Anthropic Moves Trigger Software Stock Slump as AI Agent Disruption Fears Grow

AI 智能体颠覆担忧加剧，Anthropic 动作引发软件股大跌 由于快速发展的 AI 自动化给以往稳定的经常性收入模式带来了前所未有的不确定性，投资者正在压缩软件行业的估值倍数。 主持人分析了由 Anthropic 的 AI 产品发布及广为流传的宏观经济悲观报告引发的近期股市恐慌，认为 AI 智能体带来的生产力提升最终将扩大企业需求，而非导致经济崩溃。他们讨论了数据中心扩张面临的物理限制与地方阻力，以及特朗普总统的《纳税人保护承诺》（Ratepayer Protection Pledge）。此外，本期节目还探讨了外界对长达 108 分钟国情咨文演讲的反应、一项获 FDA 批准使用山中因子逆转失明的人体试验，以及最高法院以 6 比 3 的裁决判定行政紧急关税无效的标志性判决。 Anthropic 在 2 月接连发布多款产品并对二级市场造成冲击，其中 Claude 支持 COBOL 数据库现代化的更新直接导致 IBM 单日暴跌 13%，市值蒸发 310 亿美元。 (1:22) 公开市场对软件企业的估值逻辑已从预测现金流何时收缩（When）转向担忧其模式能否存续（If），叠加 Citrini 唱空报告的发酵，促使 SaaS 板块面临市盈率倍数压缩与贴现率攀升。 (2:47) 尽管市场担忧程序员被取代，但基于杰文斯悖论（Jevons Paradox），软件开发成本大幅下降反而激增了定制化软件需求；全球开发职位发布量同比上涨约 10%，Anthropic 更是以高达 57 万美元薪资争夺顶尖工程人才。 (16:50) AI 数据中心扩张正受到电力、土地瓶颈及日益严峻的邻避效应阻碍，预计两年内导致约 1300 亿美元潜在营收受损，促使特朗普在国情咨文中提出“纳税人保护承诺”以消除居民对电费上涨的抗议。 (35:42)

The All-In Podcast hosts argue that rapid advancements in generative AI are destabilizing legacy software valuations and traditional pricing models, even as physical constraints, municipal NIMBYism, and legal battles impede the speed of actual economic disruption. AI Disruption and Software Valuation Compression (1:22) Anthropic product rollouts in February sparked immediate equity sell-offs across multiple sectors. A legal plugin announcement triggered a 10% drop in legal and business intelligence stocks like Thomson Reuters and LegalZoom on February 3, Claude Code Security prompted sell-offs in CrowdStrike and Okta on February 20, and a COBOL database modernization update on February 23 triggered a 13% single-day drop in IBM, wiping out $31 billion in market capitalization. Chamath Palihapitiya explains that the public markets have transitioned from pricing 'when' cash flows will decline to questioning 'if' legacy business models will survive, prompting P/E multiple contraction and pushing discount rates up to 12–13%. Simultaneously, a viral Substack by Citrini Research outlining a catastrophic 2028 economic collapse contributed to broader equity sell-offs. David Sacks and David Friedberg express skepticism toward such doomer narratives, noting that the Citrini report was co-authored by a partner at a short hedge fund. Sacks highlights that while prediction markets like Kalshi assigned only a 12% probability to this collapse scenario, high market uncertainty has still driven investors out of software equities and into 5% treasury yields. Jevons Paradox and Internal Agent Automation (20:06) Countering fears of mass