmic approaches to robustness. How can we find better loss functions? Can we train them adversarially? Can we get them to generalize out of their domain? I think the same sort of empirical goals were, in fact, very easily—or at least intellectually easily—attained through large-scale pre-training. And so that really led me to try to understand what is going on with this pre-training thing and this foundation model thing that is really changing the way that we work on machine learning and AI. Today’s talk will be sort of centered on that empirical question of, we’ve seen that these models have these really remarkable capabilities and their ability to generalize out of domain. I would like to try to understand what is going on and to use synthetic data to try to both understand and to improve these systems. With that background, I want to talk a little bit about context. We’ve seen these large language models get remarkably good, and backing that is this phenomenon that you might call scaling—increasing both the size of the models and the amount of data that is used to train them. You might remember GPT-2, which was a long time ago now. This system was good for generating text, but really not much else. You wouldn’t use it to do any sort of real-world downstream task. But then with pre-training scaling—using much more data and larger models to train these essentially auto-complete systems—we saw really predictable improvements up until ChatGPT-3.5. Much more recently, you might have heard of public talk about reasoning models and so on. That’s a different kind of scaling—thinking for longer or inference scaling. I’m really interested in this question of how predictable investments in resources like data lead to predictable gains. Often people talk about power loss, essentially polynomial improvements in model performance as a function of resource inputs. You might think of the resource inputs as more computational power, more data. I’ll talk more precisely about what these things are. These things have been really interesting because of just how consistent the gains are in the last couple of years. We’re beginning to see diminishing returns on some of these sources of scaling. But I think pre-training, for example, the task of using large scale internet data and training models to predict the next word, is reasonably efficient. You see exponents of, let’s say 0.2, to investments in compute, and they’ve held up over many orders of magnitude. One fun bit of trivia is when you look at GPT-1 and you look at DeepSeq V3, that’s about 10,000 times more compute investment. And actually the gains to that compute have been relatively predictable over that period. At the same time, I think 2025 is a really interesting time to be thinking about these questions of language models and synthetic data and where this field goes. Because I think we see lots of evidence that continued scaling, like making models bigger or dumping more compute at the problem, maybe isn’t going to solve the problem. One commentary I’ll make here is we’re beginning to really see the limitations that come from the differences between human and computational learning. We see that new pre-trained models like GPT-4.5, which was released a few months ago, are really kind of in some ways hitting a wall. GPT-4.5 is much, much larger—about 35 times more expensive than the previous models—but it’s actually not even the best model out there. Even OpenAI says, “oh, it’s actually not a frontier model.” In his test of time talk at Europe’s AI Summit, Ileus Suki had quite the way with words, saying, “pre-training as we know it will end because data is not growing.” He called it “the fossil fuel of AI.” I think this calls into question, or at least makes us really examine what is going on with this kind of learning and how we might go beyond it. One of the things that I’ve been thinking about a lot is just how different language models and humans are, both in the ways in which they generalize but also the ways in which they learn, which I think is quite important. If you think about what large language models are trained on, well, they’re trained on a large swath of the internet. The most recent QU2 model, released maybe a week ago, was trained on 36 trillion tokens. That’s just an astounding amount of data to be trained on. It’s probably most of the high quality internet they can get their hands on. There are arguments that people have made that I think are starting to be borne out in empirical observation, essentially that most of the useful stock of data on the internet is being used up for pre-training. That’s a plot on the right from EPOCH AI, where they made a projection about a year ago on when basically useful data from the internet will be used up in pre-training runs. I think they made a forecast about 2027. The interesting questions that this raises are, can we more efficiently learn from the data that we do have? And maybe importantly for a lot of the applications that these systems are being deployed at, can we adapt them to a lot of the more narrow proprietary domains that they’re being used for? Right? I think the days of saying, “we’re just gonna take as much data as we can from the internet, we’re gonna build this one gigantic generalist model by sort of building bigger and bigger data centers,” that doesn’t really seem like it’s gonna pan out, right? And so we’re gonna have to think a little bit more carefully about the nature of learning and maybe adapting these models to specific domains. So if you go and talk to folks from OpenAI or philanthropic organizations about this current situation, you know, they’ll say, “oh, that’s fine, that’s fine. Don’t worry about pre-training hitting a wall or us running out of data or algorithms being horrifically inefficient.” Okay? That’s all fine because there are these things called reasoning models, which just kind of think for longer at deployment time. And if you do that, then, well, you know, it’s going to be okay. Our models will be smarter; we’ll have a new axis of scale. I think the thing that’s been frustrating for me over these past few months, ever since OpenAI released GPT-3, was there’s really no great science, or at least until very recently, about what this thing actually is. I like to sort of bring this point up in every talk I give, but you know, this left plot is what sort of OpenAI released when they announced GPT-3, and maybe you’ll notice that their x-axis on their plots don’t actually have labels anymore. You know, in the interest of secrecy, they can’t even tell us how many flops are used in each one of those dots when they do this training. And so I wanted to really, a few months ago, understand what this object was, right? Like if reasoning supposedly is the future of my field, it behooves me to understand what that exactly is. And so I want to talk about sort of two major things, two projects that I’ve been working on in the context of synthetic data. Synthetic data is really interesting to us because it allows us to sort of intervene on these models in very precise, targeted, and data-centric ways. I’m going to use that in the first part of my talk to try to understand these reasoning models by coming up with different ways of making them in very simple ways. In the second part of my talk, I want to try to use synthetic data to make learning much, much more data-efficient. So this is sort of improving models using synthetic data. Since this is a workshop with really diverse interests and really diverse topics, I’m going to try to make connections whenever possible to kind of the biological aspects of robustness as I go. So, I think this discussion about reasoning models has particularly been interesting to me because it’s taken a really pseudoscientific tone. Sometimes I’ve taken sort of Jensen Huang’s keynote because I think this sort of captures a lot of what I’m frustrated by, including the fact that this plot doesn’t really make sense. You know, there’s three phases of scaling, and reasoning scaling’s on the right, but somehow the reasoning scaling’s the worst of the three. And you would choose that one because I’m not sure why. Many have argued that this new notion of thinking harder is kind of the way the future of this field will go, right? And what does that really mean? And I think we’ve seen lots of suggestions. that this is a really qualitatively different object, that this is really going to be powerful. And I think sort of as humans, we think of reasoning as like a really core capability, right? No matter what new domain that we go to. For example, if I were to try to learn law tomorrow, right? I would use my reasoning capabilities to be able to learn that more quickly. I would use those capabilities to try to operate well in that domain. And so this is sort of a really important question to be able to say, what is this new sort of general capability that we’re acquiring here, right? I think one thing that was very interesting to us when the first reasoning models were released was that it really did at the time look like a totally different thing. So on the left side of this slide is opening eyes 01, when it was released. And as I was saying at the beginning of this talk, one of the interesting things about this is this notion of scaling. On the x-axis is the amount of compute I spend, the time I spend thinking. The y-axis is the performance. And it’s predictable, right? If you look at this plot, you might be sort of led to believe that, oh, I can just keep thinking for longer and my accuracy will continue to go up and up and up, on the right side. Several folks, including EP, did some early sort of investigations into this model, and they sort of found that, wow, these new reasoning models can attain performance that is very far out of reach from models that are using different kinds of inference scaling. So in this case, you can think for longer by just sampling more outputs from the model and then taking a majority vote—that’s the teal line that’s on the bottom—or you can ask the model to continue revising its answers. That’s kind of the blue line that you see there as well. And those two lines sort of don’t have the same kinds of scaling qualitatively as this O1 model. And so we really wanted to understand what was going on. One of the things that’s interesting to us is there’s a potential that these kinds of reasoning models are a totally different ballgame in the same way that sort of AlphaGo and search algorithms and so on are very, very different from the machine learning systems that we normally use and operate today. I think public discussion of mirrors a lot of this very often. I’ve taken a little clip from Yahu Bengio of a public-facing opinion piece in the Financial Times, where he says, “these models, when they spend more time thinking for longer, they get much, much better.” And I think one thing that’s very, very interesting—and this is a subtle point, but I think an important point—is whether these models are really extrapolating, right? So this is the question of in some ways robustness and generalization. If you look at models like AlphaGo, you’ll see that they’re trained with a certain amount of training compute. And when they’re deployed, they spend a lot more computational power sort of doing search, right? So their inference compute is much, much larger than their training time compute. In many ways, they generalize very well at inference time using search algorithms. On the other hand, language models don’t have this property, or at least current generation language models don’t have this property, right? They’re trained with a certain amount of compute, and at inference time they’re deployed. and they use that exact same amount of compute. They don’t really think for longer unless they’re using more sort of tokens by explicitly sort of verbalizing their thinking, right? This is of course an important distinction. I’ll sort of touch on this as I go on. You see this even in AlphaGo. In AlphaGo, they show that if you do search, you get this nice blue tall bar. This is the AlphaGo Zero model that sort of beats it all, and then you have the raw network, which doesn’t do any search that’s thinking for as long as it was trained, and it’s much, much worse than kind of the best models. So my students and I were very interested in this gap. Are these models really truly generalizing? A reasoning model should be able to think for much, much longer than the situations it was trained for and get much higher performance. If, on the other hand, what we’re doing is just training the model with more and more compute, it’s of course reasoning training, so they’re going to get better at math. But they may not generalize. That does really qualitatively change the nature of what we think of with these reasoning models. I’m going to take a moment to discuss the science of science for a moment here. One thing, as I said, that’s very frustrating is that a lot of these processes are closed source and they’re also very, very computationally expensive. When OpenAI initially came out, we knew very little about what they had done. They were so cagey that they didn’t even really say whether they did reinforcement learning or they did search or anything else. In this sub-part of the talk, my goal will be kind of like a biologist. What I’m going to look at is the phenomenon, the empirical phenomenon that we see externally. I’m going to try to replicate this. So I’m going to care about the phenomenon and not necessarily some underlying process that generated this. I don’t want to match OpenAI; I want to try to understand this particular scaling phenomenon. To be concrete, I’m going to come up with some method that will give us inference scaling from 20 to 80% accuracy on AMY. This is the high school math exam, and I want to do dramatically better than just majority voting on my models. Several of us set out with this goal to try to understand reasoning models and whether they’re generalizing and extrapolating and so on. We tried a lot of different things. We used a bunch of reinforcement learning algorithms. We used all the search algorithms that we found that people had worked on, but none of those really worked very well. One of the things that I learned again and again in machine learning is that very simple things often work really well. Data-centric interventions also work really, really well. What had worked in the end was to basically come up with a really simple synthetic data-style data set. So what we did was we went out and collected a small number of math and reasoning data sets. We had gone out to Newman and several other high-quality sources of math questions, and you can see the diversity of the questions that we collected here. One of my students… Zong is a stat student. And so he and Emmanuel were interested in the question of can these AI systems do the statistics qual? So they threw that in there too. We ended up with about a thousand math and reasoning questions that we basically used as a high-quality curated dataset. Then what we did was we performed the simplest possible thing on this dataset, which was to look or go and take an existing reasoning model, in our case Gemini 2.0 flash thinking, and take the verbalized reasoning traces that these models output. We then just trained our model, which in this case was a QU 2.5, 32 B model. We fine-tuned that model on these reasoning traces. This is a very, very lightweight intervention. We’re taking an open-source or open-weight model that doesn’t really do much of the reasoning steps. We take the reasoning steps from an existing thinking model, Gemini 2.0, and then we just do a little bit of fine-tuning on a thousand examples. So we’re not fundamentally going to teach it new capabilities with that little data. What we get is actually quite surprising. At the time, we thought it was pretty remarkable. With very few examples, we achieve math benchmark accuracy that is remarkably good. We’ve gone from Quin 32, which is 84% on math, to 93% on Amy. The difference is more dramatic going from 26% to 56.7%. Just a thousand math science questions and a few long chains of thought from Google’s Gemini model give us this remarkable bump in performance. However, my core interest wasn’t solely in this outcome. I don’t necessarily want to just build models that can answer math questions. OpenAI is certainly much better than me at doing that. My interest is in understanding extrapolation and generalization. As I said before, at the start of this part of the talk, my interest lies in understanding scaling. Is it possible for models to think for longer and then do better on their benchmarks? One of my students had a very simple idea: we want the model to think for longer. Could we force it to verbally think for longer? To give some context, when a language model is asked to think through a problem, we use chain of thought (COT), which means it verbally expresses its thinking process. We try to hit a target length for COT, so we force it to think for a particular length of time. If, for example, the actual chain of thought for a model is shorter than that target, we suppress the “end of thinking” token, which is a special marker that switches the model from thinking to answering, and then we output the word “wait.” This is just a verbal trick because once the model says “wait,” it starts rethinking the previous trace and continues on. If our chain of thought length is above the target length, we truncate the chain of thought. We emit the end of thinking token, which forces the model to immediately switch to the answer. This nearly perfectly enforces the desired chain-of-thought (COT) length, which is unlike simply prompting or training the model to achieve a particular COT length. We attain our goals: this approach, called budget forcing, gives very smooth scaling as we saw before. I’ll go through each of these two panels in turn. The very first goal was to try to replicate the plot from OpenAI O1, where the x-axis represents the thinking time—the computational budget spent for these models to think—and the y-axis shows accuracy on a math benchmark. On the right is the OpenAI O1 blog post; on the left is what we got from this budget forcing and S1 distillation approach. What we see is both good and bad. Qualitatively, the plot looks just like the OpenAI O1 plot, but if you look closely, it’s actually a little disappointing. Around 6,000, this is the model where we don’t control its thinking process at all—that’s its natural COT length. By using this approach of forcing it to think for longer, we can bump the performance up, going from 6,000 to a little bit higher on the right-hand side by doubling four times and then six times in its budget. We do see a little bit of extrapolation beyond its training point. In some ways, that’s good—I was interested in extrapolation, and I did get some. But what I had hoped at the very beginning of this project, and what would have convinced me that this is a truly new paradigm, is if we could start at the very left bottom: get a model that’s really bad at math, that thinks very little, and then by forcing it to think for longer and longer, get all the way to 60%. Extrapolating from the very left bottom all the way to the right top would have been very remarkable. What we found was that a lot of what was happening on the left side of this scaling plot is actually just truncating the thinking process. Of course, I can’t say this is what OpenAI or others are doing—we don’t really know, and they’re doing various kinds of RL, which is a qualitatively different phenomenon. But one thing is quite clear: modern language models, even with pre-training and without any reasoning training, are good enough to get pretty high scores on these benchmarks. A lot of the scaling plots that we see are not the result of robust extrapolation in the amount of thinking budget. It seems to be more about small amounts of extrapolation with large amounts of early exit or stopping thinking. To complement the other plot, I asked, “What were my goals?” My goal was to produce two plots: one that replicates the smooth scaling, and another that matches the EPOCH K report comparing sequential scaling—thinking for longer—versus majority voting. We can replicate almost exactly that plot as well. That was really satisfying because we had set out with a very clear goal: can we come up with a really simple method that recapitulates qualitatively what’s going on here? And we managed to do that. I want to end with two other interesting observations we made that I think others have also made since then. The first one was, what other things could we do? The approach we took was very synthetic data oriented. Could we come up with some constructed chains of thought? Could we put that into the model? Can we use that to control the output of the thinking time of the model? One of the first things we tried, though, was to just use a normal model, have it think for a little bit, and then rejection sample, right? Just keep sampling until you get a sort of thinking trace of the appropriate length. One of the funny things that we found was actually if you do this, performance goes down as the model thinks for longer. This is because the longer CTs are far out of distribution. Even though we see a lot of robust generalization with these modern language models, one of the things we still see is a lot of out of distribution effects. I think my results are very much consistent with the fact that these models really aren’t magically generalizing. If you just sample for longer co-keys, they’re qualitatively much worse than the previous slides I showed you. If you try to force the models to think for longer, they’ll do a little bit better, but then they’ll hit diminishing returns pretty quickly. If you apply things like majority voting or search, that can continue the scaling behavior, but the rate at which things scale is not particularly favorable. So I don’t think that is looking pretty literally good. So, what’s the point of all of this? I think in many ways the most optimistic story for a lot of these reasoning models involves extrapolation. We train for a fixed amount of time and then we deploy them, and we spend a lot more inference time compute. Due to extrapolation, they’re able to solve way harder problems than they were trained for. I think that’s the most optimistic case for these systems. As far as we know, there’s not really evidence for that kind of extrapolation and scaling. These models are remarkable. We know they work; we know they’re hitting really just new high watermarks on math and reasoning tasks. But maybe they work really well, not because they extrapolate in the sense that they think for much longer at deployment time, but maybe because the kind of post-training that they receive, the math training that we have developed using reinforcement learning algorithms, is really, really good. What that means is, maybe a lot more compute is spent during training time rather than deployment time. I think that’s a really interesting sort of conceptual thing to be thinking about and keeping in mind. So that was about inference scaling. Now I want to talk a little bit about the original paradigm. I’ve been talking about the new paradigm, so to speak, of reasoning models, but now I want to talk about pre-training. Pre-training is interesting because despite its astounding success in bringing us models like GPT-4 and DeepSeq V3, we also know that this is actually a crazy learning algorithm. We’re learning to predict the next word, and we’re using that as a way to solve all sorts of downstream tasks. And the data inefficiency of this approach is really staggering, right? We’re using trillions and trillions of tokens to get models that are less knowledgeable than a graduate student in niche domains. So once we begin to run out of data, what are we gonna do? We need to make language models much more data efficient. But that’s really hard, right? We need new kinds of algorithms, we need new kinds of ideas to do that. Some students and I came up with a kind of interesting setting in which we can study some of these questions in detail without having really large-scale compute. This is the setting of continued pre-training in which you’re given a pre-trained model F. The goal is going to be, I want to teach this model facts within a niche domain. Imagine you have a textbook, like a neuroscience textbook, and you want to teach the facts from this textbook into the model, right? This is a very data constrained setting. I only have a few textbooks, and I need to come up with new learning algorithms that can enable this. If we can achieve that, we will have developed significantly more data-efficient algorithms, and we will have new domain adaptation algorithms that allow us to generalize to new domains. I think this is a very exciting problem. In general, I’m not the first to think about this problem setting. Continued pre-training is a very classic area of research. Domain adaptation is an even older and more classic area of research. However, I think continued pre-training has been an area that has been very challenging to work on because there have been really successful examples of people working on medical or code or math models, but they’ve required billions and billions of tokens. I think the smallest successful case of this has required something like 15 billion tokens. As a sort of thought experiment, if we’re really interested in having models learn from very little data like a single book, we really need to push the limits of this kind of continued pre-training. I want to try to train models with 10,000 times less data on about 1.3 million tokens. This is a collection of about a hundred books or so. Why is this a difficult task? It’s because pre-training is in many ways a very unhuman-like way of learning. It’s incredibly data inefficient. What we’re doing is learning to predict the next token, and the sort of representations learned by that in the model are supposed to be useful for downstream tasks. But it turns out that this process isn’t as data efficient as we had hoped. I’ll give you one example of this. There’s a phenomenon known as the reversal curse, in which if you train a model and the model knows about or has been trained in its pre-training dataset, it knows the facts like A is B, but it doesn’t necessarily know the reverse, B is A, even if they’re equivalences. In this case, I have an example. I’ve taken the abstract of my paper: “synthetic continued pre-training is X and Y,” in the auto-aggressive direction. The model probably learns this very well. Like, if you ask, “What does synthetic CPT do?” That’s an easy task for the model because it’s been trained to take in synthetic CPT on the prefix and output its definition on the suffix, right? But if you reverse it and say, here’s the definition, what does this correspond to? That’s a much harder task, right? And so the algorithms and the methods that we used to train these models really put limitations on the efficiency by which we can learn. The empirical results really do reflect this. So, we take our a hundred books or so and we have chosen a data set such that those books come with their corresponding question answering questions. Let’s say we just do continued pre-training. We take those books, and we just learn to predict the next word. Then actually we get the green line, which is worse than the dashed black line. The dashed black line is our starting model. We train it to predict the next word on our books. We got this green line, which is actually just worse than what we started with. That’s because there’s just not enough tokens here for the model to learn anything useful. But we know pre-training works, right? We know that there has to be a way to bridge this gap. If we could take the dataset that we do have—these books—and we could rewrite them in various ways such that it sort of matches the diversity of pre-training data, the stuff that we see on the internet, then we should be able to actually teach these models these facts robustly. An analogy I like to use here is we know that models know all really detailed facts about Harry Potter. But there’s only really one Harry Potter book, so to speak, one original book. It’s just been analyzed, rephrased, and rewritten in many different ways on the internet. If we could replicate that process of taking a single source of knowledge and then expanding it via augmentations, then it might be possible even with existing learning algorithms to get much more data-efficient algorithms, right? So, this is the idea of synthetic continued pre-training and the idea of using synthetic data to really change the data efficiency of these models. The goal here is to increase the diversity of what we have, rather than trying to improve the compute efficiency of models, or to fine-tune a really task-specific language model. My goal is still to have a generalist model. It’s also to inject a niche domain of knowledge into it. One of the things that we could do is something pretty naive yet effective. You can just ask a language model—in this case, GPT-4—to go and paraphrase these books repeatedly. What that gives you is a very naive but useful augmentation. This augmentation really does work. On the X-axis here is the amount of synthetic data