white-collar displacement, Sacks cites Box CEO Aaron Levie regarding Jevons paradox: when the cost of a supply-constrained input drops drastically, total demand surges. Sacks points to Citadel Securities data indicating global software engineering postings rose approximately 10% year-over-year, while Anthropic continues hiring software engineers with base pay reaching $570,000. Sacks and Palihapitiya argue that cheaper software production will expand technology spending from 1–2% of corporate cost structures up toward 50%, even as overall operational expenditures decline. Jason Calacanis illustrates this dynamic through internal operations at Launch, where 15 non-technical employees were trained to build custom OpenClaw agents. Within 30 days, staff automated sales intelligence, executive reporting, and media clipping without eliminating jobs, while increasing knowledge worker output by 10% to 20% weekly. The hosts observe that the ability to build internal agent tooling shifts pricing leverage away from traditional SaaS providers during renewal cycles. Physical Infrastructure Bottlenecks and Local Grid Resistance (35:42) While token demand is projected to increase tenfold alongside an anticipated 90% decline in token costs by year-end, physical bottlenecks restrict real-world scaling. Palihapitiya highlights research showing 100 U.S. data center projects facing active local pushback, with 5 gigawatts of capacity cancelled in 2025 and 7 gigawatts at risk in 2026—representing an estimated $130 billion in lost revenue over two years based on OpenAI projections. Friedberg warns that protracted regulatory timelines, such as the 612 days spent exclusively on environmental reviews for Micron's New York fab, risk pushing capital and tech infrastructure overseas to regions like the Middle East. To address public backlash over energy grids, Sacks highlights the administration's proposed Ratepayer Protection Pledge, which aims to shield residential consumers from rising utility bills by encouraging hyperscalers to fund off-grid, behind-the-meter power generation. Executive Tariff Reversals and Deepening Political Divide (70:17) Federal economic policy met a major constitutional hurdle when the Supreme Court ruled 6–3 against President Donald Trump's emergency-power tariffs. The decision put approximately $175 billion in collected duties into legal dispute, prompting corporate importers to file for refunds. Sacks notes the White House responded by pivoting to Section 122 of the Trade Act of 1974 for a temporary 150-day, 15% tariff to buy time for formal investigations under Sections 301 and 338. While Palihapitiya contends tariffs expose fundamental trade imbalances and should be codified by Congress, Friedberg praises the ruling as a vital check on executive overreach. This institutional confrontation mirrored the hyper-partisanship observed during Trump's 108-minute State of the Union address, where visual polarization persisted despite isolated moments of bipartisan support for lowering drug costs and banning congressional stock trading. The hosts observe that despite intense partisan conflict and mutual threats of legal retaliation, neither party has demonstrated a willingness to curtail federal spending, leaving the national deficit projected to expand by another $2.5 trillion. The hosts conclude that while enterprise software faces a painful repricing cycle driven by autonomous agents, the speed of broader macroeconomic upheaval is constrained by physical energy infrastructure, local resistance, and systemic political friction.

本期节目深入探讨了生成式AI工具如何引发传统软件行业的估值逻辑重塑与商业模式危机，并展示了企业借助自主构建智能体实现降本增效的现实案例；同时，主持人们剖析了数据中心算力扩张遭遇的电力环保阻力，以及美国最高法院推翻总统紧急关税令对行政权力与全球贸易格局的深远震荡。 软件估值重塑与AI恐慌叙事争议 (1:22) 2026年2月，Anthropic连续发布法律插件、代码安全审查工具及COBOL现代化功能，直接引发公开市场连锁震荡。由于全球85%的COBOL代码运行于IBM硬件，IBM单日暴跌13%，市值蒸发310亿美元。Chamath Palihapitiya指出，市场对软件公司的估值逻辑发生了根本性转变：从前投资者主要测算企业现金流“何时”（When）下滑，而如今则质疑其业务模式“是否”（If）还能存续。这种黑天鹅担忧迫使市盈率和市销率倍数大幅压缩，企业的加权平均资本成本（WACC）也从6%飙升至12%-13%。 与此同时，Citrini Research一份预言AI导致大规模失业和消费断崖的看空报告在社交网络上引发恐慌，但David Sacks揭露该报告合著者实为持有空头仓位的对冲基金合伙人。Sacks援引Citadel Securities数据指出，全球软件工程师招聘需求同比实际增长约10%，且Anthropic自身仍在以高达57万美元的底薪竞聘顶级工程师。他引用杰文斯悖论（Jevons paradox）强调，当原本稀缺的软件研发成本大幅下降时，非科技行业被长期压抑的技术需求将被彻底激活，总体市场需求反而会迎来指数级膨胀。 智能体落地与企业软件支出变革 (22:19) 企业应用层面正在经历从采购现成SaaS向内部搭建定制AI智能体的转变。Jason Calacanis以其机构Launch为例，说明团队仅通过6至7小时的培训，便让15名非技术员工使用OpenClaw在30天内构建出涵盖播客广告线索挖掘、高管日程追踪、音视频自动剪辑和缩略图优化的多套内部工具。这类内部智能体极大地压低了对外部软件的依赖，甚至能在周末自行部署开源Mattermost服务器以替代高成本的Slack，倒逼传统SaaS软件在续约谈判中丧失溢价优势。 在底层技术与成本分配方面，开源模型（如Kimi 2.5）能够以极低的Token成本覆盖企业80%至85%的基础任务，导致Harvey等高价垂直封装应用面临严重同质化竞争。Chamath预测，成熟科技公司为应对长期转型，必须严厉削减作为股东权益稀释主要推手的股权激励支出（SBC）以储备现金；同时，随着Token需求预计激增10倍且成本面临90%的下调，企业已普遍将Token用量直接列入员工的综合用工预算中。 算力扩张的物理瓶颈与能源政治 (35:42) AI算力基建正面临日益严峻的物理实体瓶颈与地方“邻避效应”（NIMBY）。Chamath披露的数据显示，全美已有100个数据中心项目遭遇地方阻力，第二季度25个规划项目中即有20个被迫取消；2025年因项目取消损失的电力装机达5吉瓦，2026年更有7吉瓦面临停摆风险，按OpenAI每吉瓦对应约100亿美元营收推算，两年内潜在经济损失高达1300亿美元。相比中东地区通过精简环保审批和资本扶持加速建设，美国本土繁苛的行政诉讼正迫使算力资本流向海外。 