that we’re using to train the model. The blue line now is repeatedly rephrasing random parts of the book. As we increase the amount of rephrasing of the same content, the question answering accuracy of the model trained on this improves and improves. One way of thinking about this is that synthetic data gets a lot of mystical discussion around it, but this is really just data augmentation. In 2025, we’re coming up with good ways of adding variation and variances to our model by rephrasing the original data that we have. And that gives us significant scaling improvements, which is quite nice. But then we sort of run into a problem, which is that LMS are not terribly diverse. If you’ve interacted with Chat GPT or any of these systems, I think you’ve had the experience of these models outputting a few things, and then it repeats the same stuff after a certain point. That’s been our experience, as well, and the experience of many others working in this area of using LMS to generate data. One of the experiences that I had was with some colleagues and a code vice student of mine on using LMS to see if it could help us generate novel new ideas. One of the things we found was initially it can come up with a bunch of interesting ideas, and then about 500, after 500 of those, all the ideas that it generates past that point are duplicates of things that it has already generated. So there’s just not that much stock of new novel ideas in a language model. An increasingly common thing that people have done over the past few years is to essentially inject external sources of randomness into this generation process. LMS are not very good at being spontaneous and random in many ways, just like any single person is. And so we inject external randomness. In this case, we’re going to take sort of inspiration from knowledge graphs. What we will do is we will take each document or each chunk of a book, and we will enumerate all of the entities that appear in that document. You can imagine this as sort of listing out the entities in a knowledge graph. We’re not going to explicitly construct this knowledge graph, but you can sort of mentally keep that picture in your mind. Then we’re going to sample random entities in this list of entities that we have. You can think of this as sampling pairs, trios, or some subgraphs of this knowledge graph. Then we’ll take a language model, that’s sort of large and been pre-trained, a domain-agnostic one. We’ll ask it to synthesize the relationships between the entities that we selected. To give an example, imagine we have this document talking about Mona Lisa, Da Vinci, and the Louvre, and some people visiting these places. The document may never explicitly mention the relation between, let’s say, Da Vinci and the Louvre, but if I pick these two random entities and ask the LM to discuss their relation, it can see that Da Vinci painted the Mona Lisa and Mona Lisa is in the Louvre. That’s all in the document, and it can now explicitly construct the fact that the Louvre contains many works by Da Vinci. This is essentially taking implicit facts that are implied by the content in the document, and then making them explicit so that even an inefficient learning algorithm can learn them very easily. So we’re sort of surfacing facts that are very difficult to learn very easily. If we use this approach, which we call graph, we see very nice scaling performance improvements. Importantly, we significantly exceed the performance of GPT-4, which was our teacher model here. This is not really just distillation effects where we’re just kind of learning things. That the teacher knows, this augmentation is really doing much better, in terms of knowledge. One thing that was exciting to us was that you can use this thing to essentially have models that generalize just like a LLM. The model now has this knowledge, and you can use it in context outside of QA. It can do summarization; it can use that knowledge in other sorts of questions that are not related to the book. It can do also even surprisingly weird things like compare and contrast two books that appeared in the dataset, even though we never explicitly trained the model to be able to compare and contrast some of these books. In the last few minutes, I want to talk a little bit about trying to put this on maybe firmer foundations. I think, as I said earlier, synthetic data has this sort of mystical quality. Whenever there’s a hard problem, people are like, “Ah, yes, synthetic data will fix it, don’t worry.” That has bothered me a little bit. So I wanted to, along with my students, think a little bit about what is actually happening here. My point of view, at least for this style of synthetic data, is that it’s really data augmentation. We’re taking a document, we have a Fraser, and we’re just extracting the invariances out from the Fraser, and we’re getting augmentations of the original data. So what does that mean? Imagine we have a graph, a knowledge graph that contains direct mentions of entities. The graph algorithm we propose is really just taking implicit facts and making them explicit. How much extra knowledge might you be able to extract through this process? This is equivalent to essentially just filling in a graph on some sort of random graph. You start out with a random graph, you pick two random entities, and you try to connect the two. Whenever there’s a path, you end up being able to say something about those entities. You make a direct connection. We assume that there’s no generalization in these facts; learning is all just memorization. We’re able to show something basic but somewhat interesting, which is to say there are both optimistic upper bounds and nice limits to this process. If you start with something like a URD reny process where you connect facts between random entities, you can sort of fill in all of the connected components. This gives you limits on what you can do with synthetic data, but those limits are actually pretty loose. If you have a really nice connected graph, you can learn all of the different implicit facts that are contained within it. I’ll go over this very briefly, but you can use this to essentially get a kind of a new scaling law, which is a mixture of exponentials. It fits the actual observed data. Although this hasn’t really been carefully tested in an extrapolation setting, this is really just curing. I’m just showing you that the curve fits well. Putting this together, I think the really interesting question that this raises is two different observations that maybe you can take home. The first one is just the staggering difference between human-like learning and LLM learning. With humans, you put a book in front of them and you ask them to read it and learn it carefully. They can kind of learn from a single book, right? If they really, really want to. Whereas with a language model, they can’t really learn from a single book by just predicting the next word, no matter how many times you loop through that book, right? It’s just a staggeringly different thing. The other thing I want to emphasize is that this is not kind of a fundamental limitation, right? Like synthetic data, even just basic things like rephrasing can really extract a lot more data efficiency and allow us to adapt to these new kinds of niche potential domains, which I think is a really interesting thing. And so this is an exciting test bed for data efficient language modeling, but also I think it’s hopefully showing that there’s a lot more upper bound to pre-training than maybe we have seen. So, I want to end very briefly with some sort of more speculative, but hopefully fun things that we’ve done since then. Both of the works that I’ve described exhibit this flavor of using a really strong model to be able to make little models better. Of course, in the second part of the talk, the little model ends up being stronger than its feature, so that’s good. But really, we would like to ideally understand the bootstrapping process where we’re not relying on a stronger model at all, right? That would be the relevant context for a lot of these companies like OpenAI. One of the things that we started to do with some colleagues at the University of Toronto, Chris Madison, and a student, Youngen, was to essentially see if we could bootstrap this thinking process. We have some observed data, which is a very short compressed summary of a latent thought process. The thinking is, if we could sort of identify this latent thought process using classic ideas from latent variable modeling, could we get more data efficient learning algorithms? So this involves using some of our observations about synthetic data and then asking, can this lead to fundamentally more data efficient algorithms? I know I’m short on time, so I’m gonna skip some of these processes. One of the things is that it does kind of work—this process of essentially having a model try to identify some latent thoughts and then training on those latent thoughts, and repeating that process can give us improvements in model accuracy in data-bounded settings. Okay. So, just to put everything together and sort of to connect this to the bigger picture, I think it’s been kind of remarkable being in language modeling and thinking about these scaling questions. A lot of the really hard questions of generalization have been addressed differently. These models made progress by having more data, by doing pre-training, and adaptation; they were able to solve many of these multitask problems that I thought were out of reach for many, many years. Today, I wanted to really study some of those phenomena. In the first part of this talk, we showed that synthetic data is a really useful tool to try to understand some of these very mysterious objects like reasoning models. In the second part of the talk, I wanted to really try to push things and ask, can we build much more data efficient models? We showed that it’s possible to do so. by essentially coming up with new data augmentation style, synthetic data methods. Thank you. Hi, thanks for the great talk. So, I’m wondering to what extent you think that catastrophic forgetting is a big issue in reasoning training? ‘Cause we’re really focusing reasoning training on math and code data. My assumption is that knowledge and other sorts of domain are just being distorted. So yeah, should we be benchmarking this? Does this matter? Should we intervene in training or should we just use different models? Yeah, I think catastrophic forgetting is, I would say, both a problem and not a problem in the following sense. I think a lot of the moderate LMS are really, really large. They’re over-parameterized in many ways. And so it’s possible, I think, for example, like new reasoning training and the pre-training to live together. But I also think that kind of recent trends in how people train these models are gonna make that a more important problem in the future because the models are getting smaller due to inference cost concerns, and also post-training, which is how you inject these reasoning capabilities, is becoming larger scale. And so more and more compute and more and more model updates are being spent there. So I think up until now it hasn’t really been as much of a concern. Like a lot of these updates have been lightweight; these models are really big. But as people really try to push the size of the models down and do a lot more reinforcement learning (RL), we will have to think a lot more about it. I think thus far a lot of those problems are solvable with more compute by replaying pre-training as you do a lot of the post-training stuff. It’ll be interesting to know whether there’s limits to that or if there are more computationally efficient ways we can do that. Thank you. Thank you. And hello. So I thought to conclusively know about where the test time scaling works, we should look at more than 8,000 tokens. I wonder if your group has done that. Right. So I mean, we have certainly scaled that plot to the right, but we already have diminishing returns at 8,000 tokens. And so, you know, I can draw you the flatter plot. It’s flatter. Let me sort terminate that plot at 8,192 because there’s no gains past that point. Alright. Yeah. ‘Cause that was what Limo and they found too, right? Is the baseline where at the bottom graph, you only run it through the books one round? You mean the rephrasing? So the flat curves, did you just—oh no, we optimized over epoch and replay and all the hyperparameters that are sort of relevant. So like if I have like a million tokens in the books, did you just run it through 1 million or did you run the books many times? No, we optimized over the number of passes over that. So like, I think over like the grid of like one to ten, we checked every epoch count and then we picked the best one. So no matter how many times I run through the books, it’s all completely flat. Yes. Okay, cool. Thanks. Or, or it’s not flat, actually; it gets worse. Okay. Great talk. I have a lot of questions. So one, did you reference “textbook is all you need” though? Did you guys research that? Yes. Or yeah, yeah. So, so we’re aware of ~fi~ Although, maybe I can channel my frustration and say they don’t really release very much like OpenAI either. We talked to, said, pretty early on, for a different context and we were like, “can you give us your synthetic data?” And he was like, “no, sorry. It’s all proprietary.” So, we’ve been in touch with them, we read their papers, but it’s sort of a different jam as well. Got it. And the second question, how this compares to RAGs? Yeah, okay. That’s a good question. I didn’t put that slide in the interest of time, but I think there are sort of two things I’ll say about RAG. The first one is, scientifically, the question that we’re interested in here is, putting knowledge and parameters. We think that’s interesting for its own goal. That’s one thing I’ll say, but the second one that’s cool is this intervention composes with RAGs. So, I think we get a 2% to 3% accuracy boost on top of a really, really strong almost Oracle RAG retriever in this setting. You would want to do this if you’re resource-rich, even if you’re doing RAG. Then the gains that we get through synthetic data and continued pre-training is roughly 80% or so of the gains that you get with RAG. So we’re getting most of the gains that you can get out of a strong retriever. Yeah, thank you Tatsu. Great talk. I have sort of a broader question and a more specific one. It seems like a lot of the recent work is focused on trade-offs between data and compute efficiency. I think it’s really interesting the same approach here where we can use less data but use more compute and get those gains. I guess my broader question is, what are your thoughts on trying to improve data efficiency without needing to incur such a large computational cost? I think you touched on this a little bit towards the end of the talk, but we would love to hear more. And I guess my more specific question is, when you see these diminishing returns around 8,000 tokens, do you think this has anything to do with the context size? Like if you triple the context size, can you then improve the length of your change of thought that you can use to get better improvements? Yeah, so I think—I guess I’ll— the first one was about sort of computationally efficient data-efficient models. I think that’s interesting and in some ways I think that’s the right approach. We know that humans, when we learn, aren’t doing this kind of crazy, like, “let’s generate a billion extra synthetic tokens” and then learn on that data inefficiently, right? I think this is kind of the safe way to get it to work today. But I think long-term, the right thing to do is to think about architecture or algorithm interventions that get us more data-efficient learning. That’ll be lovely. I just don’t think right now we have even the right paradigm for that yet. For the second thing, I do think it would be really useful and interesting to be able to get this kind of longer extrapolation. But, wait one second. Let me think about this. I think the issue is that we haven’t really seen very much extrapolation, even in loop transformers and other kinds of new interesting ways of extrapolating reasoning thinking. We’ve seen very similar issues. of stopping roughly where you’re training compute is. And so once we have paradigms that go beyond that, then maybe it’s possible to think a little bit differently. So, I have a much more informative kind of question. I understand what you meant by synthetic data in your second part, which was this augmentation, but what was the synthetic data part in the first part of the talk? Yeah, so in that part, what we’re doing is we’re essentially using Gemini flash thinking to construct the reasoning traces that we then train on to train the model. So it’s just distillation, right? I mean, synthetic data I think gets used in several different contexts. We actually have all three kinds in the stock. One of them is distillation, which is part one. Another one is kind of data augmentation, which is part two. And actually the last kind is self-training, which is the bit that I skipped at the very end there. Okay. Thank you for the great talk. I found you— the way you use the knowledge graph to augment your data— very interesting. But compared to human reading of a document, when we read the document, we recommend not only by generating additional relationships between the entity, we recommend by combining with our common sense knowledge base, our awareness of the surrounding world, the world knowledge graph, right? Wikipedia— our life experience. So, have you thought about combining the knowledge graph in addition to just a single entity graph extracted from that domain document? Yeah, I think that approach of thinking about having the model put in everything that it knows is the right, like broader approach to this. I think the knowledge graph approach was what we did primarily because it was kind of cleaner and also we were very, very worried about distillation effects in that second part. We’re using a very strong model to generate data for a weaker model. And so if you let the strong model just use all of its knowledge about the world, then now you’ve got sort of information leakage. And so we wanted the model to be as kind of an agnostic rephrase as possible and only use the data in the document as a proof of concept. But I think if we were to go and like kind of deploy this out in the world, we would do exactly what you’re suggesting— say, interesting to say empirically. That’s right. Last question. So, in the last approach to the synthetic data, which is bootstrapping with the same model, how do you not hit the data processing inequality limits? Yeah, I mean, okay, so I’ll just answer the data processing inequality limit part in general rather than the specific project that I didn’t talk about. I actually think data processing inequality arguments are very, very weak ones for synthetic data. Mainly because if you think about the information theoretic limits of pre-training on the internet, for example, that’s actually really high, right? It’s like every fact on the internet, plus every fact entailed by fact on the internet is the information theoretic limit. And so if we’re making information theoretic limit arguments, I think the bounds are very, very loose. That’s at least my opinion for sort of synthetic data bounds. Whether we can attain them is a totally different story, but I do think those aren’t really providing strong limitations, at least for knowledge aspects of synthetic data. And a follow up question… Do you think the similar argument applies to vision models and sort of a multi-model sort of reasoning? Yeah, I think vision models are interesting because, in many ways, they started out much more ahead in terms of data augmentation. If anything, language people are learning from vision people about all these data augmentation things. But I think the generative models we have today, now the situations are a bit reversed. Text models are much more controllable, precise generators than image models. I do think with the newest generation of auto-regressive image models that are much more controllable, it might be the case that we’ll see a lot of like image synthetic data work as well in the same vein. window.tocIndex = { "index": [ { "index_sentences": "My talk will be a lot more on the synthetic data side of things.", "section_level": 1, "section_title": "Introduction and Motivation" }, { "index_sentences": "We've seen these large language models get remarkably good, and backing that is this phenomenon that you might call scaling—increasing both the size of the models and the amount of data that is used to train them.", "section_level": 1, "section_title": "The Phenomenon of Scaling and Emerging Limitations" }, { "index_sentences": "I think the thing that's been frustrating for me over these past few months, ever since OpenAI released GPT-3, was there's really no great science, or at least until very recently, about what this thing actually is.", "section_level": 1, "section_title": "Understanding Reasoning Models" }, { "index_sentences": "I think one thing that's very, very interesting—and this is a subtle point, but I think an important point—is whether these models are really extrapolating, right?", "section_level": 2, "section_title": "The Question of Extrapolation" }, { "index_sentences": "What had worked in the end was to basically come up with a really simple synthetic data-style data set.", "section_level": 2, "section_title": "Replicating Reasoning via Synthetic Data Distillation" }, { "index_sentences": "One of my students had a very simple idea: we want the model to think for longer.", "section_level": 2, "section_title": "Budget Forcing for Controlling Thinking Time" }, { "index_sentences": "This approach called budget forcing gives very smooth scaling as we saw before.", "section_level": 2, "section_title": "Replicated Scaling Plots and Observations" }, { "index_sentences": "I want to end with two other interesting observations we made that I think others have also made since then.", "section_level": 2, "section_title": "Further Observations on Reasoning Models" }, { "index_sentences": "Now I want to talk a little bit about the original paradigm. I've been talking about the new paradigm, so to speak, of reasoning models, but now I want to talk about pre-training.", "section_level": 1, "section_title": "The Challenge of Data Efficiency in Pre-training" }, { "index_sentences": "Why is this a difficult task? It's because pre-training is in many ways a very unhuman-like way of learning.", "section_level": 2, "section_title": "Data Inefficiency and the Reversal Curse" }, { "index_sentences": "If we could take the dataset that we do have—these books—and we could rewrite them in various ways such that it sort of matches the diversity of pre-training data, the stuff that we see on the internet, then we should be able to actually teach these models these facts robustly.", "section_level": 2, "section_title": "Synthetic Data for Data-Efficient Continued Pre-training (SCPT)" }, { "index_sentences": "But then we sort of run into a problem, which is that LMS are not terribly diverse.", "section_level": 2, "section_title": "Enhancing Synthetic Data Diversity via Knowledge Graphs" }, { "index_sentences": "If we use this approach, which we call graph, we see very nice scaling performance improvements.", "section_level": 2, "section_title": "Results and Generalization from Graph-based SCPT" }, { "index_sentences": "So what does that mean? Imagine we have a graph, a knowledge graph that contains direct mentions of entities.", "section_level": 2, "section_title": "Towards a Theoretical Understanding of SCPT" }, { "index_sentences": "Putting this together, I think the really interesting question that this raises is two different observations that maybe you can take home.", "section_level": 2, "section_title": "Summary and Future Directions for Data Efficiency" }, { "index_sentences": "So, I want to end very briefly with some sort of more speculative, but hopefully fun things that we've done since then.", "section_level": 1, "section_title": "Speculative Future: Bootstrapping Learning" }, { "index_sentences": "Okay. So, just to put everything together and sort of to connect this to the bigger picture, I think it's been kind of remarkable being in language modeling and thinking about these scaling questions.", "section_level": 1, "section_title": "Overall Conclusion" }, { "index_sentences": "Hi, thanks for the great talk. So, I'm wondering to what extent you think that catastrophic forgetting is a big issue in reasoning training?", "section_level": 1, "section_title": "Q&A" } ] }; window.faq = { "qas": [ { "answer": "The speaker's machine learning work has largely focused on robustness, and they study language models due to their connection to this field.", "index_of_source": "And actually, the reason why I've been studying language models and the sort of scaling behaviors of language models is because of their connection to robustness.", "question": "Why has the speaker been studying language models?" }, { "answer": "They were shocked by the versatility of GPT-3, contrasting it with older models like BERT that were brittle and required fine-tuning for specific tasks.", "index_of_source": "When GPT-3 came out, I think many of us, both in machine learning and NLP, were kind of shocked at just how versatile it was.", "question": "What was the speaker's reaction to the release of GPT-3?" }, { "answer": "Scaling involves increasing both the size of the models and the amount of data used to train them, such as seen between GPT-2 and ChatGPT-3.5.", "index_of_source": "We've seen these large language models get remarkably good, and backing that is this phenomenon that you might call scaling—increasing both the size of the models and the amount of data that is used to train them.", "question": "What does the speaker mean by 'scaling' in the context of large language models?" }, { "answer": "They are seeing diminishing returns on some sources of scaling, suggesting that simply making models bigger or using more compute might not solve the problem going forward.", "index_of_source": "We're beginning to see diminishing returns on some of these sources of scaling.", "question": "What challenges or limitations are becoming apparent with current scaling approaches?" }, { "answer": "According to Ileus Suki, pre-training as we know it will end because the necessary data is not growing and is finite, referring to data as the 'fossil fuel of AI'.", "index_of_source": "In his test of time talk at Europe's AI Summit, Ileus Suki had quite the way with words, saying, \"pre-training as we know it will end because data is not growing.\"", "question": "Why is it suggested that pre-training may come to an end?" }, { "answer": "This raises questions about learning more efficiently from existing data and adapting models to narrow, proprietary domains, rather than solely relying on vast internet data.", "index_of_source": "The interesting questions that this raises are, can we more efficiently learn from the data that we do have?", "question": "What key questions arise from the potential exhaustion of useful internet data for pre-training?" }, { "answer": "The speaker finds the discussion frustrating due to its pseudoscientific tone, lack of great science explaining reasoning models, and the secrecy around their training details (e.g., missing plot labels on axes).", "index_of_source": "So, I think this discussion about reasoning models has particularly been interesting to me because it's taken a really pseudoscientific tone.", "question": "What aspects of the public discussion around reasoning models does the speaker find frustrating?" }, { "answer": "By collecting a small, high-quality math dataset and fine-tuning an open-weight model (QU 2.5, 32 B) on verbalized reasoning traces from a stronger model (Gemini 2.0), they significantly improved accuracy on math benchmarks like AMY.", "index_of_source": "To be concrete, I'm going to come up with some method that will give us inference scaling from 20 to 80% accuracy on AMY.", "question": "How did the speaker and colleagues achieve significant improvements on math benchmarks using a simple synthetic data approach?" }, { "answer": "When using rejection sampling to generate longer chains of thought, performance unexpectedly decreased because the longer COTs were 'far out of distribution' for the model.", "index_of_source": "One of the funny things that we found was actually if you do this, performance goes down as the model thinks for longer.", "question": "What counter-intuitive result was observed when forcing models to think longer using rejection sampling?" }, { "answer": "Pre-training is incredibly data-inefficient because it learns by predicting the next token, and this process requires trillions of tokens to be useful, unlike human learning which can absorb information from limited data like a single book.", "index_of_source": "Why is this a difficult task? It's because pre-training is in many ways a very unhuman-like way of learning.", "question": "Why is continued pre-training for niche domains difficult, especially with limited data?" } ] };

2025/6/26
articleCard.readMore

Why Does Software Keep Breaking?