针对居民对电费上涨的抗议，David Sacks阐释了特朗普提出的“纳税人保护承诺”（Ratepayer Protection Pledge），主张由科技巨头全额出资构建“表后”（Behind the Meter）自备供电系统，并通过回输盈余电量摊薄公共电网固定成本，避免普通家庭为AI发展承担通胀代价。然而Chamath补充提醒，由于受监管公用事业公司依资本支出规模享有法定回报率，即便没有数据中心介入，常规电网扩建仍可能持续推高终端零售电价。 国情咨文党派极化与关税司法博弈 (52:13) 特朗普长达108分钟的国情咨文暴露了国会内部不可弥合的党派鸿沟。除冰球队获胜与平抑处方药价等少数议题外，两党在边境管控与犯罪治理上针锋相对，仅在禁止国会议员炒股的议题上因特朗普对伊丽莎白·沃伦的即兴调侃形成了罕见共鸣；而在联邦预算赤字预计继续激增2.5万亿美元的背景下，两党在控制财政开支上均未展现实质作为。这一政治对立态势延伸至宏观经贸领域，最高法院以6比3裁定总统动用紧急权力征收关税违宪，成为自1935年以来对总统行政权力的最强司法审查。 该裁决令已征收的1750亿美元关税面临超2000家进口商的退款诉讼。白宫随即换道援引1974年《贸易法》第122条推行为期150天的15%临时关税，并借此窗口筹备依据第301条及第338条重建法定基础。主持人们对此分歧鲜明：Friedberg赞赏司法机构超脱党派意志捍卫了宪政制衡；Sacks认为政府在沉重赤字压力下绝不会轻易退还关税；Chamath坚称关税矫正了损害美国中产阶级的失衡外贸，呼吁国会正式立法固化；Calacanis则力劝华盛顿停止两党间的司法清算与行政越权，重归理性的温和治理通道。 山中因子人体试验与细胞逆龄再生 (63:51) 在生命科学前沿，David Sinclair联合创立的Life Biosciences与美国食品药品监督管理局（FDA）达成协议，正式启动首个将山中因子（Yamanaka factors）直接应用于人体的I期临床试验。山中因子是四种能够重置哺乳动物细胞表观基因组至年轻状态的转录蛋白。该试验瞄准青光眼及缺血性视神经病变，采用腺相关病毒（AAV）载体向眼球玻璃体注射编码基因，并开创性地使用常见抗生素多西环素（doxycycline）作为基因表达的开关控制机制。 David Friedberg强调，该技术的核心突破在于不改动原生DNA序列的前提下逆转表观遗传时钟。如果临床试验能成功修复神经损伤并恢复患者视力，不仅将攻克致盲性疾病，更将首次在活体人类组织中确证细胞逆龄重编程的医学可行性。目前已有十余家初创团队正在围绕山中因子及人造蛋白展开布局，尝试将细胞重编程拓展至骨关节炎、皮肤衰老及系统性器官退化等更广泛的衰老领域。 从软件行业估值体系的瓦解危机，到算力基建遭遇的电力与环保重围，AI革命的实际扩散速度正在剧烈受制于物理资源供给、体制惯性以及司法宪政审查的现实硬约束。虽然技术内部的生产力裂变与生命科学逆龄突破展现出巨大的增量价值，但如何协调分配红利、化解政策阻力并遏制政治报复，依然是决定新技术能否平稳落地的核心变量。

## Background 本期节目由知名科技与财经播客频道《All-In Podcast》于 2026 年 2 月 28 日在 YouTube 平台发布。节目简介重点导览了 Anthropic 旗下 Claude 系列新品引发的 SaaS 市场大震荡与 Citrini 末日信件热议、AI 恐慌叙事背后的认知偏差、数据中心遭遇地方阻力与白宫用电户保护承诺、长达 108 分钟的国情咨文两党交锋、山中因子逆转失明人体试验突破，以及最高法院裁定关税越权后特朗普的紧急关税转向等关键议题。 ## TL;DR 四位主持人深入剖析了由 Claude 密集发布产品引发的传统软件巨头股价暴跌，讨论了从 SaaS“确定性年金”到市场追问“现金流是否存在”的估值重构，并援引杰文斯悖论与自身团队部署 Agent 的亲身实践反驳了 AI 毁灭就业的末日论。随后，节目围绕全美数据中心建设受阻与白宫提出的用电户保护承诺展开辩论，探讨公用事业垄断和环保诉讼对基础设施落地的实质卡扣。在时政与前沿科技板块，嘉宾们深度复盘了创纪录时长的国情咨文演讲及背后折射的深层党派极化，解读了山中因子用于治疗失明的首个人体临床试验，并就最高法院 6:3 裁定特朗普紧急关税越权及其替代法律工具、司法独立与三权分权制衡展开了激烈交锋。 ## Atmosphere & style 整场节目交织着硅谷顶尖投资人的硬核商业洞察与好友间无拘无束的即兴互怼，节奏紧凑且信息密度极高。从开场调侃“单飞阴谋论”、恶搞德州口音，到中段就经济通缩与科技失业展开针锋相对的学术及实务辩论，现场张力十足；在讨论两党极化与政治法律战时，言辞更是辛辣犀利、火花四溅。收听体验既像旁听一场由顶级宏观资本操盘手主导的闭门头脑风暴，又宛如置身于四个性格迥异的老友围坐客厅、各执己见的畅快茶话会。 ## Bulletin - [01:22](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=82s) Anthropic 推出针对法律、网络安全与 COBOL 遗留系统现代化的多款工具，重挫 Thomson Reuters、CrowdStrike 等传统 IT 及 SaaS 龙头，IBM 更是创下单日暴跌 13%、市值蒸发 310 亿美元的 25 年来最惨纪录。 - [03:08](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=188s) 资本市场对 SaaS 与科技软件的定价逻辑发生质变，估值模型从研讨现金流受损的“时间点（When）”转向质疑其商业模式存续的“是否存在（If）”，高 WACC 与低倍数重估迅速蔓延。 - [35:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2140s) 全美数据中心建设因地方阻力与社区环保抵制已造成高达 1200 亿至 1300 亿美元的远期潜在产值蒸发，白宫为此推出“用电户保护承诺”以推动大科技公司自建独立供电。 - [01:03:58](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3838s) Life Biosciences 与 FDA 达成协议，将正式推进山中因子首个人体临床试验，通过 AAV 载体注射及分子开关逆转视网膜衰老并尝试治愈失明。 - [01:10:16](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4216s) 最高法院以 6 票对 3 票裁定特朗普依据紧急状态征收关税违宪越权，但特朗普随即启用《1974 年贸易法》第 122 条款出台 15% 临时关税作为反制路线图。 ## Key takeaways - [01:22](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=82s) Anthropic 二月发布的产品组合精准冲击法律、安全与遗留系统维护三大软件板块，打破了传统 IT 护城河。 - [03:08](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=188s) 对冲基金大幅降杠杆叠加 AI 黑天鹅事件风险，令资本市场放弃按坐席收费的高估值模型，导致科技企业股权激励（SBC）体系面临重构压力。 - [07:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=460s) Citrini 爆火的“2028 全球智力危机”虚构报告引发消费金融板块抛售，但背后做空基金联合作者的披露引发了市场对舆论操纵的质疑。 - [10:26](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=626s) 德里克·汤普森与大卫·萨克斯指出当前辩论多为“伪装成严谨分析的科幻小说”，预测市场对西特里尼式大崩溃的认可概率仅约 12%。 - [12:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=720s) SaaS 从曾经具 120% 净金额留存率（NDR）的增长型年金资产，沦落为增长上限与商业模式皆不明朗的高风险品类。 - [14:02](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=842s) 弗里德伯格提出全球消费能力可能存在物理上限，广义知识工作或许只是计算机诞生与 AI 成熟之间的过渡性历史现象。 - [19:47](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1187s) AI 带来的并非程序员失业，而是契合杰文斯悖论的生产力杠杆释放，企业潜在软件需求将被数十倍级工程效率迅速激活。 - [22:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1339s) 初创企业全面向“AI 原生”演进，基于开源模型与自主 Agent 的低成本自动化流程让内部自研平替传统 SaaS 成为可能。 - [30:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1840s) 末日叙事之所以广受欢迎源于固定馅饼谬误与反乌托邦心理偏见，而新技术真正的价值在于充当灌溉经济花园的活水。 - [40:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2419s) “用电户保护承诺”主张科技巨头在表后自建发电设备独立供电，避免 AI 基建能耗转嫁为居民电费上涨。 - [42:47](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2567s) 公用事业公司的垄断机制与 Capex 驱动利润模型才是电价潜在上涨的核心病灶，数据中心往往被误立为替罪羊。 - [52:13](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3133s) 长达 108 分钟的国情咨文凸显两党极化加剧，而弗里德伯格讽刺两党唯一真正的共识就是继续无节制推高国债赤字。 - [01:06:35](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3995s) FDA 批准山中因子针对眼部衰老的首次人体临床试验，通过多西环素分子开关调控表观基因组重编程，开启再生医学新纪元。 - [01:11:52](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4312s) 最高法院对关税的否定裁决彰显了司法独立与三权分立的有效制衡，但行政部门已借助第 122 与 301 条款确保关税政策常态化延续。 - [01:18:38](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4718s) 跨党派合作在法律战和党派清算情绪中举步维艰，呼吁中间温和派介入以遏制权力单边扩张与政局恶性循环。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=0s) 开场调侃：阴谋论角与最佳阵容的各自“单飞” 节目在轻松互怼的氛围中拉开帷幕。Jason 开玩笑地称这期节目为“阴谋论角”（Conspiracy Corner），宣称要彻底聊聊 9/11 内部真相、地平说以及肯尼迪遇刺案，甚至笑称今天的特邀嘉宾是 Alex Jones。 