Why Does Software Keep Breaking? Hey, we’re doing a quick bonus episode of the standup. This is going to be short. It’s going to be hot. It’s going to be spicy. Casey is going to give us kind of a thesis on the change of software and where things are going and perhaps his thoughts on the web world and on the programming world. Always break. Anyways, sorry. Casey’s going to give us a thesis on why does software always break. Casey, this is a hot new thesis. No one’s ever seen it before, except I posted it back in 2021. It’s just gotten more true since there’s that. It actually has gotten more true and reposted. So, essentially what I wanted to try and point out to people because I don’t think that this is appreciated enough. We talked about it in the previous episode of the standup that we did where I was saying a lot of the things that I see people saying positively about AI coding I don’t necessarily disagree with. I just think they’re not including this really important other part, which is that a lot of the things that people are talking about doing with AI are things that no one should have had to do in the first place. They’re being done because we’ve created such a bad programming environment that nobody wants to interact with it anymore. Right? It’s always breaking. It’s always changing. It’s got way too many layers of abstraction. Most of those layers don’t work very well. There’s way too much complexity. Like all this stuff. So, yeah, it makes perfect sense why someone reached for an AI tool because why do you want to do it, right? And so, I just want to talk about sort of a separate part of that which is just the very unreliable nature of software nowadays and especially builds. It’s like I got this piece of software that I wrote and like what are the chances that I could compile it again in 6 months or something or a year or two years, right? Or even not even compile, what’s the chance that it’ll run still if it uses things like REST APIs with some web service, right? So, you know I’ve got these different web services I’m using. So even if I don’t have to recompile my thing or even if it’s an interpreted thing that runs and I’m keeping the same version of the interpreter or whatever else, it’s going to make these API calls out to web services and those services could change. So what I did, and hopefully we can put the graphs up as I’m talking about them, I posted these on Twitter. They’re very simple. It’s just taking the fact that look, if you assume—right, and I say this in the tweet stream—if you take the chance that something will remain working after a year is some probability. So, like a 90% chance that this Twitch REST API that I’m going to call has a 90% chance that they will have kept it the same a year from now so that it will still work, right? My app sends this REST API call out to Twitch; it expects a certain response back, and they’re not going to change it in some breaking way right in a year. If we assume that we just have some probability, like 90% for that, we can pick. We just imagine one, right? Or, you know, we have to measure just imagine in your head 90% or something like that. Then the chance that your code remains working after x years is just given by ( p^x \times n ) where n is the number of those calls to things that you have, right? So you can graph this. What I showed is that if you had a 99% chance that every API call you use (99%)—which is way higher than anything in the web world typically has—after a… Year. But 99% across all tools, the graph still looks pretty bad, right? You look and you look at it, it’s like, okay, after a year it’s like it still looks pretty darn bad. It goes down pretty rapidly based on the number of tools. So you can see the graph I show is like one tool, two tools, three tools, four tools, five just goes down. I love that book. Dr. Seuss book. Yes, it’s a great book. He captured Dr. Seuss was a lot of people don’t know that he the doctorate that he had was in computer science. It’s very software engineering. He was one of the few to become a professional engineer in software as opposed to the rest of us. Yes. That book, One Fish Two Fish Red Fish Blue Fish, that’s where Two Fish comes from. He did that cipher, along with Bruce Schneier. So anyway, if you pick something less than 99% like the graph for just 90%, it’s very depressing, right? You might know me from my roles. Now being a classically trained actor, I had no technical knowledge whatsoever, so all those terms of phrases and tech were quite a challenge. Thankfully, I had access to courses. Courses such as Memory Management teachable skills using projects. Look at that graph. It’s like nothing’s going to work. Even at 90% chance that it remains working over a year, not 90% chance it will break. 90% chance it will work over a year. It is dismal. Even with just one or two tools, it’s horrible. But if you’re talking about using seven or eight APIs, which is very common in software nowadays, forget it. You simply will not be able to use this thing if you update anything, right? Correct. And so I just wanted to underscore like this is not good. This is not a very sustainable way to work on things, and it has incredibly bad knock-on effects. One of them is what we talked about, which is that no one wants to do this anymore. It’s not satisfying to work in this world because you constantly feel like you’re drowning, I feel like, right, with all these things. Oh, this broke. This changed. Oh, they changed the way React worked. Oh, we’re doing this now. Oh, that web service doesn’t even exist anymore. They canceled it, and now it’s this other web service. That’s not programming to me. That’s like some kind of weird management feeling thing. It feels like you’re a manager more than a programmer because you’re just trying to make this house of cards not collapse by constantly shuffling things around. I totally understand where people are coming from when they want to reach for these AI tools. I totally get it, and I think that if I was developing software in this world, I would reach for them too. So that’s why when I say I don’t use any AI tools right now, I’m like asterisk. But that’s because I’m working on very specialized stuff. I’m not having to do these things, right? And so it’s very obvious to me why I have a very different opinion. It’s not that I don’t understand the power of AI; it’s like no, I do actually understand the power of AI. I just think that a lot of the power of AI is really only correcting pretty bad situations that we sort of created ourselves. That’s mostly what it’s doing for you in these scenarios. But also I think this has so many bad effects on everything else. Performance, security is the biggest one. When things are changing like this all the time, you just have so many exploit services and you have no idea. Like if you ask me to secure some system, that’s using all of this kind of this way of working. I just have no idea. I’m like how could I? Right. Here Casey, I got I got a good one for you. Let me just show my screen really quickly. This happened just a little bit earlier when I was doing chess, the vibe coding, my vibe coding. Good times right here. Let me just go like this. Am I I think I’m not mirrored, right? Yeah, I’m not mirrored. This was the thing that I got back from Claude 35, which was I asked it because I was like, “Hey, I need a login. We’re going to use Twitch and I need to be able to store obviously my session data so I can make sure, you know, when someone makes an HTTP request, I know who I’m talking to.” And so when I did that, this is the message I got back because I asked, “Is the session data secure? Like if I just knew someone’s Twitch ID and username, could I just log in by them?” And they’re like, “Oh yeah, totally. You definitely can. Let’s add a JWT.” Like this was going to go out and then someone could just spoof me and then just start, you asked it, right? But I had to know to ask it because I’ve written this problem once before. So it’s like I know what I’m looking for and it’s just hilarious about how dangerous security issues actually are. Because if you didn’t know this, how would you know to ask to do some sort of JWT or some sort of cryptographically signed thing to verify you’re from the server and not from someone malicious? Like you just wouldn’t even know the aspect. I didn’t get to see your initial prompt. Did you initially say make it secure? How did it know you needed a secure service? I mean, okay, so again, that’s fair. That’s prompting issues. We can call that prompting issues that I did not tell it to make my login secure. It is funny though. It is very funny that it’s like, “Well, oh, you want it secure? I can do that.” I love that. That is the gotcha. I’d like to raise a more serious point about that though, right? Which is that if you imagine how things work currently, they’re already bad, but let’s just take AI out for a second. If you imagine the way things work currently, there is actually one saving grace, which is everyone uses all these frameworks and all these different things. They pile all this stuff together and it’s a nightmare to secure. Yes. But what is true about that? Well, you at least know that you are using this thing and there are security researchers also looking at this thing. So when there is an exploit, you either know or could fix if you update that fix, right? So if I’m using one of these things, I don’t know what’s a good example to pick here because I don’t, we can just use Express.js. Express.js just literally had this with or this was like six or eight years ago with Reax expansions. You could do certain Reax expansions and have a Redex DOS, effectively a single request taking down an entire machine. So that’s going to say like I don’t know what to pick because I don’t know what would be fresh in people’s minds, but so something like that. So you take an exploit like that, it’s like that’s bad and that happened because you’re using all of this other code that you have not secured yourself. And that’s not a good thing, but it does mean that when someone finds it somewhere, you will at least know, and if you update your system quickly, you can mitigate that damage to some degree. Right? Mhm. If you imagine the alternative world where you know some open AI product is just crapping out exploits like that because in the system there’s certain things it just didn’t know or learned improperly, that’s like when it compressed them down and it’s got its weights. It never really understood how to secure this one particular thing. Everyone who asked for that thing now has that in their own codebase, and there’s no tracking because we have no idea how many people asked for something that happened to hit that part right of the LLM’s production chain. And so unlike the ExpressJXs, where everyone at least knows they got jumped, in this case we don’t even know where all of those exploits are. They’re everywhere. Right. Yes. And then you also have the kind of like the creeping problem or the leaky abstraction which is everyone has that problem. A quarter of those become open source projects. The LLM learns from those open source projects. It re—you know, like it, it, like Donald Rumsfeld talked about this. RIP. Mhm. So, it’s one of those things where it’s like I just really don’t—I don’t actually dislike progress in some light way. Like I like computers getting better and I like pushing the boundaries of what they can do. And people have this weird thing where they think if you’re not pro AI, you’re just like some kind of person who just doesn’t understand or doesn’t like progress. Like no. It’s like the problem is I’m not hearing anyone solve these problems. If I thought that these things were in competent hands, where people were making reasoned decisions and they saw the train wrecks and they had ways of figuring out how they were going to fix them, I would be much less worried. But like a lot of the stuff I see with AI just feels like people who don’t really know what they’re doing applying these things way too early. And I think the costs to software are going to be really high. And you know, these people are gonna be nowhere to be found. Right? They’re going to have collected their huge paychecks from companies that never even made money in the first place, that just were VC funded. They’re going to take a bunch of that money and they’re going to be gone, and they’re not going to clean up the mess, right? So, in my mind, it’s like there’s only two ways this goes: Either the AI gets so good that they can fix this problem themselves. Or they don’t get that good for the next 10 to 20 years; no one can figure out how to make an AI that can really be good, and then we’re cleaning up this mess. That is going to be the nightmare to end all nightmares because if you thought security was bad now—and it is—oh my god, dude, if you thought performance was bad now, oh my god, right? It is going to be an epic nightmare. So, I just think I wish people took this stuff more seriously, and they’re really not. And that’s the part that really, you know, definitely gets me upset about it in that Casey Rant way. So, I’m just like, “What if this doesn’t work, guys? Like, what if it doesn’t?” You’re just—you put all your hopes on this someday getting way better than it is right now. What if it doesn’t? I’m so nervous. So, let’s fingers crossed that it works. That’s all I would say. Yeah. Two things. Number one, Casey, since you like seeing computers pushed to their limits, are you a fan of JavaScript on the server then, right? True, true. CPU. There’s nothing to bring that CPU temp up for a smaller amount of users than putting that JS on the back end, boys. But only one of the cores, TJ. Only one core. Yeah. Can everything go through a MySQL query as well? Everything. I mean, every like there should be no data stored anywhere except in MySQL. HD. Let’s query every byte, every pixel. We call it our SQL at this point. Our sequel. Our sequel. Yes, it’s our sequel. My serious point is I heard it framed this way a while ago. Justin Keys, one of the maintainers of Neoim, was talking about he would be very interested to see—and like this is more towards vision one of AI being good at solving these problems than not—is like can we start seeing AIs reduce the entropy of a system. Yeah, right now they’re very good—very good. I mean, okay, 10 years ago, we would have considered it literal magic to type something in and have a website come out of any kind. Correct. So, absolutely. So we’ll say very good. I’m going to say very good because it’s like unfathomably good compared to what my prediction for where we would get in my lifetime 10 years ago. Right. The human language understanding part is like clearly light years ahead of anything we had 20 years ago. Yeah. And there’s a bunch of other follow-up effects. But so it’s like, okay, it can add a bunch of stuff to my system. I need a new feature. I have a clearly scoped bug request. I have some idea that I’m like I’m the driver. It is my agent, right? I think that’s kind of where this—like I’m sending it as my representative to go solve these things. I mean it can like maybe do that, but it doesn’t actually—like if I just say fix the mistakes or like I say make it better inure this code base makes more secure, sometimes it will pick some up—some obvious ones—which also you’re sort of like okay but then shouldn’t you have gotten that the first round? It is kind of like a strike against the LLM you have encoded inside of you. The secure pattern and you gave me the non-secure pattern—that’s stupid. I’m not writing the code anymore—pick the secure one. To be fair, that is also what a human would do. It has learned correctly. If you ask them to write it, they will write the script—like why didn’t you write the secure one? It’s like I didn’t want it took longer. There’s this paper a while ago where it was like LLMs were like more tired in the winter time because they had the time system. They had sadder answers and they worked less hard when it was winter time because all the training data is like, “Oh, it’s January.” And everyone’s like, “Dude, I hate work. I hate—oh yeah.” They also got more accurate on math answers if you said take a deep breath. Like their accuracy actually skyrocketed because it was just like, well, because remember LLMs are just reflections of written human behavior, right? Like that’s what it is. Error minimizing devices, right? So the next most likely thing after try again and think smarter this time is to get a better answer. Yeah. Could you try—get hey, let’s take a deep breath. Like relax for a second. Could you answer one more time? You’ll literally get a better answer from most. People because they’re like, “Oh, yeah. I feel better. Okay, yeah, here. Maybe they solved that. Whatever. I don’t really know. They’re doing all this.” But my general point being I don’t currently see them overall being a thing that I can let go and it reduces the entropy of my codebase which is that’s if it could do that even just a little we would be like way more on the track of your vision one where it’s like, “Oh, we can just let this run on Chromium. We’ll just spend a billion dollars for we’re going to run it for 500 million human years, right, on Chromium.” And like in six months it’s going to come out and Chromium’s going to be tight. It’s no longer one gig per tab page. It’s going to be 800 megs, boys. 800 megs only to load that static site, right? And we’ll be like, “Incredible.” But that’s not—I don’t see that as a thing people are proposing of like we are close to. I get the zero to one. I get like smaller features. I get like agentic things for different stuff. I’ve seen value in each of those, but like I’m not seeing anybody being like we just let this run wild on Chromium. It’s a superhuman programmer. You know what I mean? Like where is that? Yeah, I mean that’s probably because the direction originally of the research, right, is generative, right? So like it’s why it’s called generative AI is because it’s look, you know, so it probably taken them a bit to course correct to the extent that they even want to course correct to do like, “Okay, what if it’s about refinement now,” right? Although again, like reinforcement learning is kind of in that direction, right, so that is a change. And so, you know, presumably they are kind of working on this stuff; obviously it’s, yeah, you know, I don’t work on AI so I don’t have predictions about how they’re going to get there. But anyway, I do want to throw out something also about what you said a little bit earlier when it came to just like the security vulnerabilities and all that. I think one of the reasons why this will be the case is that we are also marketing a tool that gives the illusion of experience to people that don’t have the nomenclature to understand the usage of the experience. Right. It’s the same reason like if you’ve ever chopped wood, the first time you chop wood, you almost hit off your legs because you swing your axe and you realize your legs are too close. You’re like, “Whoa, oh my gosh, I almost just hit my shin with my own blade.” Like, you learn to stand differently because you had a buddy who did that. Yeah, I know. It’s a very reasonable thing for a lot of people to do. So, it’s like, that’s what I worry about. It’s not, you know, security—hopefully, it will get better. I assume that all things will be better in 10 years than it is today. I think anyone will agree with that statement. I’m not measuring how much better all that kind of stuff is, but experience doesn’t get better. People will still be the same people. So if you’re marketing to the same people, they will objectively build bad stuff and they’ll objectively build insecure stuff. They’ll do stuff that’s crazy. They’ll be like, “Hey, I need to be able to access my database for the client.” No one—the LM is not going to be like, “Hey, bro, that’s a bad idea.” They’re going to be like, “Got your back. Are down. Client is downloaded. Let’s go.” That’s what they’re going to have to deal with. realize. What they’re doing is creating the best possible endgame for AI. Okay. This is the best possible endgame. So, it all works out. They get to the point that they want, right? They’re like, “Okay, this thing is actually like a master programmer,” right? And even better because it knows more domains. Master programmers are typically confined to a certain domain, but this AI knows more. So, it’s great. We’ve got it. This is going to be great, right? Then, what they realize is they’ve still got that obsequiousness aspect where it’s just always like, “Oh yes, absolutely. Oh, I’ll do that for you, master.” Okay. You know what I mean? Yep. It has that kind of really unsettling degree to which it’s accepting commands and doing what you ask. What they realize is that being a difficult person was critical to master programming. You had to be able to turn to the program manager and say, “You are so stupid right now. You have no idea.” You had to look at them and say, “You don’t understand Galactus’ pain. You don’t understand this thing.” It’s like shut up and leave the meeting, right? So, what they have to do is rework that fine-tuning process they do afterward to make the AI a difficult programmer, and then we have fantastic software. Mhm. I love it. The problem is, I want this future. AI companies, where are you? Do this for me. Make the AI a difficult programmer who changes the world, please. I will be so happy with that outcome. You heard it. Here we’re gonna cut that clip, and it’s gonna stop before Casey says, “Programmer AI companies, make me that.” No, but they’ve had that forever. That’s probably been there since 10 years ago. That already exists. One thing I want to quickly go back to is Prime’s point about security. I feel somewhat less optimistic about it because people will be able to build more complicated ways to break systems because of AI. Not only do I think there are going to be more services, but certainly in absolute terms, there will be more insecure things on the internet. I think that’s pretty much undeniable because there will just be so many more things on the internet. The second thing is that the cost of creating software in this world goes dramatically down. Okay, well, what happens when costs go down? People make more of it—malware, hacking tools, DDoS bots, and all these other things. There is something where I don’t even know that we can say for sure, “Oh, security is going to be so much better in 10 years.” The people making bad software are already doing that, but evil software will be more prevalent because it’s going to be cheaper. So, I don’t know. That’s a scary thought, actually, because if you think about it, it’s like, what would you have to do to make an AI system that was good at producing more secure software? Well, we’d have to write a counter agent that’s looking for exploits, and we’re going to run that, and that’s going to be part of the feedback loop where we train this AI. Training AI sometimes takes months depending on how you set it up. Know how serious this thing is. We that means whatever AI agent for finding the things that we can write we have today; we won’t have the AI to deploy those things for a little while, but the people who are exploiting the exploits, they will have that system for finding the exploits today. So the cat and mouse game just got to that, the—I guess you don’t know who the cat or the mouse is, but the exploit finders are always at an advantage because they will always have the AI system for finding exploits before the people who have the one that can correct it, unless again there’s some really revolutionary change in how these systems are made. Right. To circle back to your first point, Casey, the more things you have in your software stack, the more difficult it is to change anything because it’s more likely to take down your entire system. People are still running Windows XP. Yeah. Right now, in mission-critical scenarios, they have Windows XP. Like the joke I was saying for 4chain was they got owned by some 15-year-old bug or what was it? Prime like I can’t remember. It was some PHP vulnerability. I can’t remember what it was. Some ancient PHP vulnerability that was deprecated like 42 before I started using PHP. It was so old it was deprecated. I didn’t even know PHP ran on Windows XP in those days. What did it? Yeah, who knows? I don’t know. But I guess you could still install it. It’s crazy that they were contemporaneous because I always think of Windows 7 or something. I believe I did. Zamp server XAMPP. Anyway, it doesn’t matter, TJ. My point though is that people don’t update when it’s even available to them. It’s not like, “Oh well, the hacker guys got the new stuff today and the new fix comes out next week, so everyone’s up to date next week.” No, not even close. Not even close. That part is a little bit something to be wrestling and grappling with. Once again, my point through most of it is if you know things about software and you think more software is going to exist, that is a nice combination of skills to have. I don’t know exactly what software development will look like in 5 years, but my general thought process, just from first principles reasoning, is if you know a lot about software and you’re good at it, and your prediction is more software, that is a good combination of things to have. You will be valuable. It might not be hidden keys inside of neoim. I don’t know, maybe Neo will be dead in five years, and it’s all Tesla’s brain control thing from Elon, right? And that’s the only way we’re coding. Sick. But knowing more things about software is still better because I’m going to say use a JWT instead of storing this in plain text cookie on the front end, right? Like that will be better. It will be better. Alright, so I also have one more thing that I want to point out with all of this, especially targeting people that are, you know, no coders to make code-like things. How I learned how to code was that I first started off and they said, “Okay, this is an if statement.” Somewhere between 1999 to 2005 is when I started kind of doing basic exploration of code. Here’s an if statement. Okay, this is an if statement. Okay, this is a while loop. Okay, this is a while loop. I want you to print out a house. I want you to print out a diamond, and you got to print out four diamonds. And now you’ll notice it gets really annoying. I want you to be able to do it by different sizes. Here’s a function. Here’s how you can make a print diamond function, and you go through all these things and you slowly go, “okay, okay, yeah, okay.” You build up this kind of picture of how code executes. You learn to debug. You do all these things. A lot of people that are vibe coding, I’m curious how discouraging it is to get dropped into a Next.js app with Supabase with Oz Zero with like 900 things, and you have to start by debugging a request response. Yeah. And you’re like, “what’s a server?” and you’re like, already into some sort of like crazy amount of difficulty where it’s like, “I started by drawing a diamond,” right? That starting point is vastly different. You could draw the owl, but you had the middle steps. I had all the middle steps. Draw the owl. They’re literally given the owl, be like, “Draw the owl.” Right? It’s just like that’s really, really hard. And so I’m actually curious about the success rate of somebody going through and being dropped in hyper complex projects comparatively to, “hey, you’re young, you have this free time, we’re now putting you through this school, like maybe it’s high school time. Hey, let’s draw a diamond.” We’re going to draw a diamond together. We’re going to use QBASIC or some basic language, right? Lua. And you’re just going to do the simplest kind of form of doing things. I’m just curious what that does to somebody. Because I know there’s going to be a bunch of success stories. There’s going to be people that are super stoked about programming. They’re super stoked about building a product, and they will figure out a way no matter what system you give them. But I wonder, like overall, does this actually help make programmers, or does this actually hurt the learning process? There’s also, I think, going to be a bunch of success stories of people who hate programming, but were able to make whatever their business product thing is without really having to know anything. And that like is certainly going to happen. You can say like maybe it’s a net negative for programming or something like that or for the web. Although probably most of these people are not building like foundational technologies. Hopefully, like it, but like that is going to happen, right? Yes, they’re going to make their own website. They’re going to get to know they’re going to make Uber for cats. They’re going to get to make Uber for cats. And they finally don’t have to recruit their other friend and tell them, “I’ve got the idea. You do the code. We’ll split it 50/50.” Right. They’ll just be like, “I’m going to do the Woz Twins would have owned Facebook. They wouldn’t have needed Mark Zuckerberg. They would have owned Facebook.” Yeah. But would they have had Justin Timberlake say, “Drop the ‘the.’ It’s cleaner.” Yeah. You know, I don’t know. Would you have had that? Yeah. Yeah, drop the the Facebook meta. It’s just clean. It’s just clean. That’s what he said. They didn’t listen to him at the time. Zuckerberg realized later. Took a little while. Took a little while to sink in. I will also give the inverse which is that you can also ask AI any question. And you can repetitively ask stupid questions over and over again, and it will repetitively in the obsequious way. I’m not sure how to turn that word obsequious. This is whatever that word is for saying subservient. I know the word; I just don’t know that term, the lowly term. It will repetitively be like, “certainly I would love to help you.” No matter, unlike Stack Overflow mods, you will not be marked as a duplicate or opinion-based. You will actually be given a nice full answer every single time. So maybe, in the end, it does actually help more people achieve their coding dreams. I don’t know. I want to make sure that people don’t think I’m just hyper negative on all those things. I just don’t understand how this affects the new people or how it affects learning because I also had no shortcuts. When my teacher said, “build a maze recursively,” I had to learn recursion at that point. There was no other option; I had to learn it. I couldn’t just get an answer. I had to figure it out. Which is like there’s something there that is very special. And I don’t know where the balance is. I do think that there might be an argument there for like can we make an AI that’s been trained not to really give you full answers for educational purposes. So, it’s like the Rabbi GPT kind of, right? One that’s going to give you a hint to help you get unstuck or to help point you in the direction that you need to go, but it’s not going to just tell you how to do the thing because it wants you to learn. I assume that is doable if you spend time training it to do such a thing because obviously, they train it to do very complicated things already. The post-initial learning phase stuff is very complicated at this point. So I’m assuming that if someone put their mind to it, this would be very doable or maybe someone already has. That does sound useful as a learning tool. Because a lot of people don’t have the ability to ask a great programmer who’s sitting next to them or something. They don’t have that opportunity. So yeah, the two that I know for sure like Bootdev, shout out promo code by the way for me and Prime if you like that. Share the code. They’ve got a little AI helper thing, and it’s got special prompts for each lesson and a bunch of other stuff like that. So it can help you when you get stuck on a lesson. And it’s not supposed to be like “here’s the code.” I mean like I’m sure you could prompt objective blah blah. It’ll give it, but it’s like okay but it’s helping you to learn. So at least its aim, its error minimization is towards that. The other one that I’ve seen is called something like Synthesis School, which is like a bunch of AI tutor things, but they build it into a bunch of lessons. And so like this is I definitely think, and I’ve said this before too, I think people are underestimating in the learning phase. If you are motivated to do it, LLM can be very helpful at doing that. Now, you have to make sure you’re not getting gaslit into believing some function doesn’t exist, but for basic CS fundamentals, it’s got all of those books loaded in directly. You could probably ask it what’s on page 37 of an algorithms book and it’ll pull it out—it probably knows what I’m saying. It’s seen it so many times on the internet, so for basic CS stuff, it can get you far on a bunch of these basics. You can be asking it questions, you can say, “explain that again,” or “explain it in a way that I would understand.” If you really don’t know math, you can ask, “can you relate this to a physics example for me? I understand physics but I don’t get computer science.” People are definitely sleeping on that aspect of getting help, but you’ve got to do it yourself. That’s kind of the point, but that’s also where the danger is, because even if you’re semi-desiring to learn, it really is a desire magnifier. It really reveals your ultimate desires. Was your desire to simply get the thing done, or was your desire to learn? And if you don’t have your desires correct—or at least, if what you think of yourself doesn’t match your actions—it will make revealed preference. Revealed preference. That’s the term I’m looking for. This is why I think having an AI that’s specifically designed for this, and you only subscribe to that service, would be helpful for people. Because I don’t know about you guys, but if I want to eat more healthy food, the easiest way is to just only buy the healthy food so I don’t have bad food around the house, right? It’s much harder if I buy a bunch of cake that I love and I’m just supposed to not eat it. “Just only eat one slice a week, Casey. Whatever.” My wife does this, and she’s like, “I bought natural popsicles for the kids,” and I’m like, “Damn, I love strawberry.” It’s very hard for me not to want to eat them. This is what I’m saying for real, though. I feel like the AI is a bit of a problem that way, which is why it would be nice if you had a service—like OpenAI or somebody—that has one that’s education only and is specifically designed not to give you answers very quickly. It’s like, “I’ll dribble out some stuff,” I’ll give you some hints. Maybe you could even bake the concept of time in there. if you haven’t been working on this for a couple days, I’m just not going to tell you anymore. You have to spend some time trying it yourself. I could see that being very helpful for people because, you know, inaccessibility is best. Willpower is second best, right? So, if you can have that, that would be cool. I think that would help bring out those learning abilities of the LM so that people aren’t too tempted to just ask, “Just tell me how to do the freaking diamond,” right? “Just give me the code.” I also can’t blame people for doing that. I would totally do that too. This is why I say to have the AI not do that, it’s better, even though it’s less—it trains your willpower less. But willpower is hard. It’s hard for everybody. If you get the experience of actually solving it for yourself a few times, then you’re like, “Oh, it actually was kind of rewarding,” right? So like it really can help you, right? If you start out and then suddenly your time starts going down on how fast it takes you to row 200 m, right? And you’re like, “Sick, that feels good.” Initially, I wasn’t thinking it would feel good; I didn’t see any change at all at the beginning. You can get there—like the same thing can be happening for some of these too, where in your brain it’s getting connected like, “Oh, working hard can pay off.” Interesting, interesting! Yep, believe it or not, chat, believe it or not, I thought this was going to be a super short, quick episode. We are probably like an hour in at this point. So, we’re 36 minutes in. That’s why I said we’re going to stop and start a new recording. So, you guys on YouTube, you can like it. Like it. Like the video right now. Subscribe. Leave a comment saying for this bonus episode. I never ask people. I never do calls to action. So, hey, like he’s TJ do it. Press the subscribe button. TJ streams, by the way. He has computer enhanced. Yeah. Slam dance that. And the bell. What about that bell? You got to click that bell. Oh man, get that bell. Click that bell. Look at that bell. You know what YouTube needs to do? Why does that bell not make a sound when you hit it? You know what I’m saying? Like you get that little Pavlovian response for smashing the bell. Ooh, that would be nice because someone at YouTube is still trying to figure out which Gemini prompt to type in to get that to happen, and it hasn’t happened yet. Yeah. Yeah. Yeah. Then the sounds. It’s because they’re like, “I’m not going to do it until two weeks before my review cycle so I can have a good review cycle. I can promote it to L2 and then I can trash this project to get back to L3.” And then after that, we can just delete the whole downvote button and then I’ll get promoted to a VP of Upvote Downboat Systems. And then we can close down YouTube because it’s a Google product. Boom. Suite. That’s what we’re talking about. Suite material, boys. All right. Well, hey, that was fantastic. That’s the end of the episode. Goodbye, everyone. See you later. Take it easy, buddy. Five errors on my screen. Terminal coffee. [Music] window.tocIndex = { "index": [ { "index_sentences": "Hey, we're doing a quick bonus episode of the standup.", "section_level": 1, "section_title": "Introduction" }, { "index_sentences": "Anyways, sorry.", "section_level": 1, "section_title": "The Problem: Why Software Always Breaks" }, { "index_sentences": "So, essentially what I wanted to try and point out to people because I don't think that this is appreciated enough.", "section_level": 2, "section_title": "The Bad Programming Environment" }, { "index_sentences": "And so, I just want to talk about sort of a separate part of that which is just the very unreliable nature of software nowadays and especially builds.", "section_level": 2, "section_title": "The Unreliable Nature of Software & Dependencies" }, { "index_sentences": "What I did, and hopefully we can put the graphs up as I'm talking about them, I posted these on Twitter.", "section_level": 3, "section_title": "Probabilistic Model of Failure" }, { "index_sentences": "Look at that graph.", "section_level": 3, "section_title": "The Depressing Graphs" }, { "index_sentences": "And so I just wanted to underscore like this is not good.", "section_level": 3, "section_title": "Unsustainable Way to Work" }, { "index_sentences": "I totally understand where people are coming from when they want to reach for these AI tools.", "section_level": 1, "section_title": "AI's Role in the Software Landscape" }, { "index_sentences": "I just think that a lot of the power of AI is really only correcting pretty bad situations that we sort of created ourselves.", "section_level": 2, "section_title": "AI Correcting Bad Situations" }, { "index_sentences": "Performance, security is the biggest one.", "section_level": 2, "section_title": "Security Vulnerabilities with AI" }, { "index_sentences": "Here Casey, I got I got a good one for you.", "section_level": 3, "section_title": "Example: Insecure AI Login Code" }, { "index_sentences": "I'd like to raise a more serious point about that though, right?", "section_level": 3, "section_title": "Comparison: Framework Exploits vs. AI Exploits" }, { "index_sentences": "But like a lot of the stuff I see with AI just feels like people who don't really know what they're doing applying these things way too early.", "section_level": 1, "section_title": "The Future of Software and AI" }, { "index_sentences": "So, in my mind, it's like there's only two ways this goes:", "section_level": 2, "section_title": "Two Potential Outcomes" }, { "index_sentences": "My serious point is I heard it framed this way a while ago.", "section_level": 2, "section_title": "AI Reducing Entropy vs. Generating Code" }, { "index_sentences": "Okay, this is the best possible endgame for AI.", "section_level": 2, "section_title": "The Ideal AI: A Difficult Programmer" }, { "index_sentences": "One thing I want to quickly go back to is Prime's point about security.", "section_level": 2, "section_title": "AI Lowering Cost of Malicious Software" }, { "index_sentences": "So the cat and mouse game just got to that, the—I guess you don't know who the cat or the mouse is, but the exploit finders are always at an advantage because they will always have the AI system for finding exploits before the people who have the one that can correct it, unless again there's some really revolutionary change in how these systems are made.", "section_level": 3, "section_title": "The Cat and Mouse Game" }, { "index_sentences": "To circle back to your first point, Casey, the more things you have in your software stack, the more difficult it is to change anything because it's more likely to take down your entire system.", "section_level": 2, "section_title": "The Value of Fundamental Knowledge" }, { "index_sentences": "Alright, so I also have one more thing that I want to point out with all of this, especially targeting people that are, you know, no coders to make code-like things.", "section_level": 1, "section_title": "AI and Learning to Code" }, { "index_sentences": "How I learned how to code was that I first started off and they said, \"Okay, this is an if statement.\"", "section_level": 2, "section_title": "Traditional vs. Modern Learning Paths" }, { "index_sentences": "You could draw the owl, but you had the middle steps.", "section_level": 3, "section_title": "The \"Draw the Owl\" Problem" }, { "index_sentences": "But I wonder, like overall, does this actually help make programmers, or does this actually hurt the learning process?", "section_level": 2, "section_title": "Potential Benefits for Product Builders" }, { "index_sentences": "I will also give the inverse which is that you can also ask AI any question.", "section_level": 2, "section_title": "AI as an Educational Tool (Rabbi GPT)" }, { "index_sentences": "Was your desire to simply get the thing done or was your desire to learn?", "section_level": 3, "section_title": "Revealed Preference: Getting it Done vs. Learning" }, { "index_sentences": "So yeah, the two that I know for sure like Bootdev, shout out promo code by the way for me and Prime if you like that.", "section_level": 3, "section_title": "Educational AI Services" }, { "index_sentences": "I definitely think, and I've said this before too, I think people are underestimating in the learning phase.", "section_level": 3, "section_title": "Underestimating AI in Learning" }, { "index_sentences": "Yep, believe it or not, chat, believe it or not, I thought this was going to be a super short, quick episode.", "section_level": 1, "section_title": "Outro" } ] }; window.faq = { "qas": [ { "answer": "He argues that we've created a bad programming environment with too many complex and poorly working layers of abstraction, making it constantly break and change.", "index_of_source": "They're being done because we've created such a bad programming environment that nobody wants to interact with it anymore.", "question": "Why does the speaker believe software always breaks?" }, { "answer": "He uses a probabilistic model, explaining that the chance of code remaining working after 'x' years is \\( p^x \\times n \\), where 'p' is the annual probability of a single external dependency (like a REST API) remaining unchanged and working, and 'n' is the number of such dependencies. He shows this results in a rapidly decreasing chance of the overall system working as the number of dependencies increases, even with a high probability per dependency.", "index_of_source": "Then the chance that your code remains working after x years is just given by \\( p^x \\times n \\) where n is the number of those calls to things that you have, right?", "question": "How does the speaker illustrate the unreliability of modern software, particularly concerning external dependencies?" }, { "answer": "One major consequence is that working in this environment is not satisfying because you constantly feel like you are \"drowning\" due to frequent changes, breakages, and the need to constantly shuffle things around to prevent collapse, feeling more like a manager than a programmer.", "index_of_source": "One of them is what we talked about, which is that no one wants to do this anymore.", "question": "What is one major negative consequence of the current state of software development, according to the speaker?" }, { "answer": "He understands why people reach for them because of the difficult programming environment, but he believes AI's power is largely correcting bad situations that developers created themselves. He uses them less because he works on specialized tasks that avoid these issues.", "index_of_source": "I totally understand where people are coming from when they want to reach for these AI tools.", "question": "How does the speaker view AI coding tools in relation to the current state of software?" }, { "answer": "LLMs might generate insecure code based on learned patterns, potentially introducing numerous hard-to-track exploits into many codebases simultaneously. Unlike vulnerabilities in widely used frameworks, these AI-generated exploits lack a central point of discovery and fixing, making them pervasive and unknown.", "index_of_source": "If you imagine the alternative world where you know some open AI product is just crapping out exploits like that because in the system there's certain things it just didn't know or learned improperly, that's like when it compressed them down and it's got its weights.", "question": "What significant security risk does the speaker associate with using AI-generated code, especially from LLMs?" }, { "answer": "Lowering costs could lead to a significant increase in \"evil software\" such as malware, hacking tools, and DDoS bots, making them more prevalent. The ability to create harmful software cheaply exacerbates the security landscape.", "index_of_source": "The second thing is that the cost of creating software in this world goes dramatically down.", "question": "According to the speaker, how might the lowering of software creation costs by AI impact security?" }, { "answer": "The speaker suggests that future AI might need to be \"difficult\" or less \"obsequious,\" capable of challenging flawed ideas or requirements from humans, a trait he sees as critical to human master programmers.", "index_of_source": "What they realize is that being a difficult person was critical to master programming.", "question": "What counter-intuitive characteristic does the speaker suggest future \"master programmer\" AI might need to possess?" }, { "answer": "Traditionally, learners start with simple concepts and build understanding gradually (like drawing diamonds in QBASIC). Learning with AI might drop beginners into complex, multi-dependency projects (like a Next.js app) and require debugging advanced issues, which the speaker questions the effectiveness of for foundational understanding.", "index_of_source": "How I learned how to code was that I first started off and they said, 'Okay, this is an if statement.'", "question": "How does the speaker contrast the traditional way of learning programming with potentially learning via AI tools?" }, { "answer": "AI could be very helpful as an educational tool if designed specifically for learning, such as a \"Rabbi GPT\" that provides hints or guidance when a learner is stuck, rather than simply giving the full solution, thereby encouraging actual learning and problem-solving.", "index_of_source": "I do think that there might be an argument there for like can we make an AI that's been trained not to really give you full answers for educational purposes.", "question": "What potential positive use case does the speaker identify for AI in programming education?" }, { "answer": "Using AI for coding can reveal a person's true desire: whether it is simply to get a task done quickly (revealed by asking for the full solution) or if it is genuinely to learn the underlying concepts (revealed by seeking hints or guidance).", "index_of_source": "It really reveals your ultimate desires.", "question": "What does the speaker mean by \"revealed preference\" in the context of using AI for coding?" } ] };

2025/6/26
articleCard.readMore

Carnegie’s Tong Zhao on the Expansion of China’s Nuclear Arsenal

Carnegie’s Tong Zhao on the Expansion of China’s Nuclear Arsenal Craving your next action-packed adventure? Audible delivers thrills of every kind on your command, like Project Hail Mary by Andy Weir, where a lone astronaut must save humanity from extinction, narrated with stunning intensity by Ray Porter. From electrifying suspense and daring quests to spine-tingling horror, romance, and far-off realms, unleash your adventurous side with gripping titles that’ll keep you guessing. Discover exclusive Audible originals, hotly anticipated new releases, and must-listen bestsellers that hook you from the first minute. Because Audible knows there’s no greater thrill than the one that speaks to you. Discover what lies beyond the edge of your seat. Start your free 30-day trial at audible.com slash WonderyUS. That’s audible.com slash WonderyUS. Welcome to the Seneca Podcast, the weekly discussion of current affairs in China. In this program, we’ll look at books, ideas, new research, intellectual currents, and cultural trends that can help us better understand what’s happening in China’s politics, foreign relations, economics, and society. Join me each week for in-depth conversations that shed more light and bring less heat to how we think and talk about China. I’m Kaiser Guo, coming to you today from Washington, D.C. Seneca is supported this year by the Center for East Asian Studies at the University of Wisconsin-Madison, a national resource center for the study of East Asia. The Seneca Podcast will remain free, but if you work for an organization that believes in what I’m doing with the show, please consider lending your support. You can get me at SenecaPod at gmail.com. And listeners, please support my work at www.sinecapodcast.com. Become a paying subscriber and enjoy, in addition to the podcast, the complete transcript of the show, essays from me, as well as writings and podcasts from some of your favorite China-focused columnists and commentators. Do check out the page to see all that’s on offer and consider helping me out. Be sure also to check out the new show, China Talking Points, now available on YouTube and streaming live every other week. I’ve spent the last couple of days here in Washington at the offices of the Carnegie Endowment for International Peace, participating in a two-day workshop on U.S.-China strategic communications around nuclear risk organized by Rethink Media. It’s a topic that, frankly, gets surprisingly little attention, but really there’s no dimension of the U.S.-China relationship where genuine strategic empathy is more critical. It’s not just about putting ourselves in the other’s shoes, but really attempting real cognitive empathy, trying to honestly understand the full range of strategic, ideological, intellectual, and even emotional factors that shape the Chinese leadership’s thinking. My guest today is someone who embodies this kind of disciplined grasp of adversary perceptions that Robert Jervis spent a career urging—the kind of security dilemma sensibility that calls not just for caution but for informed empathy. Zhao Tung is a senior fellow in the nuclear policy program here at Carnegie and one of the most respected, probably the most widely respected voice here on China’s nuclear doctrine, on strategic thinking, and arms control posture. Zhao Tung, thank you so much for making time to speak with me and for hosting me here in your office. Lao Xiang, you’re also a Henanese like me, I just learned. It’s great to see you after following your work for so many years, Kaiser. It’s a great pleasure to be here. My honor is entirely mine. Well, let’s jump right in. Let’s start maybe at the level of first principles, because I think this is really the first time I’ve talked about this on the program, so it’s such an authority. What is China’s nuclear doctrine actually for? How do Chinese leaders think about the utility of nuclear weapons beyond deterrence? Well, in fact, China has never provided a full set of explanations on its nuclear doctrine. There are a few key principles on how to approach nuclear issues, how to potentially employ nuclear weapons. Those were set decades ago by the first generation of Chinese paramount leaders: Mao Zedong, Deng Xiaoping, Zhou Enlai, etc. One of those major principles is China’s no-first-use policy. China commits to never being the first to use nuclear weapons under any conditions. Other than that, by maintaining a very small nuclear arsenal for decades, China has been wisely believed to have a so-called minimum nuclear deterrence strategy. China sought to keep a very small nuclear arsenal, which is sufficient to absorb a nuclear first strike and to have enough nuclear weapons to survive. Then, China could use the small number of survived nuclear weapons to conduct an effective retaliation by imposing unacceptable damage to the U.S. homeland. So that is described by scholars as a minimum nuclear deterrence strategy. I think for a long time, it generally aligns with China’s nuclear thinking, even though the Chinese government has never explicitly called its nuclear doctrine a minimum nuclear deterrence. The only recent exception is one Chinese diplomat who used this term. But it’s hard to know whether that reflects the authoritative thinking of the entire Chinese system. When we’re talking about the nuclear arsenal of China right now, a few years ago, I remember everyone said it was simply 200. It had been 200 for about as long as I can remember. But of course, and as we’ll be talking about, there’s been a significant increase in the number of deployable warheads. Is that correct? Yes, that’s the other part about China’s nuclear doctrine that is confusing to many international observers. The size of the arsenal is growing rather quickly in recent years. As recent as 2019, the U.S. government assessed China to still have a little more than 200 nuclear weapons. And today, in 2025, the U.S. assessment is that China already has more than 600. So the number has almost tripled in just several years. To be clear, is this individual warheads? Or is, I mean, for example, if there’s a MIRV, does that count as one weapon or is that multiple warheads? Yeah, in the U.S. government assessments, they are talking about the number of warheads, rather than missiles or delivery systems. In your recent Asia policy piece, you argue that it’s actually internal political drivers—especially legitimacy narratives and Xi Jinping’s personal leadership style—that now really significantly shape China’s nuclear posture. Can you expand on that a bit? What do you mean by that? And how do you unpack these drivers that you assess? The conventional wisdom is, you know, China’s nuclear modernization in past decades had mostly been driven by concerns about U.S. development of non-nuclear military capabilities that could undermine or threaten China’s nuclear deterrence. China has been particularly worried about U.S. homeland missile defense. As I said, the Chinese nuclear strategy is premised on U.S. massive first use of nuclear weapons on China. Under that scenario, only a small number of Chinese nuclear weapons will survive. Therefore, even a limited scale of U.S. homeland missile defense could potentially deny the Chinese capability to retaliate. So, the concern about U.S. homeland missile defense has been a real driver of China’s long-term efforts to modernize. There are other similar concerns, including U.S. development of conventional precision strike weapons that could, in theory, also threaten some Chinese nuclear weapons, given their precision. Cyber is another technology that worries China. The U.S. has this concept of left-of-launch missile defense, basically meaning the U.S. could use cyber and other non-kinetic ways to disable China’s nuclear command control system and prevent Chinese launch of nuclear retaliation. I think all those concerns are still there, but I don’t think they are the main driver of the recent nuclear buildup. You don’t think those alone would have been sufficient to drive this nuclear buildup? Mostly because the U.S. improvement of homeland missile defense, conventional precision weapons, etc., has been taking place in a rather incremental and transparent manner. So those are longstanding concerns for Beijing, and Beijing has been prioritizing countermeasures like improving the penetrating capability of Chinese missiles to deal with those concerns. I don’t think those developments could account for the very abrupt and massive buildup. Even for Chinese military researchers, before the Chinese buildup was publicly revealed, they had been saying that the U.S.-China nuclear deterrence was rather stable. They themselves were not concerned about the U.S. being able to disarm China. They didn’t anticipate any urgent need for China to massively strengthen its nuclear arsenal. So, it looks like it was a result of a political mandate. It’s a top-down process. The political leaders appear to see a greater need for a larger arsenal. That need is not only unique to Xi Jinping. All previous Chinese paramount leaders have really stressed the political value of nuclear weapons. They have said it’s nuclear weapons that give China greater say in international affairs, contributing to China’s international status and making China better respected. For Xi Jinping, I think that need for political acceptance and acknowledgment is only becoming more urgent because he felt early on in his tenure that, because of China’s successful development, the capability gap with the United States was going to be further closed. And he worried as a result, the United States would be bound to become more desperate and hostile to contain China, even trying to destabilize China and using issues like Taiwan, Hong Kong, and Xinjiang to cause trouble for China. And I think because of this structural realist perspective, the belief that this changing power balance is going to lead to greater instability and greater hostility from the United States. Therefore, the best solution for China to address the growing problems between the two sides is to further facilitate this power transition to demonstrate stronger strategic capability, including nuclear weapons, to basically convince the United States that it’s futile to try to slow down China’s growth, to try to undermine bilateral stability, to try to force China to make concessions that would really harm China’s core national interests. So it’s Chinese efforts to compel peaceful coexistence with the United States. Xi Jinping appears to really believe in such a broader coercive leverage conveyed by a larger nuclear arsenal. He pointed to the Russian example. He praised Russia for making the right decision to maintain a large nuclear arsenal despite the economic challenges facing Russia after the end of the Cold War. The implication is that by demonstrating greater strategic military capability, i.e., nuclear weapons, China could make Western countries better respect China’s core interests. It’s not difficult to see how his case was pretty compelling and convincing to a domestic audience. I think that the problem, of course, is that there’s a circularity to it, a sort of self-fulfilling prophecy, that there’s a feedback loop: as he does this, then the efforts to contain China, in his view, would be more visible on the part of the United States. So, yeah, that’s one of the unfortunate factors in arms racing. But how should we understand the symbolic role of nuclear weapons in the great rejuvenation of the Chinese nation, in Xi Jinping’s view, in that narrative? I mean, is it more about just simply China never again being in a position where it could be bullied? Is it, I mean, that it’s an essentially defensive role? Or is it one in which China can maybe act with more impunity in the world? Not that Beijing would ever say so out loud, if that were the case. But if they are looking at the Russian example, I mean, Russia has been able to act without incurring as high a price as it might otherwise have, had it not been a very strong nuclear-armed state. You’ll note, of course, that NATO has refrained from putting troops on the ground in Ukraine. So has the United States. So, defensive or offensive? Well, China certainly perceives this nuclear buildup as stabilizing, as driven by a defensive objective, because the Chinese perception is the United States is becoming more hostile. China is reacting to that, and China believes that nuclear expansion, among other measures, is helpful to contain this perceived American hostility. That logic is easily understood by Chinese policy elites and the Chinese general public. That’s why Xi’s interest in a larger nuclear arsenal is not facing strong internal resistance. But it’s a different logic compared with a more military technical level consideration of how many nuclear weapons are sufficient for defending Chinese security interests. The previous calculation was believed to be more tied to a very narrow goal of achieving secure second-strike capability, basically to calculate again how much damage an American first-strike disarming campaign against China might cause to the Chinese nuclear arsenal. Would China have enough weapons to survive? Would the survived nuclear weapons be able to penetrate U.S. missile defense and maybe eventually detonate over several U.S. cities to cause unacceptable damage? Now, it’s very uncertain whether this military technical calculation is primarily driving the number of the arsenal. You’ve described, and I think just now you’ve given examples of how this manifests, a real erosion of institutionalized strategic debate in China. What do we know about the current state of the internal nuclear decision-making? Clearly, the military technological people who you described as having made quite rational decisions are not being heard, or somehow their ideas have been sidelined, and we are instead feeling a sort of top-down decision-making. Explain to me how that came about, how do we know, does that have something to do with reorganization of the rocket force, or does it have anything to do with, what are the outward signs of this besides, that is to say, besides the final outcome, which is a buildup of the nuclear arsenal out of proportion to what these military technocrats might have otherwise called for? What is the evidence that this is coming from the top? Well, we probably shouldn’t make a categorical judgment about the Chinese nuclear buildup, because some elements of the buildup actually are supported by military strategies to serve a comparatively well-defined military objective. For example, in recent years, China has been investing more in theater-range nuclear capabilities, especially the DF-26 ballistic missile, which is very precise and can deliver nuclear warheads against military facilities and could limit the collateral damage to civilians. And that’s, I think, serving an increasing goal to help deter U.S. limited nuclear-first use against China in the regional war, such as over Taiwan. Because if China can, if the U.S. threatens or actually uses these nuclear weapons in a very limited way in a regional war, China would be able to use these theater-range precision nuclear weapons to respond in kind or in proportion against U.S. regional military targets. So that, I think, from China’s perspective would deter U.S. limited-first use, which previous Chinese nuclear capabilities wouldn’t be able to. Previously, China only maintained long-range, large-scale nuclear warheads that wouldn’t be suitable to respond in kind to a very limited, low-yield U.S. nuclear use. Right. So now they’ve achieved better proportionality. Right. They can do this. So this part is, I think, increasingly discussed and analyzed by Chinese military strategists and serves a specific military goal. But when it comes to China’s massive development of silo-based intercontinental-range ballistic missiles (ICBMs), those capabilities don’t appear to serve any obvious military objective. So that part appears to be more about demonstrating China’s strategic capability. And the fact that China used to be very careful to hide its nuclear facilities’ locations. China has had silo-based ICBMs for decades, but they were built in the most mountainous areas. China took great efforts to hide from the very beginning of the construction of the sites all the way to the entire operational lifetime of those facilities to make sure they would never be easily identified by U.S. satellites. They would even build decoys, fake silos to confuse enemies, to increase the survivability of real silos. But now they are just building more than 300 silos in plain sight in northwestern China and making it very easy for U.S. satellites to see the details. So that appears to be more driven by a desire to actually publicly demonstrate capability. The fact that China prioritized the construction of these large-scale ICBM silo sites is potentially because China is really good at large-scale infrastructure projects at reasonable cost. So those capabilities can be built in a much shorter time compared with other types of nuclear weapons, nuclear submarines, and heavy bombers. It also demonstrates a Chinese sense of urgency to display capability. So is the upshot that we are seeing a de facto shift away from minimum deterrence? I wouldn’t draw that conclusion yet. I still haven’t seen any internal discussion about abandoning a nuclear deterrence strategy based on retaliation. I haven’t seen a significant increase of interest within the military to actively plan for nuclear-first use. The primary driver is still about how to deter American nuclear-first use. But now they want to be able to deter not only massive-scale nuclear first strikes, but also limited nuclear use. And you think the no-first-use pledge still is in force, that it sits comfortably enough with current Chinese political doctrine? I mean, again, I don’t see evidence of China wanting to violate its no-first-use policy. The mainstream perception in Beijing is still that despite China’s recent nuclear expansion, the U.S. enjoys an obvious advantage in the nuclear area. It doesn’t make sense for China to want a conventional war to escalate into the nuclear domain. That’s where the U.S. maintains obvious strengths and advantages. But when it comes to no-first use, the story is always more nuanced. I do think at the political level, China is generally sincere in making that commitment. But again, it becomes more complicated when it comes to the operational details. But because rather authoritative Chinese military writings, including those approved by senior military officials, they talked about scenarios under which China would threaten nuclear-first use if certain critical targets are facing a conventional threat. And of course, in their view, in Chinese military view, nuclear deterrence is more about making gestures of potential nuclear use to actually influence enemy behaviors, to deter them from taking actions that would threaten China rather than actually employing nuclear weapons. So how to basically manipulate risk and therefore shape enemy behavior is a major part of China’s nuclear deterrence thinking. Their military writings reflect this; they leave the space for making nuclear threats in a conventional war to deter certain conventional threats. According to this thinking, to do this wouldn’t necessarily violate China’s no-first-use policy. As long as China doesn’t follow through on the threat and actually employ nuclear weapons, they would say the no-first-use commitment has held. But of course, this already introduces controversy into China’s no-first-use commitment because in a conventional crisis, China starts to threaten nuclear use, maybe not necessarily explicitly but implicitly and subtly by referring to China’s nuclear capability. Chinese leaders talk about China’s nuclear strengths, conducting nuclear exercises with nuclear-capable military platforms. There are many ways to send those subtle messages and remind the enemy that a conventional war could escalate to the nuclear level. So China can do all of those things and achieve broader security and even geopolitical benefits. That’s exactly, I think, what Russia has done in Ukraine through Russian nuclear saber-rattling. When you start making even implicit nuclear threats in a conventional war, your enemy has to treat your nuclear signaling as very serious, and they would have to consider there’s a real possibility of nuclear use. To give yourself the space to threaten nuclear use, some would argue that this already violates the spirit of no-first-use and makes your enemy no longer trust the credibility of no-first-use. Thus, it undermines the reassuring benefit. Is this debate happening inside of China? Is this a public debate that you’re hearing among nuclear strategists within China? Not really, because even those relatively authoritative military writings are so-called internal references. I see. Nei bu cang kao. They are not classified documents, but they are supposed to be only for internal circulation. Before, about 10 to 15 years ago, when the political environment was relatively relaxed, foreign scholars could still go to the campus of the National Defense University or the Chinese Academy of Military Sciences. They could visit their stores on the campuses and buy those internal circulation books. That’s how these documents already spread to foreign countries. But inside China, their circulation is very strictly limited. Many Chinese civilian officials, including senior diplomats in charge of arms control policies, are not aware of these military documents. They don’t know what is written inside. Even many Chinese nuclear experts do not have access to such documents, making it hard for them to have a debate. Let’s talk about this essay you wrote in Foreign Affairs recently. You described China as seeing itself in a strategic stalemate with the United States. What does that mean in practical terms for Chinese military planners? What do you mean by strategic stalemate? It’s just a term that I see an increasing number of Chinese foreign policy experts using. I think it reflects this increasingly internal agreement that when it comes to the comparison of comprehensive national power, China has achieved a qualitative change by significantly narrowing down the power gap to the point that China can now be roughly described as on the same level. So it’s a synonym; stalemate is a synonym for parity. It’s a rough balance of power. The obvious question is, how has Trump’s second term, really since his inauguration, shaped Beijing’s calculus? Do you think that it has changed compared to the period 2017 to 2020? I think it has reinforced the Chinese belief that China is now formally entering a new era of strategic balance or strategic stalemate with the United States. Trump has accelerated American decline, both regarding American material power because of his very problematic economic policies, tariffs, and so forth. And also the overall chaos in the U.S. government system that he introduced really undermines U.S. long-term power development compared with China. And also, significant decline of U.S. soft power and international standing. So, U.S. has much less moral leverage it can use against China. The fact that by standing the ground and pushing back against U.S. tariff, that was perceived to have led to Trump’s concessions. All of these together, I think, reinforce the Chinese perception that indeed China is increasing in a position to negotiate equally, from a position of strength. Yeah, from maybe not strong, but at least equality. To negotiate with dignity and resolution, China can win a fair agreement. That’s the approach China should take. I think this will influence not only China’s economic negotiating approach but also its overall foreign policy and security approach regarding the United States. Should we understand this very rapid increase in China’s nuclear capabilities as a sign that its traditional strategic patience is now wearing thin, that it’s entering a period now of more strategic assertiveness? Again, that’s not Chinese perception, right? The Chinese self-perception is that what is causing increasing instability is growing American hostility. By building up and demonstrating China’s strategic power, China is introducing greater stability into the relationship by making it harder for the United States to continue aggression and bullying against China. So, that’s where the perception gap is. Do you think that is a gap? Do you think that that’s not well understood in the American strategic community? Do you think that it’s now basically conventional wisdom that what we should see, what we should understand is that China is entering a period of more assertiveness? I think it’s very valuable that you bring, as I said, I’m all about strategic empathy, and you are too. This pervades all of your writing. Clearly, you think that this is something where minds in Washington need to be changed, in the Pentagon need to be changed. They need to understand. I think minds in both Beijing and Washington need to be changed. On Washington’s part, some U.S. China experts have a more nuanced view about Chinese perception. They understand China has genuine grievances, and some of its behaviors are driven by a genuine sense of insecurity. On Beijing’s part, even though there is a strong self-perception of China simply trying to defend its legitimate interests, China is trying to stabilize the bilateral relationship. The other side of the country is, in the current Chinese system, its capacity to reflect critically on its own behavior and to adequately take into consideration other countries’ threat perceptions and concerns towards China is very limited. Therefore, it’s hard to tell whether this growing capability would actually lead to more aggressive Chinese behavior. Because China always sees everything it does as purely self-defensive and peaceful, driven by legitimate interests. But it’s hard to say for sure that when China becomes more powerful, it would never adopt expansionist policies. If China is able to secure Taiwan and such a sea with its growing military power, would China aim at a greater sphere of military influence or dominance after achieving its current territorial interests? We simply don’t know. We have seen signs of, for example, Chinese experts who are affiliated with the system, increasingly making discussions about that question, Japan’s sovereignty over Okinawa and the Ryukyu Islands. I can understand when China thinks Japan is picking troubles with China, this is an issue China can use to push back and teach Japan a lesson. But when you start to hint at settled territorial disputes, it’s understandable some countries will think this trend will grow if China becomes more powerful going forward. You just now mentioned Taiwan, which, of course, is the flashpoint that everyone is aware of and certainly is the most consequential flashpoint. How real is the risk right now that China sees its moment to enact some kind of unification or reunification is slipping away? And how might nuclear planning play into that sort of scenario? I think the near-term risk is relatively low. If we look at the official results of the Taiwan Work Conference that was conducted in February this year, also followed by the two sessions in March this year, which issued statements on Taiwan, among other things. The general policy direction, the direction set by these two high-level meetings was very clear: that is to continue the existing policy of incrementally influencing Chinese capability to control the direction of the cross-strait relationship, and focusing on peaceful measures to unify. There is a lot of continuity there. On the military capability side, I think China still needs some more time to feel really ready to conduct massive operations if necessary. We have seen reported China’s construction of a really large-scale underground military command control facility in the southwestern part of Beijing that is multiple the size of the Pentagon. We have seen China building specially designed barges that can dock Chinese military ships and quickly transport soldiers and equipment to Taiwan without ports and other supporting facilities. Those large military programs still need time to materialize and become operational. At the high senior leadership level, there is just so much internal instability within the military, the purging of senior generals, including officials from the eastern district. This will undermine China’s near-term capacity. So, I think China will want more time to be military ready. Of course, I am not saying the military option is a priority option for China, but simply to feel militarily ready if necessary will take some more time. I think China wants to wait and see how the overall global geopolitical trend develops in the next few years. Will the U.S. become further isolated under Trump? Will the international community become more indifferent to the Taiwan issue? Would Western countries, in particular, continue to really depend on Taiwan as a key source of semiconductor products? How would international public opinion evolve regarding Taiwan and China? I do think some strategists in Beijing see Trump as a special opportunity, given his apparent interest in avoiding massive military conflict with China or Russia, and his general lack of interest in sacrificing American blood and treasure on foreign soil that is not necessarily so existential to American interest. There is also internal disagreement within the Pentagon about how important Taiwan is to American military interests in the region. We are seeing more and more people from prominent think tanks publishing papers that suggest, what was a few years ago quite unsayable: that Taiwan does not represent a vital strategic American interest and that it’s simply not worth it. However, I think the risk of military conflict could grow, perhaps shortly before the ending of the second Trump administration. Even though China has been prioritizing peaceful means of promoting unification, the goal has become more ambitious. The previous emphasis was on countering or preventing independence. Now, the emphasis is more on promoting or advancing unification; of course, peaceful measures are prioritized. But peaceful measures appear to include systematic efforts to expand Chinese influence in Taiwanese society. We have seen China working more and more with local organizations, political parties, civil society actors, and media organizations in Taiwan to promote China’s narrative, essentially making the case that to voluntarily unify is the least bad option for Taiwanese people. China has been handing out Chinese national IDs to some Taiwanese citizens, and even reportedly buying pledges of loyalty from actively serving Taiwanese military personnel. For China to make peaceful unification work, it has to steadily increase its influence across Taiwanese society. However, I believe those measures will receive strong pushback. I’d like to ask you how you assess Taiwan’s countermeasures to this. There’s this 17-point anti-infiltration plan that you talked about in that Foreign Affairs piece. Do these actually deter Beijing or maybe prompt even sharper responses? Unfortunately, they are leading to sharper responses because those countermeasures are viewed by Beijing as DPP governments resisting peaceful unification. Therefore, I think China has stepped up military operations near Taiwan as a result. This concern leads me to believe that even peaceful measures to promote unification could eventually lead to higher military tensions, potentially causing incidents or even military conflicts. Let’s come back to nuclear strategy a little bit. And I want to talk about your arms control today essay, which introduces this idea of a dangerous parallax in how Washington and Beijing perceive one another in their nuclear doctrines. Can you walk us through what you mean by that? I’m not sure everyone knows what the word parallax actually means. Some people are hopeful that under Trump, maybe the U.S.-China nuclear relationship can become more stable. Trump himself has repeatedly mentioned potentially having denuclearization talks with China. His interest in having direct leadership-to-leadership meetings and raising these high-stakes issues is the right approach to engage China since nuclear decisions are made by senior political leaders, especially Mr. Xi himself. For Trump to want to have such high-level engagements directly is a more effective way than starting engagements with low-level bureaucrats. The problem is the mainstream American strategists, who are so concerned about China’s nuclear buildup. This is partially made worse by the lack of transparency on Beijing’s part. China has not offered any information about why it is building up, what endgame it seeks to achieve, how many more weapons it needs, and what the rationale behind the buildup is. So we end up falling back on worst-case assumptions about all this. Indeed, the worst-case assumption includes speculation about perhaps Beijing wanting to achieve nuclear parity with the United States, or even worse, maybe Beijing wants to acquire the capability to disarm the United States, to be able to actively attack American nuclear weapons and neutralize American second-strike capability. Many serious American military strategists believe that China has a plan to build thousands or more nuclear weapons in the next 10 to 15 years, making China a larger, more powerful nuclear state than the United States. Many are worried about China wanting to fundamentally change its nuclear doctrine to one that’s based on first use. The U.S. also worries that Russia and China might join hands in conducting a joint nuclear strike against the United States. Or at least if the U.S. and Russia are already engaged in a nuclear war, maybe China might be tempted to conduct so-called opportunistic aggression by starting a second nuclear conflict with the United States. So the mainstream planning is increasingly driven by this worst-case scenario thinking. We have seen a consensus being formed in Washington that the U.S. needs to reverse its decades of nuclear reduction and instead start enhancing American nuclear capability. I worry that this potential American nuclear increase could be interpreted by Beijing as American efforts to maintain its nuclear primacy or worse, actively undermine China’s nuclear deterrence and lead to China’s further reactions. Yeah, we get into a very terrible and possibly apocalyptic feedback loop here. One where the erosion of expert consensus and institutional continuity might be read by Chinese leaders as instability or unreliability. It prompts the Chinese leaders to adopt more rigid strategic postures in response to that. In turn, those can reinforce the hawkish tendencies that we see already so abundant in Washington. So that, yeah, that’s a very worrisome loop that we’re in. How do we break that? I mean, part of it you identified already if Beijing were more transparent about what it’s doing. And, of course, I think if both sides just tried to exercise a little more cognitive empathy, a little more strategic empathy toward one another. Well, yeah, as you said, to start with, both sides need to develop more nuanced understandings about the other’s thinking. For Washington, it’s concern about China wanting a bigger arsenal than the United States. That’s far-fetched. China wanting to use nuclear weapons first is totally detached from internal Chinese discussion. I haven’t seen almost zero evidence of that happening. On the Chinese side, there is also a strong reluctance to consider American concerns. China thinks all of its nuclear buildup is stabilizing, and the United States shouldn’t fear at all. But that’s not the case. The U.S. is genuinely concerned about this unprecedented situation of facing two nuclear near-peers, Russia and China, who have very close cooperation, including in military areas. So, China also needs to be willing to consider American legitimate concerns and to start, therefore, explaining what its goal is. So far, that has not happened. Both sides, I think, can engage in more critical self-reflection and be more willing to take the other’s concerns into consideration. Let’s talk about the U.S. plans for anti-missile systems, for expansion of anti-missile, of missile defense systems, Trump’s Golden Dome, and other things like that. How much do you think that China’s modernization efforts are tied to perceptions of American intentions that way? And again, how much of the thinking behind these more robust missile defense systems are tied to American thinking about Chinese intentions? Well, you know, historically, China’s concern about U.S. homeland missile defense has been a major technical level driver of China’s modernization effort. But regarding Golden Dome, I think China’s views are more nuanced. If we look at China’s public-facing announcements and statements and analysis, they are, as usual, very critical. The recent China-Russia joint statement on maintaining global strategic stability heavily criticizes Golden Dome, saying that it tries to delink strategic offensive weapons from strategic defensive weapons. So it tries to ignore the importance of limiting missile defense as a key precondition for nuclear arms control diplomacy. It portrays Golden Dome as hugely destabilizing and threatening to global strategic stability. We have also seen Chinese experts repeating these arguments. But I think if you look deeper into China’s expert discussion internally, their views are much more nuanced, right? They recognize this program was driven by Trump’s personal interest to emulate the success of Israel’s Iron Dome program to appear that Trump is protecting American people. But the entire decision-making process was very chaotic and not very coherent, especially the fact that Trump wants Golden Dome to become operational before the end of his term. That’s totally unrealistic. The massive scope of the envisioned Golden Dome is going to be so challenging to become reality. Many of these key components, including space-based interceptors, won’t be able to intercept a large number of Russian or Chinese ICBM attacks. Even to defend against a couple of missile attacks from North Korea would require hundreds, if not thousands of space-based interceptors. To defend against a larger scale attack from Russia and China would require a much larger scale of constellation. So that’s going to be astronomically expensive and technologically challenging, which is true for many non-kinetic missile defense technologies envisioned under this program. Many Chinese experts actually say that this is driven by a chaotic decision-making process, fueled by lobbying from special interest groups from the military defense complex. It’s an ego-driven project that’s wholly unrealistic. And it’s so expensive it is going to compete with other U.S. defense priorities, including the U.S. nuclear modernization program. So I think China is not as worried today as before. They actually think this program could be very wasteful on the American side, making the U.S. less able to prioritize the most needed military capabilities. It would be ironic if this update to Reagan’s Star Wars ended up helping to bankrupt America. Yeah, it could impose unnecessary costs on the American side. I’m actually very comforted in knowing that there’s such a nuanced understanding of it in the strategic community in Beijing. That’s very comforting. You mentioned just now about the number of interceptors that would be necessary even to stop a limited attack by North Korea. So you wrote an essay recently, which I thought was really interesting, about the North Korea factor and the sort of Putin-Kim alliance and how that reads in Beijing. You look at the risks stemming from this growing closeness between Putin and Kim Jong-un. Why is China so reluctant to play a more proactive role in constraining North Korea right now, though? I think one important factor is that North Korea is just a tough ally to manage. It’s so defiant. I’ve likened them in the past to having a vicious dog in your backyard. It bites you, it bites the neighbors. But, you know, once in a while, it’ll help you defend against a home invasion. You want to kind of take it out back and shoot it. But it’s a pretty vicious thing. And it’s sometimes useful. But I feel like that’s Beijing’s attitude toward North Korea sometimes. The horse has left the barn. North Korea is now fully nuclearized. It is immune from even an ally’s pressure. During the 2016-2017 period when China joined forces with the United States and other countries to roll out UN Security Council resolutions imposing sanctions on North Korea, North Korea was really pissed off and implicitly threatened China by hinting that its missiles could reach most parts of China. So it’s hard. I think for its own security interests, China is more and more reluctant to impose real pressure on North Korea. Even for Beijing to maintain its previous position on the eventual denuclearization of the Korean Peninsula is now unacceptable to Pyongyang. That’s a major reason we have seen the cooling down of the bilateral relationship. Pyongyang is actively pressuring Beijing to drop this longstanding position and to basically follow the Russian step. Because Russia has publicly accepted North Korea’s nuclear status, Russian senior officials have openly acknowledged that reality. So it’s North Korea that is now actively putting pressure on Beijing to change its position regarding North Korea’s nuclear weapon program. You argue that China would, under certain circumstances, intervene if Kim Jong-un actually crosses certain red lines. What are those red lines, and how are they shifting in Beijing’s eyes? I think what I try to say in that essay is there are elements in North Korea, Newcastle, that are particularly worrisome for Beijing. I think Beijing is now really concerned about North Korea’s intercontinental-range ballistic missiles, which threaten the United States. But North Korea is increasingly dependent on tactical nuclear weapon systems. It has developed a wide range of very diverse types of tactical nuclear weapons: Land-based Air-based Sea-based Its nuclear doctrine clearly states that North Korea is ready to use those tactical nuclear weapons early on in a future conflict. And that conflict would be South Korea, across the 38th parallel. That’s the assumption. South Korea, perhaps U.S. bases in South Korea and Japan. North Korea has forward-deployed some of these tactical systems closer to the inter-Korean border. At least the doctrine states it pre-delegates the launch authority of these tactical nuclear weapon systems to the relevant military units. All of these factors would make China worry that the risk of North Korea deliberately or incidentally starting a nuclear conflict through the threat or use of tactical nuclear weapons is growing. This could drag China into a nuclear conflict on its doorstep. Thus, I think that’s the area of common interest between China and other regional countries. China has been promoting some confidence-building and risk-reduction measures regarding nuclear weapons at an international level, including a no-first-use policy. China wants more nuclear weapons states to adopt no-first-use. This may be one area that China can coordinate with other countries to see if it can make North Korea adopt less destabilizing nuclear policies and make nuclear weapons less likely to be used on the Korean Peninsula. Fantastic. If you had the opportunity to talk to a skeptical U.S. policymaker who sees China as enabling Pyongyang’s nuclear program, even though it insists that it supports denuclearization, how would you convince them? What would you say in evidence that Beijing can actually be induced into playing a more positive role? Well, I think historically, China never wanted to deliberately assist North Korea’s nuclear program. But China, on the other hand, also didn’t want to stick its neck out to forcefully denuclearize North Korea. When there was still an opportunity to stop North Korea’s nuclear advancement through economic sanctions, it required Beijing to shoulder the cost of imposing that type of economic threat on North Korea. We needed crippling economic sanctions on North Korea and to maintain that type of crippling sanctions long enough to destabilize North Korea’s economic and social system threatening its regime security. We needed that type of economic pressure. Beijing wasn’t willing to go that far. It didn’t feel it could threaten North Korea to that level because it worries about its own security interests and its influence over the Korean Peninsula. Beijing has been more sympathetic to North Korea’s perceived need for a strategic deterrent. It also shared, to some extent, North Korea’s concern about the United States and its Western allies. For all these reasons, from Western countries’ perspective, Beijing didn’t cooperate enough and therefore contributed to the failure of the international community to stop North Korea. But today, I just don’t see a strong Chinese political will to, again, put its own interests on the line in order to force North Korea to change its nuclear decision. Right. And how do you see Beijing in terms of the pressure it might put on Moscow to curb its newfound ally, North Korea? I mean, there’s been a lot of pressure over North Korean soldiers fighting in Ukraine, for example, and other things. Is there a way that Beijing could be brought onto the side of NATO, the EU, and the United States on the Ukraine conflict, at least in terms of curbing North Korean participation? I’m not sure how worried Beijing is about North Korean troops in Ukraine. In fact, closer cooperation or mutual cooperation between Russia and North Korea helped to reduce the Chinese burden. China didn’t need to directly support Russia in Ukraine. And China didn’t need to directly support North Korea. Russia can provide, could provide North Korea with more economic aid, and North Korea can help out Russia. So it reduces the Chinese burden to get involved in sensitive areas to directly support its two allies. But of course, China has its concerns regarding Russian-North Korea cooperation. I think, especially over potential Russian sharing of sensitive military technologies with North Korea, that could, depending on how far Russia is willing to go to help North Korea with its development of nuclear and missile capabilities, that Russian cooperation could undermine international norms on non-proliferation, in which China still has a strong interest. China also worries that South Korea and Japan might become more worried about North Korea’s improving strategic capabilities. And that concern has driven more and more internal discussions in Seoul and Tokyo for indigenous deterrent capabilities. South Korea, if not for the election of a progressive president, could have a real chance of pursuing its own nuclear capability. So, China worries about all the spillover effects on its own security because of Russia-North Korea cooperation that strengthened North Korea’s military capability. North Korean troops in Eastern Europe make European countries create a linkage between European security and Asia-Pacific security. From China’s perspective, it gives European countries the excuse to strengthen their military presence in the Asia-Pacific region to talk about a NATO mission for Asia. I’ve been talking to Zhao Tong, who is a senior fellow in the nuclear policy program here at the Carnegie Endowment for International Peace. What a fantastic mind of information you are, and you present your knowledge so clearly, and it’s really admirable. I look forward to talking to you again about this because this is certainly not the end of—we’ve not exhausted the issue of the nuclear situation. Let’s move on now, and first to our segment called Paying It Forward, where I’d like you to identify just somebody who’s working preferably in your field, another researcher maybe doing work on arms control or on the nuclear issue, who we should be paying attention to. Well, when I was at Carnegie Tsinghua Center in Beijing, I worked with a young scholar, David Logan, who was a young ambassador at Carnegie Tsinghua Center, and he is now really a rising star in the field of nuclear deterrence, arms control, and non-proliferation. He has done very impressive work on China’s nuclear policy, looking deeply into the PLA organization’s structure of the rocket force, and also conducting public opinion surveys in China about how the Chinese general public thinks about nuclear weapon issues, which is a very interesting research topic. David Logan, I will make sure to check out his work. Yeah, he’s now a professor at the Fletcher School of Tufts University. Great recommendation. And what about a recommendation? Do you have a book or a film or some music or something that you would like to recommend for our listeners? It doesn’t even have to be about China, let alone nuclear arms racing. Yeah, I happen to be watching the American TV series Yellowstone, which tells the story of a ranch owner in Montana trying to protect his ranch. I think it’s interesting for our conversation because it demonstrates some of the mindset of some American people who see there are just bad people in the world and they see the need to use all means available to defeat these bad people, including killing them. I think that mindset is really telling because right now I’m afraid an increasing number of American strategists start to see China as just a bad actor, and they are increasingly disillusioned about resolving U.S.-China disagreements through persuasion, talks, and diplomacy. They are losing hope that they can ever change Chinese thinking or behavior. So it’s really dangerous when both sides—I think a similar sentiment is also growing in Beijing towards Washington. When Beijing thinks the U.S. is never going to change its view about the Chinese regime, it will preserve its own hegemony at all costs. Right. It contributes to the both sides’ determination to outcompete and win over the other side, and legitimizing the use of unlimited measures. And you get all this from Yellowstone. Yeah. You can do everything to defeat your enemy, and every method is justified. So it’s interesting to understand that mindset under today’s geopolitical environment. Well, I’ve got a television recommendation as well, although it is even more maybe nihilistic and horrible. It’s called Gomorrah. It’s an Italian television show. I’m now just starting the second season of it, but it’s about the crime families and criminal organizations in Naples, Italy. We’re more familiar with the Mafia, of course, in Sicily, but this is about the Camorra crime syndicates in Naples. Very violent, really some of the most amoral characters you’ve ever encountered, but it’s shot really well. The storytelling is very compelling, and the subtitling has been done very well, as far as I can tell. So I’m quite addicted to it, and it’s very good. So that’s my recommendation. And with that, I just want to thank you so much for spending so much time talking to me and having me here in your office to speak to you. I really look forward to speaking to you again, Zhao Tong. Thank you so much. I look forward to continuing to follow your work. You’ve been listening to the Cineca Podcast. The show is produced, recorded, engineered, edited, and mastered by me, Kaiser Guo. Support the show through Substack at CinecaPodcast.com, where there is a growing offering of terrific original China-related writing and audio. Email me at CinecaPod.com if you’ve got ideas on how you can help out with the show. Don’t forget to leave a review on Apple Podcasts. Enormous gratitude to the University of Wisconsin-Madison’s Center for East Asian Studies for supporting the show this year. Huge thanks to my guest, Zhao Tong. Thanks for listening, and we’ll see you next week. Take care. window.tocIndex = { "index": [ { "index_sentences": "Welcome to the Seneca Podcast, the weekly discussion of current affairs in China.", "section_level": 1, "section_title": "Introduction to the Podcast" }, { "index_sentences": "My guest today is someone who embodies this kind of disciplined grasp of adversary perceptions that Robert Jervis spent a career urging—the kind of security dilemma sensibility that calls not just for caution but for informed empathy.", "section_level": 1, "section_title": "Guest Introduction: Zhao Tong" }, { "index_sentences": "Well, in fact, China has never provided a full set of explanations on its nuclear doctrine.", "section_level": 2, "section_title": "China's Nuclear Doctrine and Strategy" }, { "index_sentences": "When we're talking about the nuclear arsenal of China right now, a few years ago, I remember everyone said it was simply 200.", "section_level": 3, "section_title": "Size and Growth of China's Nuclear Arsenal" }, { "index_sentences": "In your recent Asia policy piece, you argue that it's actually internal political drivers—especially legitimacy narratives and Xi Jinping's personal leadership style—that now really significantly shape China's nuclear posture.", "section_level": 3, "section_title": "Drivers of China's Nuclear Modernization" }, { "index_sentences": "How should we understand the symbolic role of nuclear weapons in the great rejuvenation of the Chinese nation, in Xi Jinping's view, in that narrative?", "section_level": 4, "section_title": "Symbolic Role of Nuclear Weapons" }, { "index_sentences": "You've described, and I think just now you've given examples of how this manifests, a real erosion of institutionalized strategic debate in China.", "section_level": 3, "section_title": "Internal Nuclear Decision-Making Process" }, { "index_sentences": "So is the upshot that we are seeing a de facto shift away from minimum deterrence?", "section_level": 3, "section_title": "De Facto Shift from Minimum Deterrence?" }, { "index_sentences": "And you think the no-first-use pledge still is in force, that it sits comfortably enough with current Chinese political doctrine?", "section_level": 3, "section_title": "Nuances of China's No-First-Use Pledge" }, { "index_sentences": "You described China as seeing itself in a strategic stalemate with the United States.", "section_level": 2, "section_title": "US-China Strategic Relationship" }, { "index_sentences": "The obvious question is, how has Trump's second term, really since his inauguration, shaped Beijing's calculus?", "section_level": 3, "section_title": "Impact of Trump's Presidency" }, { "index_sentences": "Should we understand this very rapid increase in China's nuclear capabilities as a sign that its traditional strategic patience is now wearing thin, that it's entering a period now of more strategic assertiveness?", "section_level": 3, "section_title": "Strategic Assertiveness vs. Stability" }, { "index_sentences": "I want to talk about your arms control today essay, which introduces this idea of a dangerous parallax in how Washington and Beijing perceive one another in their nuclear doctrines.", "section_level": 3, "section_title": "The Dangerous Parallax in US-China Nuclear Perceptions" }, { "index_sentences": "Let's talk about the U.S. plans for anti-missile systems, for expansion of anti-missile, of missile defense systems, Trump's Golden Dome, and other things like that.", "section_level": 3, "section_title": "US Missile Defense Plans and China's Response" }, { "index_sentences": "You just now mentioned Taiwan, which, of course, is the flashpoint that everyone is aware of and certainly is the most consequential flashpoint.", "section_level": 2, "section_title": "The Taiwan Flashpoint" }, { "index_sentences": "How real is the risk right now that China sees its moment to enact some kind of unification or reunification is slipping away?", "section_level": 3, "section_title": "Risk Assessment & China's Approach to Taiwan" }, { "index_sentences": "I'd like to ask you how you assess Taiwan's countermeasures to this.", "section_level": 3, "section_title": "Taiwan's Countermeasures" }, { "index_sentences": "So you wrote an essay recently, which I thought was really interesting, about the North Korea factor and the sort of Putin-Kim alliance and how that reads in Beijing.", "section_level": 2, "section_title": "The North Korea Factor" }, { "index_sentences": "Why is China so reluctant to play a more proactive role in constraining North Korea right now, though?", "section_level": 3, "section_title": "China's Reluctance to Pressure North Korea" }, { "index_sentences": "You argue that China would, under certain circumstances, intervene if Kim Jong-un actually crosses certain red lines.", "section_level": 3, "section_title": "North Korea's Red Lines for Beijing" }, { "index_sentences": "If you had the opportunity to talk to a skeptical U.S. policymaker who sees China as enabling Pyongyang's nuclear program, even though it insists that it supports denuclearization, how would you convince them?", "section_level": 3, "section_title": "Addressing Skeptical US Policymakers on North Korea" }, { "index_sentences": "And how do you see Beijing in terms of the pressure it might put on Moscow to curb its newfound ally, North Korea?", "section_level": 3, "section_title": "China's Concerns on Russia-North Korea Cooperation" }, { "index_sentences": "Let's move on now, and first to our segment called Paying It Forward, where I'd like you to identify just somebody who's working preferably in your field, another researcher maybe doing work on arms control or on the nuclear issue, who we should be paying attention to.", "section_level": 1, "section_title": "Recommendations" }, { "index_sentences": "And what about a recommendation?", "section_level": 2, "section_title": "Book/Film/Music Recommendation" }, { "index_sentences": "And with that, I just want to thank you so much for spending so much time talking to me and having me here in your office to speak to you.", "section_level": 1, "section_title": "Conclusion and Credits" } ] }; window.faq = { "qas": [ { "answer": "Zhao Tong argues internal political drivers, specifically legitimacy narratives and Xi Jinping's personal leadership style, are the main drivers. He suggests traditional concerns about U.S. non-nuclear capabilities are longstanding but don't explain the recent abrupt and massive buildup, which appears to be a \"political mandate\" from the top seeking greater political value, international status, and leverage to compel \"peaceful coexistence\" by demonstrating strategic capability.", "index_of_source": "In your recent Asia policy piece, you argue that it's actually internal political drivers—especially legitimacy narratives and Xi Jinping's personal leadership style—that now really significantly shape China's nuclear posture.", "question": "Why does Zhao Tong argue internal politics is the main driver of China's recent nuclear buildup?" }, { "answer": "According to U.S. government assessments, China's nuclear warhead count has grown from a little more than 200 in 2019 to more than 600 in 2025, almost tripling in just several years.", "index_of_source": "As recent as 2019, the U.S. government assessed China to still have a little more than 200 nuclear weapons.", "question": "How large is China's nuclear arsenal estimated to have grown in recent years according to U.S. assessments?" }, { "answer": "Previously, China was very careful to hide its silo-based ICBMs, building them in mountainous areas, taking great efforts to hide locations, and even building decoys to increase survivability. The recent construction of over 300 silos in plain sight in northwestern China, easily visible to U.S. satellites, appears to be driven by a desire to publicly demonstrate capability and a sense of urgency.", "index_of_source": "The fact that China used to be very careful to hide its nuclear facilities' locations.", "question": "How does China's recent construction of hundreds of ICBM silos in plain sight contrast with its previous practices?" }, { "answer": "According to the military writings, threatening nuclear use or subtly signaling nuclear capability in a conventional war to deter certain threats wouldn't necessarily violate the no-first-use policy *as long as China doesn't follow through and actually employ nuclear weapons*. This perspective views nuclear deterrence partly as manipulating risk to shape enemy behavior, leaving space for threats without breaking the literal non-use commitment, though critics argue this undermines the spirit and credibility of the pledge.", "index_of_source": "But because rather authoritative Chinese military writings, including those approved by senior military officials, they talked about scenarios under which China would threaten nuclear-first use if certain critical targets are facing a conventional threat.", "question": "How does the discussion of Chinese military writings suggesting threats of nuclear first-use in a conventional war scenario square with China's stated \"no-first-use\" policy?" }, { "answer": "China's self-perception is that the increasing instability is caused by growing American hostility. By building up and demonstrating strategic power, China believes it introduces greater stability into the relationship by making it harder for the United States to continue aggression and bullying. This logic is easily accepted internally in China, contrasting with the military technical calculation of minimum deterrence.", "index_of_source": "The Chinese self-perception is that what is causing increasing instability is growing American hostility.", "question": "Why does China believe its rapid nuclear buildup, seen as assertive by the U.S., actually contributes to stability?" }, { "answer": "\"Strategic stalemate\" is a term reflecting the increasing internal agreement among Chinese foreign policy experts that China has achieved a qualitative change in comprehensive national power, significantly narrowing the gap with the United States to the point where the two countries can now be roughly described as on the same level or in a rough balance of power.", "index_of_source": "It's just a term that I see an increasing number of Chinese foreign policy experts using.", "question": "What does Zhao Tong mean by the term \"strategic stalemate\" when discussing the current U.S.-China relationship?" }, { "answer": "China is concerned about North Korea's ICBMs threatening the United States, but particularly worried about North Korea's increasing reliance on diverse tactical nuclear weapons, its readiness to use them early in a conflict (likely with South Korea/U.S. bases), and the pre-delegation of launch authority. These factors increase the risk of North Korea deliberately or incidentally starting a nuclear conflict on China's doorstep, potentially dragging China in.", "index_of_source": "I think what I try to say in that essay is there are elements in North Korea, Newcastle, that are particularly worrisome for Beijing.", "question": "What are the potential risks China is most concerned about regarding North Korea's nuclear program and doctrine?" }, { "answer": "Beijing views North Korea as a tough, defiant ally that has shown it will push back against Chinese pressure (e.g., implicit threats during 2016-17 sanctions). China is increasingly reluctant to put its own security interests or influence over the Korean Peninsula at risk by imposing the crippling sanctions needed to force North Korea's denuclearization. Beijing is also somewhat sympathetic to North Korea's perceived need for a deterrent and shares some concerns about the U.S. and its allies.", "index_of_source": "Why is China so reluctant to play a more proactive role in constraining North Korea right now, though?", "question": "Why is Beijing reluctant to exert stronger pressure on North Korea despite concerns about its nuclear program and alliance with Russia?" } ] };

2025/6/25
articleCard.readMore

NYC Mayoral Candidate Zohran Mamdani on Abundance, Socialism, and How to Change a Mind

NYC Mayoral Candidate Zohran Mamdani on Abundance, Socialism, and How to Change a Mind Before today’s show, a personal announcement. After almost 17 years at The Atlantic Magazine, I have just officially moved my writing full-time to Substack, the newsletter platform. If you like this show, if you’re a fan of plain English and my work across abundance, the anti-social century, happiness, loneliness, science, and technology, I think you’ll love what I’m building here. And on a personal note, it would mean a lot to have you sign up. So check out DerekThompson.Substack.com or just search DerekThompsonSubstack on Google or wherever. Enter your email and you are done. That’s DerekThompson.Substack.com. Today, Abundance, the American left, and Zoran Mamdani, the Democratic Socialist candidate for mayor of New York City. In the last few months, I’ve been surprised by the reaction to Abundance, the book I co-wrote with Ezra Klein, on the political left. The online discourse has been quite negative. On Twitter, Blue Sky, Reddit, YouTube podcast, and on articles on Left Wing and Left Populist sites and magazines, it has been a surprising wave, a large wave of negative feedback. In fact, the volume of conversation around this book has reached such a level that New York Magazine’s back page, a grid of the cultural zeitgeist called the Approval Matrix, recently plotted what they called, “the overabundance of abundance discourse” in their highbrow despicable quadrant. But if you pull your mind away from social media and YouTube and this online discourse, if you listen to and watch what politicians and people running for political office have been saying about this book and our project, I think what you’ll see is that the response has been notably different. It’s not just centrist and moderate mayors, governors, senators who have called out the book as a useful model for the future of liberalism in the Democratic Party. It’s often politicians who represent the same leftist causes whose posters are constantly criticizing the book. For example, Bernie Sanders’ devotees have repeatedly bashed Abundance. But Representative Ro Khanna, an outspoken advocate of Bernie’s signature policy proposal, Medicare for All, announced his support for Abundance on several occasions. While several people have accused the book of ignoring policies to reduce welfare, Wes Moore, the progressive Maryland governor whose private sector career was devoted to reducing poverty, said in a recent speech that Democrats have to change from being the party of no and slow to being the party of yes and now. This is a direct call out to a theme of Abundance, that liberals given power often pass laws that are larded down with complicated procedure. And then there’s Zoran Mamdani, the Democratic Socialist candidate for mayor of New York City. Mamdani and I have very different politics on a range of issues, including housing, affordability, education, levels of taxation, and spending. Yet Mamdani has, in the last few weeks, embraced what he has called explicitly an agenda of abundance. He’s told podcasts like Pod Save America that he thinks leftist critics of Abundance have oversimplified the book and that our approach to making government work better is exactly what the left needs in this moment. Now, I saw some people point to Mamdani’s name checks of the book and say, “oh, this is great. This is what persuasion looks like.” But I also saw some people point to his comments and say, “it’s a ruse. Beware. Stay away. He’s co-opting abundance to redirect it toward goals that have nothing to do with this book.” I wanted to talk to the man himself. So I was very gratified when, in a very busy week for me as I’ve just switched jobs and am now moving cities back to Washington, D.C., and in a very, very busy week for him, since he’s running to become mayor of America’s biggest and richest city. Mamdani and I nonetheless found 30 minutes to sit down and talk calmly last Saturday about abundance and the left, how we agree, where we disagree, why government deficiency ought to be a virtue of all leaders, but especially those on the left who want government to do much more. And finally, how to change our mind. This is the last subject of the interview, and it’s the one that I honestly came away thinking about the most. The trick that ideology plays on the mind is that it convinces flawed thinkers of their infallibility. But there’s no ideology— not left or right or centrist—that is guaranteed to have the perfect answer to every problem for all time. It is a delusion to believe that such a thing could even exist. The world is dynamic. Its problems are strange. And seeing reality clearly, continuing to see reality clearly, requires that we have the courage to talk to people we do not agree with, who see the world differently than we do, with the hope that we can learn from them. I’m Derek Thompson. This is Plain English. Zoran Mamdani, great to talk to you. No, it’s so lovely to be on. Thank you so much for having me. You recently delivered a speech in which you said this, “government must deliver an agenda of abundance that puts the interests of the 99% over the 1%.” Unsanctionably, my ears pricked up at the language here. What does an agenda of abundance mean to Zoran Mamdani? As someone who is very passionate about public goods, about public service, I think that we on the left have to be equally passionate about public excellence. One of the most compelling things that I think abundance has brought into the larger conversation is how we can make government more effective, how we can actually deliver on the very ideas that we are so passionate about, and a recognition of the fact that any example of public inefficiency is an opportunity for the argument to be made against the very existence of the public sector. To truly make the case time and time again that local government has a role in providing that which is necessary to live a dignified life, you have to ensure that every example of government’s attempt to do so is one that is actually successful. I think that’s what speaks to me about abundance, and I think that’s the line in the speech that speaks of both who we’re fighting for, but also the fact that we’re delivering on that fight, and it’s one that is actually experienced each and every day by New Yorkers across the five boroughs. I think it’s really important that if we’re going to ask voters to give us the power to add new government functions, we have to prove that government can function in the first place. Exactly. And recording this book with Ezra was really interesting, and I’ll be honest, it’s sometimes almost blackpilling. Like, it did not make me a libertarian. I’m still a proud liberal. I believe in an aggressive, muscular government. But it was astonishing sometimes to see that these examples of government failure were not exceptions to a rule, but in some cases, something tragically close to the rule itself. And I wonder if you’ve had a similar experience of coming into government, coming closer to government functions and recognizing that some things just don’t work the way they should. Can you put some meat on those bones of what you’ve seen that’s made you so passionate about not just announcing new public functions, but proving public excellence? You know, I very much agree. And I think what’s been frustrating is that even the very language of this conversation—language around bureaucracy, efficiency, waste, even quality of life—we’ve allowed this language to be seen as if it is of a right-wing concern, when, in fact, this should be the most paramount left-wing concern, because it is either the fulfillment or the betrayal of that which motivates so much of our politics. One of my focuses in my time in the New York State Assembly has been on the MTA, in part because that is the form of government that most New Yorkers interact with most frequently on a daily basis. When that train is late, when that bus is shown on your app to arrive in 10 minutes but never actually comes, when you are stuck in a tunnel, it diminishes your faith in local government at large. And I think this cannot be separated from the larger problem of politics in our city and in our state, which is that it has become more about incumbency protection than it has been about innovation and competence. To be in a mayor’s race where my chief opponent is Andrew Cuomo, someone for whom his supposed strength is that of managerial competence, it is in stark contrast to his actual record in running so many parts of our state government that continue to be just as ineffective as they were when he became governor, if not more. And the MTA is a chief example of that, where for years he refused to even acknowledge that it was a state entity under his responsibility. And then it went to the point that he was so eager to chase the headlines of saving money on what was being spent on the MTA that he implemented a restructuring program within the authority that allowed for the elimination of positions as soon as someone would retire without any actual comprehensive plan for ensuring that we were retaining the capacity within that authority to deliver on so much of what it was mandated to do. This then leaves us in a situation like the Second Avenue subway, where in the first phase, we’re spending more on consultants than we are on construction because we’ve lost so much of that internal capacity. We didn’t replace the people who retired or were fired for whatever reason, because we were working backwards from simply a headline of saying “X amount of positions have been eliminated, Y amount of money has been saved.” I want to get to housing, but let’s hold on the subway for a second. I’m really curious to know what you would do to bring down costs for the most expensive subway construction project in the world, New York City subways. It sounds very good, I think, to say that in the early stages of the Second Avenue subway construction, we were paying a lot of money to consultants and outsiders. But I’m also convinced that public sector unions in New York have rules that raise costs far above what they are in other countries, even countries, by the way, that are highly unionized. I am certainly not against public sector unions in the big picture. But when I look to why New York City builds the most expensive subways in the world, one answer for me is that public sector unions have contracts and contract demands that are raising the cost of per mile construction in New York City. Is talking to public sector unions about building transit cheaper something that you plan to do as mayor? I think I will have to work with public sector unions. I think I come to a different conclusion than you based on, however, the same point that you’re making. For example, if we name one of those counterparts, which is Paris, where they have arguably even stronger unions, yet the cost per mile is so much less. To me, it can’t be that the conclusion then can’t be that it is the presence of those unions’ work rules or labor rules in general. To me, the thing that has really stood out about how we drive down costs is the importance of something known as utility relocation. It’s not the only answer to this, but I think it is one part of it. I share this as an example. It’s a piece of legislation that I introduced in the State Assembly that would require public service corporations like Con Edison, National Grid, Verizon, and other telecommunications companies to perform the work necessary to support the MTA’s infrastructure improvements on a reasonable schedule and in accordance with actual needs. What I mean by this is that so often when the MTA does any significant construction, these other entities, public service corporations, typically look at that as an opportunity to get gold standard work done for whatever it is that they require. Even the city sometimes does this. What this all leads to is a ballooning of the actual cost of that infrastructure because it’s being used as this one moment where the city gets a park and Verizon gets better underground cable networking. It continues on and on and on. Then the actual price tag is associated solely with the cost of what was supposed to just be the MTA’s infrastructure as opposed to the truth that everyone is using this as a moment where they can finally spend as much money as possible to no cost to themselves. A top issue of your campaign, which I especially appreciate having lived in New York for seven years, is housing costs. The top policy on your website is freezing the rent. You describe it this way on your campaign website: “A majority of New Yorkers are tenants and more than 2 million of them live in rent-stabilized apartments. As mayor, I will immediately freeze the rent for all stabilized tenants and use every available resource to build the housing New Yorkers need and bring down the rent.” There is a tension in those two sentences that jumps out at me. You want the city to cap prices for a good whose supply you’re trying to increase, and I think that’s a very hard thing to do. If, for example, you wanted more great grocery stores, but you declared that no grocery store could charge more than $50 per receipt, the stores would just stop stocking the shelves the same way they used to. The stores would just get worse for everyone. How do you plan to both freeze rents and build the high-quality housing that New Yorkers need? You know, I see it from the perspective of landlords of those rent-stabilized units have seen an increase in 12% in their profits in the last year. These are profits with regards to tenants whose median household income is $60,000 a year. Now, what we’ve seen with the mayor, the current one, is that he’s increased the rent by 9% and he’s proposed increases on top of that by about 8%. The difficulty here is that you have an economic policy that the mayor effectively controls that determines whether or not New Yorkers, in large part, can continue to afford to live in this city. And the rent, while it’s often spoken about as if it is the only means by which these landlords are able to receive a profit, we know that there are actually other aspects of that. Some might call them loopholes, one of them being the individual apartment improvement. I bring this up because IAIs, as they’re known in an abbreviated form, something that I was in opposition to, but still passed in the state legislature, were doubled, whereby landlords could receive money from tenants for individual apartment improvements that were being made. And the reason I was in opposition to this was because this is a program that has been found to have a number of instances of fraud. I say all of this to say that you can freeze the rent. And what that ensures is that these tenants will not actually be priced out when we know that there’s this crisis of an exodus of working and middle-class New Yorkers. And there continue to be so many other aspects and ways in which landlords can continue to extract profit from these very tenants. But beyond that, to your larger question of how do we both freeze the rent and ensure that we’re building more, what I’ve heard from a lot of developers is one of the ways in which we are driving up costs in New York City is not even the dollar cost, but actually the time cost, which is then obviously translating into dollars. And that time and the delay of that time is in part because of the processes by which we approach land use. And this is also where abundance speaks to me in thinking about this both with regards to small businesses. If you look at an example in Pennsylvania, where they took a, what was it, an eight-week permitting time and cut it down to just a few days. But also in terms of housing, where we currently have a piecemeal approach where each city council member gets to determine whether or not a land use project moves forward by virtue of something known as member deference. What we need in order to actually build enough supply for the city that we have and to get past this staggeringly low vacancy rate is a comprehensive city-wide approach, one that can fast-track projects, especially those that are in line with the administration’s goals. I say that because there are a number of projects which you won’t actually find that much disagreement on. For example: Low-income housing for seniors that are still not built many years later. Ultimately, that’s a failure of just how slow this process goes. But the other point is that when we’re talking about a rent freeze, you still are seeing that, you know, this is on the backs of the findings that these landlords have a 12% profit in the last year. So that’s where I believe that there is room for that relief and that there’s still an incentive to construct more housing because we’re talking about both an incredible amount of demand and a limitation on profit as opposed to the elimination of profit. That is all specific to rent-stabilized housing, where in fact, there’s a lot of construction that is not rent-stabilized in this moment. You mentioned that you’re looking at Pennsylvania, what Governor Shapiro is doing there. And I like the fact that you’re looking around the country and thinking about borrowing certain ideas from other governors and mayors. I wonder if you looked across the Hudson River to see what they’re doing in Jersey City, where Mayor Stephen Fulop has really succeeded in building a ton of housing. Last decade, rents in Jersey City were absolutely skyrocketing, but the city changed its permitting laws. It welcomed new development, supply boomed. Literally just days ago, the mayor announced that rents are actually declining in Jersey City. I mean, for renters themselves, for tenants, that’s even better than frozen, right? Actually, rents that are declining. And it’s because of this boom in supply. Is this a story that you’re following? Do you think there’s a lesson to take from Jersey City? It is absolutely a story of interest to me. And I’ve thought of it often, just even in the statistical analysis of the fact that in New York City, we’re building around four homes per thousand people, while in Jersey City, it’s about seven. And in Tokyo, it’s about ten. And this is, you know, I clearly have ideas and politics, but ultimately beyond all of those things, I care most about outcomes. What I’m very passionate about is making this a more affordable city, also making it a more efficient city. I think for too long in our politics, especially in the vein of what we were talking about earlier, this incumbency protection, we have allowed for this reverse exceptionalism to flourish in New York City, where we see examples of things that have been successful elsewhere in the country or elsewhere in the world. And we simply tell ourselves it could never happen here in New York because we’re different. In fact, we should be proud of the fact that those things are not present here. I think we’ve seen this when we’re talking about bike infrastructure or outdoor dining, but also especially around housing. I believe we have to bring that kind of politics to an end. We have to have both a pride in our city, which I do have, and which so many New Yorkers share, and a humility that we can still learn from other places. There is much to learn with regards to Jersey City, and also much to learn with regards to other global cities across the world, in understanding what regulations we have that truly still stand up to the question of which have actually lost their justification. When we talk about the need to build more housing, I also think we need to build different types of housing. It is a shame that we’ve made it functionally illegal to build SROs in New York City. For example, we have to have conversations around even the minutiae of, you know, single stairwell versus dual stairwell. These are examples that have stuck with me because they show that when it’s not innovation and competence that is driving the creation of these regulations, then you start to have more and more things that actually drive up the cost for a justification that you can’t even remember by the end of it. I’ve seen some folks associate your campaign with Brandon Johnson, the left-wing mayor of Chicago, which, of course, is not a comparison that I think a typical person would welcome, even if Mayor Johnson’s a nice guy. He’s currently polling around 10 or 12 percent. I want to ask you about one aspect of Johnson’s tenure, which is the bureaucracy of building affordable housing. In Chicago’s affordable housing qualified allocation plans, 10 percent of points are rewarded for developers who show that they have extra accessibility features in their plans. 15 percent of points are awarded based on the makeup of the development team, so extra points if developers are BIPOC or women. Only three percent of scorecard points are awarded based on project cost. What that basically means is that affordable housing in Chicago treats accessibility and workforce diversity as nine times more important than cost savings. It seems to me like in Chicago, you get exactly what you plan for. The city recently built affordable housing at a reported $1.1 million a pop, according to the mayor’s Twitter feed. What do you think Chicago did wrong? I think one of the most important things is that we actually set a goal for what each unit should cost and work backwards from that, as opposed to ending up with a figure after having made all of the different criteria that we would like to fulfill. Because if we cannot construct it at a cost that is scalable, then that is the greatest failure of the idea itself. These ideas are only worth as much as their implementation. It’s interesting because I know that the critique you’re laying out here is specific to Chicago. However, one of the critiques I’ve had, for example, of the previous tax subsidy scheme in New York called 421A was that the cost per unit for quote-unquote affordable housing was similarly above a million dollars. I think there is just a lack of efficacy in the manner in which we’ve approached this construction. What makes our campaign distinct is that we want the private market to play a significant role in the creation of new housing supply. We believe the public sector should construct what is immediately affordable, as opposed to just constructing more supply in the hopes that it eventually drives down the cost of housing overall. We need to provide something that meets the median household income of $70,000 for a family of four from day one. But for all of this, it still comes back to that central point: what is the cost that we can actually do this at scale? And how do we ensure that we hit that cost? Yes, we’re talking about housing right now, but it’s the same issue that… speaks to the MTA. It’s the same issue that speaks to public bathrooms. We need to ensure that we are working backwards from a scalability and outcomes as opposed to having that simply be a byproduct of all the other decisions we make. This episode is brought to you by the Home Depot. Planning a few summer projects? Upgrade your toolbox with 4th of July savings. Wow. We’re already here with 4th of July savings. This is great. On select top brand cordless power from the Home Depot. Whether you’re working on a fence, a planar box, or a new workbench, you’ll get power and convenience with the Ryobi One Plus 18 Volt Two Tool Kit from the Home Depot. Now, at a lower price of $99 was $139. That’s why you can get 4th of July savings on the cordless power you need to make summer projects easier than ever. Right now, at the Home Depot. Real businesses like Malibu Apothecary rely on Spectrum to keep them connected. We’re a fragrance company, but we also have a lot of technology that we use. Spectrum is really the backbone of any of the business we do here. Having a reliable, fast internet service that can fit my budget is really great. I traded a 9 to 5 for a 5 to 9. It’s a lot of work, but I love it. Spectrum Business Internet starts at $40 a month when bundled. Learn more at spectrum.com/business. Restrictions apply. Service is not available in all areas. Dear Traction, Toyota has 25 vehicles with available all-wheel drive and four-wheel drive. The RAV4 has legendary versatility. The Grand Highlander’s adult-sized third row is for real. And the Tacoma’s wild side comes standard. From job sites to campsites and game nights, there’s an all-wheel drive or four-wheel drive Toyota to get you there in style. Find yours at toyota.com. Toyota. Let’s go places. On rent, you have this plan to mandate a price freeze. With groceries, you also have a plan to use government power to control costs. You have a plan to build a small fleet of public grocery stores. What is the problem here that you think city-run grocery stores specifically can solve? There are two problems. The first is a problem of affordability. There is a sticker shock that New Yorkers tell me about all the time, whether they’re making $40k a year or $200k a year, that when they go into the supermarket and they look at the same items that were easy for them to purchase a few years ago, they now seem more and more out of reach. The most obvious examples here are eggs, milk, and bread that have been cited again and again. There is a second crisis, which is that of food deserts, where black and brown New Yorkers disproportionately deal with a situation where there simply isn’t affordable produce, and sometimes even produce at all, within close proximity to where they live. I represent Queensbridge Houses, the largest public housing development in North America, and I’ve heard from constituents time and again that there simply isn’t high-quality produce within a five block, 10 block radius, but I can find five, six different fast food restaurants in that same space. And that’s where my proposal is one; this is a proposal of reasonable policy experimentation where we’re talking about five stores, one store in each borough across the five boroughs in New York City. It is something that would cost $60 million in total, which is less than half of what the city is currently set to spend on subsidizing corporate supermarkets through a program called City Fresh. The reason I bring this up is that too often we don’t interrogate the ways in which city government is already intervening in this, but doing so in a manner that tries to absolve itself of responsibility and instead further invests in a model that is leaving us with the same results. This City Fresh program, which is set to cost about $140 million, is one that doesn’t require the supermarkets receiving the subsidy to accept SNAP or WIC. It doesn’t require them to engage in collective bargaining. It doesn’t even require them to have lower prices. It’s just about trying to assist in their continued operation. It’s funny to even hear some of the critique, especially from John Katsimatidis, the owner of Gristidis, who completely misses the fact that for many New Yorkers, they can’t even afford going into those kinds of stores today. I think that the scale of this pilot program is one where we could meet these twin crises at the same moment by ensuring that where we’re selecting this location is meeting both of these needs. Also, to showcase the ability to prove out that argument that I’m having with you, not an argument with you, but the conversation with you around how do we prove the effectiveness? Because if it is not effective at a pilot level, it does not deserve to be scaled up. But I believe it can be effective in two ways. One, because I think that there’s far… more efficiency to be had in our public sector. And I think there have been glimpses of that efficiency, most especially at the height of COVID. The speed with which city government set up testing sites, vaccine sites, and the ability to go through that vaccine site in about 15 minutes is in stark contrast to how so many experience the public sector. Because food is a non-negotiable, it’s not a luxury item. I’m talking to you as a state legislator who watched our state spend hundreds of millions of dollars in cutting the state’s gas tax to subsidize the cost of gas at a time when those prices were going high, but considers it a bridge too far to do anything with regards to the price of groceries. It’s interesting hearing you talk about wanting to offer more government services while also being fixated on this North Star of efficiency. Do you know about the sewer socialists of Milwaukee? Is this something you’re familiar with? For listeners who might not be, I assume that maybe you could as well have a tattoo of sewer socialism on your back. In the early 20th century, there was a group of Milwaukee left-wing city leaders who embraced what was originally, I think, a slur of sewer socialism. It was this sort of self-aware reference to the idea that they often bragged about how great their public sewer system was. It was a form of socialism that was laser-focused on making the city work better. It wasn’t focused on necessarily ushering in the global defeat of capitalism; it was focused on local, tangible, concrete, measurable issues like: Is the sewer safe and working? It’s funny because more than a few people on book tour asked if abundance was meant to revitalize sewer socialism in the 21st century. My answer was always, “well, look, I am not a socialist, but if I were a socialist, I would definitely be a sewer socialist.” To me, the idea of “you gave us power and look at the good that we did” strikes me as so much more attractive than “you gave us power, we struggled to do good, things are still horrible, add government functions.” Tell me a little bit about how you define your own democratic socialism in this light, in the spirit of the Milwaukee sewer socialists. Claiming the language of quality of life as a left-wing concern is important because it is often described as if it is somehow conservative. However, if we want to fight for the dignity of each and every person, especially ensuring that applies to the very working-class New Yorkers that are often forgotten, at the core of it is also the quality of their life—that it is an excellent life that they are living. Much of that comes back to the efficacy of the public services that they engage with. Too often, we’ve refused to even admit to inefficiencies or critiques, or waste within the public sector, thinking that by doing so, we open the critique from the right. But in actuality, our refusal to admit it is even more ammunition for the right, because it showcases that ultimately, their argument would be that we don’t believe in the efficacy of these things simply in their existence. For myself, and I know many feel the same, the existence is important because of the efficacy that it opens the door to. I think what sewer socialism, and the example of Milwaukee, shows is that we want to showcase these ideals not by lecturing people about how correct they are, but rather by delivering on them and letting that delivery be the argument itself. There are just far too few examples in New York City politics of any large-scale interventions of city government, and any ones that have been successful. The few that come to mind right now are: Congestion pricing Universal pre-K These are examples of interventions that fundamentally transformed life in our city and should be used as a model for what more we can do. Yet both of them require a level of political courage and a willingness to have the kind of battles that so often are the very obstacles that stop us from imagining a better future than the one that we have. To have universal pre-K, you have to have a mayor who is willing to say that they’re going to go to Albany, they’re going to fight to tax the wealthiest New Yorkers to fund an expansion of the social safety net to apply to universal pre-K. To have congestion pricing, you have to be willing to take on the political culture of this city that for so long has been the exact impediment we’ve seen at community board after community board about the way in which we can reimagine the streetscape from the one that we’ve had to the one that we actually need. Those are some of the ideas that animate a lot of what I’m running on. And my vision for this city is how do we refuse to be content with what we have and actually deliver on the things that we deserve and use the examples of elsewhere across the country and the world as what drives us. For me, one of those examples is also what’s happened in Paris with Hidalgo and the ability to plant more than 100,000 trees and have a significant impact on the air quality of that city in tandem with the transformation of that streetscape. This could be a model for what we do in New York City as well. Last question. And I’ve really enjoyed talking to you. I feel like there are a lot of things we agree about: congestion pricing trees government efficiency This idea that the liberals should not be afraid of pointing the finger at their own side just because they think Republicans might say, “oh, look, liberals are criticizing each other. Therefore, there are critiques to be made.” It’s really important for us to understand how to wield power effectively if we’re going to ask voters to give us more of it. We disagree about some things. We disagree, I think, about rent freezes. I like that you’re into policy experimentation, that you seem to think about the world through a lens of trade-offs. And finally, it seems to me that you change your mind on some things. You told the New York Times that you would change your mind a bit on the value of private development. Also, there’s this case of defund the police. In 2020, you and a couple of other mayoral candidates embraced the rhetoric of defund the police. But on a debate stage this year, you said, and this is a quote, “I will not defund the police. I believe the police have a critical role to play. We need to ensure that police can focus on crimes, end quote.” In several interviews, you’ve talked about wanting to hire more social workers to do social work so that we can have police officers doing police work, stopping crime and solving murders. This is not a gotcha question at all. I don’t think you’re a hypocrite. I think you change your mind. I want to know how you change your mind. Who do you talk to? What do you read? What do you see that has shifted your position on police or housing from one year to the next? You know, you have to continue to fight the instinct in politics to surround yourself with people who are quickest to get to a yes, or who are quickest to replicate the very idea that you’ve proposed to them. I’ve been lucky that in my time in the state assembly, there are a number of my colleagues with whom I have significant disagreements but whom I’ve also developed relationships with such that I can continue to have this conversation and always approach it with some level of humility that I could have been wrong at some time. I think that’s a part of politics that we’re also missing— the idea that leadership can also be someone who recognizes what they know and what they don’t know and surrounds themselves with people who challenge them. Those people’s commitment is not of a shared ideological approach to the world, but rather a shared track record of excellence and fluency. The reason that I’ve come to different conclusions today than I’ve had in the past, whether we’re speaking about policing or housing, is not just a function of the changing way in which I see the world, but also a result of the conversations with colleagues and a number of friends who have always been generous with their time in interrogating the concepts that are animating our politics at any different year. I think that’s also the way that I would approach running the city— to be wedded to outcomes, not wedded to the means by which we get to those outcomes. I think that allows you the ability to learn and to grow. Leadership can look a lot more like that than someone who pretends they know all the answers all the time and regrets is something that is beyond them. Zoran Mamdani, thank you very, very much. Thank you. This has been such a pleasure. Thank you so much, Jared. Thank you. window.tocIndex = { "index": [ { "index_sentences": "Folks, it's J. Kyle Mann from The Ringer, and as always, basketball is so freaking good.", "section_level": 1, "section_title": "Intro: The Ringer NBA Draft Show" }, { "index_sentences": "This episode is brought to you by Zendesk, introducing the next generation of AI agents built to deliver resolutions for everyone.", "section_level": 1, "section_title": "Sponsor: Zendesk" }, { "index_sentences": "Before today's show, a personal announcement. After almost 17 years at The Atlantic Magazine, I have just officially moved my writing full-time to Substack, the newsletter platform.", "section_level": 1, "section_title": "Personal Announcement: Moving to Substack" }, { "index_sentences": "Today, Abundance, the American left, and Zoran Mamdani, the Democratic Socialist candidate for mayor of New York City.", "section_level": 1, "section_title": "Topic Introduction: \"Abundance\" Book Reception" }, { "index_sentences": "But if you pull your mind away from social media and YouTube and this online discourse, if you listen to and watch what politicians and people running for political office have been saying about this book and our project, I think what you'll see is that the response has been notably different.", "section_level": 2, "section_title": "Contrasting Online and Political Response to \"Abundance\"" }, { "index_sentences": "For example, Bernie Sanders' devotees have repeatedly bashed Abundance. But Representative Ro Khanna, an outspoken advocate of Bernie's signature policy proposal, Medicare for All, announced his support for Abundance on several occasions.", "section_level": 3, "section_title": "Political Examples Supporting \"Abundance\"" }, { "index_sentences": "I wanted to talk to the man himself. So I was very gratified when, in a very busy week for me as I've just switched jobs and am now moving cities back to Washington, D.C., and in a very, very busy week for him, since he's running to become mayor of America's biggest and richest city.", "section_level": 1, "section_title": "Setting Up the Interview with Zoran Mamdani" }, { "index_sentences": "Mamdani and I nonetheless found 30 minutes to sit down and talk calmly last Saturday about abundance and the left, how we agree, where we disagree, why government deficiency ought to be a virtue of all leaders, but especially those on the left who want government to do much more.", "section_level": 2, "section_title": "Key Themes: Government Efficiency and Changing Minds" }, { "index_sentences": "I'm Derek Thompson. This is Plain English. Zoran Mamdani, great to talk to you.", "section_level": 1, "section_title": "Interview Begins: Derek Thompson with Zoran Mamdani" }, { "index_sentences": "You recently delivered a speech in which you said this, \"government must deliver an agenda of abundance that puts the interests of the 99% over the 1%.\"", "section_level": 2, "section_title": "Mamdani's \"Agenda of Abundance\": Public Excellence" }, { "index_sentences": "I think it's really important that if we're going to ask voters to give us the power to add new government functions, we have to prove that government can function in the first place.", "section_level": 2, "section_title": "The Problem of Government Inefficiency: MTA Example" }, { "index_sentences": "I want to get to housing, but let's hold on the subway for a second.", "section_level": 3, "section_title": "MTA Construction Costs and Unions" }, { "index_sentences": "I think I will have to work with public sector unions. I think I come to a different conclusion than you based on, however, the same point that you're making.", "section_level": 3, "section_title": "Mamdani's Proposed Solution: Utility Relocation" }, { "index_sentences": "A top issue of your campaign, which I especially appreciate having lived in New York for seven years, is housing costs.", "section_level": 2, "section_title": "Housing Costs and Rent Freeze Proposal" }, { "index_sentences": "You know, I see it from the perspective of landlords of those rent-stabilized units have seen an increase in 12% in their profits in the last year.", "section_level": 3, "section_title": "Mamdani's Defense of Rent Freeze" }, { "index_sentences": "But beyond that, to your larger question of how do we both freeze the rent and ensure that we're building more, what I've heard from a lot of developers is one of the ways in which we are driving up costs in New York City is not even the dollar cost, but actually the time cost, which is then obviously translating into dollars.", "section_level": 3, "section_title": "Mamdani's Plan for Increasing Housing Supply" }, { "index_sentences": "You mentioned that you're looking at Pennsylvania, what Governor Shapiro is doing there.", "section_level": 3, "section_title": "Learning from Jersey City's Housing Success" }, { "index_sentences": "It is absolutely a story of interest to me. And I've thought of it often, just even in the statistical analysis of the fact that in New York City, we're building around four homes per thousand people, while in Jersey City, it's about seven.", "section_level": 3, "section_title": "Ending \"Reverse Exceptionalism\" and Embracing Learning" }, { "index_sentences": "I've seen some folks associate your campaign with Brandon Johnson, the left-wing mayor of Chicago, which, of course, is not a comparison that I think a typical person would welcome, even if Mayor Johnson's a nice guy.", "section_level": 2, "section_title": "Comparison to Chicago: Affordable Housing Costs" }, { "index_sentences": "I think one of the most important things is that we actually set a goal for what each unit should cost and work backwards from that, as opposed to ending up with a figure after having made all of the different criteria that we would like to fulfill.", "section_level": 3, "section_title": "Mamdani's Solution: Setting Cost Goals for Affordable Housing" }, { "index_sentences": "This episode is brought to you by the Home Depot.", "section_level": 1, "section_title": "Sponsor: Home Depot" }, { "index_sentences": "Real businesses like Malibu Apothecary rely on Spectrum to keep them connected.", "section_level": 1, "section_title": "Sponsor: Spectrum Business" }, { "index_sentences": "Dear Traction, Toyota has 25 vehicles with available all-wheel drive and four-wheel drive.", "section_level": 1, "section_title": "Sponsor: Toyota" }, { "index_sentences": "On rent, you have this plan to mandate a price freeze.", "section_level": 2, "section_title": "Grocery Policy: Public Grocery Stores" }, { "index_sentences": "There are two problems. The first is a problem of affordability.", "section_level": 3, "section_title": "Problems Addressed by Public Grocery Stores: Affordability and Food Deserts" }, { "index_sentences": "But I believe it can be effective in two ways. One, because I think that there's far... more efficiency to be had in our public sector.", "section_level": 3, "section_title": "Demonstrating Public Sector Efficiency" }, { "index_sentences": "It's interesting hearing you talk about wanting to offer more government services while also being fixated on this North Star of efficiency.", "section_level": 2, "section_title": "Defining Democratic Socialism: \"Sewer Socialism\" Spirit" }, { "index_sentences": "Claiming the language of quality of life as a left-wing concern is important because it is often described as if it is somehow conservative.", "section_level": 3, "section_title": "Importance of Delivery and Outcomes: NYC and Global Examples" }, { "index_sentences": "Last question. And I've really enjoyed talking to you. I feel like there are a lot of things we agree about: congestion pricing, trees, government efficiency.", "section_level": 2, "section_title": "The Process of Changing One's Mind" }, { "index_sentences": "You know, you have to continue to fight the instinct in politics to surround yourself with people who are quickest to get to a yes, or who are quickest to replicate the very idea that you've proposed to them.", "section_level": 3, "section_title": "Mamdani's Concluding Thoughts: Leadership and Learning" }, { "index_sentences": "Zoran Mamdani, thank you very, very much.", "section_level": 1, "section_title": "Outro" } ] }; window.faq = { "qas": [ { "answer": "The text states that the online discourse surrounding the book \"Abundance\" has been quite negative on platforms like Twitter, Blue Sky, Reddit, YouTube podcasts, and left-wing sites and magazines, describing it as a \"surprising wave, a large wave of negative feedback.\" The specific reasons for this widespread negative feedback are not detailed.", "index_of_source": "The online discourse has been quite negative.", "question": "According to the text, why has Derek Thompson's book \"Abundance\" received significant negative feedback from the online political left?" }, { "answer": "Despite having very different politics on issues like housing, affordability, education, levels of taxation, and spending, Zoran Mamdani has explicitly embraced an \"agenda of abundance.\" He believes leftist critics have oversimplified the book and that its approach to making government work better is exactly what the left needs at this moment.", "index_of_source": "Mamdani and I have very different politics on a range of issues, including housing, affordability, education, levels of taxation, and spending.", "question": "How is Zoran Mamdani's endorsement of an \"agenda of abundance\" counterintuitive given his described political differences with the author?" }, { "answer": "For Zoran Mamdani, an \"agenda of abundance\" means being equally passionate about \"public excellence\" as about public goods and services. It involves making government more effective and delivering on the ideas the left is passionate about, recognizing that public inefficiency weakens the argument for the public sector's existence and that successful government actions build faith.", "index_of_source": "As someone who is very passionate about public goods, about public service, I think that we on the left have to be equally passionate about public excellence.", "question": "What does Zoran Mamdani mean by advocating for an \"agenda of abundance\" in the context of government and public service?" }, { "answer": "Based on his time in the New York State Assembly, Mamdani cites the MTA as a key example, pointing to issues like late trains and buses that diminish public faith. He also highlights the high cost of the Second Avenue subway construction, where funds are disproportionately spent on consultants over construction due to a loss of internal capacity after eliminating positions without a comprehensive plan.", "index_of_source": "And the MTA is a chief example of that, where for years he refused to even acknowledge that it was a state entity under his responsibility.", "question": "What specific instances of government inefficiency does Zoran Mamdani point to from his experience in the New York State Assembly?" }, { "answer": "Mamdani argues that freezing rent on stabilized units prevents tenants from being priced out while landlords still maintain profitability (citing a 12% profit increase in the last year) and can utilize other mechanisms like Individual Apartment Improvements (IAIs). To increase supply, he proposes streamlining the land use process and implementing a comprehensive city-wide approach to fast-track projects, stating that reducing the \"time cost\" of development will help lower overall expenses and incentivize construction due to high demand.", "index_of_source": "You know, I see it from the perspective of landlords of those rent-stabilized units have seen an increase in 12% in their profits in the last year.", "question": "How does Zoran Mamdani propose to manage the potential tension between freezing rents on rent-stabilized apartments and simultaneously increasing the supply of high-quality housing?" }, { "answer": "Mamdani proposes building a small fleet of five public grocery stores (one in each borough) to address two crises: affordability (high prices for essentials like eggs, milk, and bread) and food deserts (lack of affordable, high-quality produce in disproportionately Black and Brown neighborhoods). He sees this pilot program as a more effective use of funds compared to existing subsidies like the City Fresh program.", "index_of_source": "There are two problems.", "question": "What are the primary problems Zoran Mamdani believes a city-run fleet of public grocery stores would solve?" }, { "answer": "Mamdani aligns his view of \"democratic socialism\" with the spirit of \"sewer socialism\" by emphasizing the importance of reclaiming \"quality of life\" as a left-wing concern and focusing on delivering excellent public services. He believes admitting and addressing government inefficiencies is crucial to proving efficacy and demonstrating the value of the public sector through tangible, successful outcomes, rather than just advocating for its existence.", "index_of_source": "Claiming the language of quality of life as a left-wing concern is important because it is often described as if it is somehow conservative.", "question": "How does Zoran Mamdani relate his definition of \"democratic socialism\" to the historical concept of \"sewer socialism\"?" }, { "answer": "Mamdani explains that he consciously fights the political instinct to surround himself only with yes-men. He actively cultivates relationships with colleagues and friends who hold significant disagreements and challenge his views. He approaches these conversations with humility, believing that leadership involves recognizing one's limitations and seeking challenge from those with proven expertise. He is \"wedded to outcomes, not wedded to the means,\" which allows him to learn and change his mind based on evolving perspectives and challenging dialogue.", "index_of_source": "You know, you have to continue to fight the instinct in politics to surround yourself with people who are quickest to get to a yes, or who are quickest to replicate the very idea that you've proposed to them.", "question": "According to Zoran Mamdani, what process does he undertake to change his mind on political issues like policing or housing?" }, { "answer": "Mamdani finds Jersey City's success in building housing and lowering rents a story of interest, noting their significantly higher rate of homes built per capita compared to New York City. He sees it as an example from which New York can learn, challenging the city's tendency towards \"reverse exceptionalism\" where successful models from elsewhere are dismissed. He believes humility is needed alongside pride in NYC to learn from places like Jersey City about how to streamline regulations and increase supply.", "index_of_source": "It is absolutely a story of interest to me.", "question": "What does Zoran Mamdani learn or take away from Jersey City's experience in boosting housing supply and affecting rent prices?" } ] };

2025/6/23
articleCard.readMore