Chamath 随即打趣滑雪周播出的那一期只有极少播放量（“只有 9 次观看”），直言四个好友（Besties）缺席其三很难撑场面。Jason 幽默反驳自己一个人就能扛起至少 40 万播放量。 几人顺势调侃起彼此近期在主节目之外的动向：Chamath 创立了自己的个人 YouTube 频道，被调侃是为播客这列“失控列车彻底烧成废墟”后准备的逃生舱与对冲策略；Friedberg 也在搞个人项目。Jason 笑称现在的状态就像当年的披头士乐队（The Beatles）各自搞单飞、进行先锋实验，全员陷入了一种“小野洋子（Yoko Ono）式”的各自为政状态。随后 Jason 提到，节目专属 AI 机器人选出的头号话题本是 Anthropic CEO Dario Amodei 与国防部官员 Hegseth 之间的争端，但因裁决结果定于周五播客上线当天公布，该话题将留待下周详谈。 ### [01:22](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=82s) Claude 的“狙杀名单”与惨烈波及的传统软件巨头 Jason 迅速切入正题：Anthropic 堪称开启了一场“跨时代级别的连击”，整个二月接二连三发布产品，精准狙击了三个不同的二级市场板块，犹如接过了“砸盘之王”的接力棒。这串被称为“Claude 狙杀名单”（Claude hit list）的产品发布接连重创了美股上市公司： ```mermaid flowchart TD accTitle: Claude 产品发布对二级市场股票的连锁冲击 accDescr: 流程图展示了 2 月 Anthropic 发布的三大工具及其导致的相关行业巨头股价重挫 subgraph Feb03["2 月 3 日: Claude Co-work 法律插件"] L1["Thomson Reuters"] L2["LexisNexis"] L3["LegalZoom"] end subgraph Feb20["2 月 20 日: Claude Code Security 预览版"] S1["CrowdStrike"] S2["Cloudflare"] S3["Okta"] end subgraph Feb23["2 月 23 日: COBOL 数据库现代化工具"] I1["IBM (单日暴跌 13%，市值蒸发 310 亿美元)"] end Feb03 -->|"法律科技股全线重挫 10%+"| Feb20 Feb20 -->|"网络安全龙头股价受挫"| Feb23 Feb23 -->|"大型机与遗留系统遭受根本性冲击"| Market["传统 IT 与 SaaS 市场陷入估值恐慌"] ``` - **法律服务软件（2 月 3 日）**：Anthropic 推出针对 Claude Co-work 的法律插件，Thomson Reuters、RELX（LexisNexis 母公司）、LegalZoom 等法律科技标的应声下跌超过 10%。 - **网络安全工具（2 月 20 日）**：Claude Code Security 开启受限研究预览，直接引发对传统网安护城河的质疑，CrowdStrike、Cloudflare 与 Okta 等公司股票集体大跌。 - **COBOL 遗留系统现代工具（2 月 23 日）**：Anthropic 宣布 Claude 能够对 COBOL 代码进行现代化重构。COBOL 作为最古老的编程语言之一，驱动着全美 95% 的自动取款机（ATM）、支撑着联邦社会保障金的划拨发放，并在银行、薪酬管理、医疗及政府机构的底层深度运行；而其中 85% 的 COBOL 代码运行在 IBM 的主机硬件与系统上。该功能一经公布，IBM 周一股价暴跌 13%，创下自 2000 年互联网泡沫破裂以来的单日最大跌幅，市值单日蒸发高达 310 亿美元。 面对这些剧烈波动，Jason 抛出核心疑问：这到底只是市场在历史高位上获利了结、寻找借口减仓的暂时性抛售，还是 AI 彻底压缩传统软件利润空间、颠覆其商业模式的新现实？ ### [03:08](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=188s) 市场逻辑的质变：从“何时（When）”到“是否（If）” Chamath 对市场异动给出了兼顾短期策略与底层结构的双重分析框架。在战术层面，当前顶尖量化与对冲基金正在大规模执行“去杠杆与降风险敞口”（degrossing），通过同步压缩多头与空头仓位来缩小头寸规模，这在整个市场中形成了自上而下的普跌抛压。 然而在更根本的战略结构层面，资本市场对资产定价的核心逻辑已经发生了颠覆性的范式转移。Chamath 指出，在过去的常态市场下，投资者争论的从来都是“时间点（When）”——即高确定性的未来现金流将在“什么时候”开始受到竞争侵蚀（例如 Coca-Cola、Eli Lilly 或 Meta 的现金流何时受损）。这种讨论映射在公共市场的三大估值指标上：市盈率（P/E）、市销率（P/S）以及加权平均资本成本（WACC）。 - **市盈率（P/E）**：作为投资收益率的倒数，20 倍 P/E 等价于 5% 的资本回报率。 - **加权平均资本成本（WACC）**：用来折现未来 20 至 30 年的预期收益。低 WACC 代表市场相信现金流极度稳固抗跌；高 WACC 则意味着未来现金流面临剧烈贴现。 当前市场的定价核心已经彻底从“何时发生（When）”转变成了“是否存在（If）”。市场参与者开始惊恐地质问：这笔现金流是否根本连三年都撑不过去？拐角处是否正潜伏着某个尚未面世的全新 AI 模型，会以无人知晓的方式将整家公司的业务彻底夷为平地？ 面对这种无法量化黑天鹅概率的“事件风险（Event Risk）”，股票持有者开始疯狂寻求巨大的安全边际（Margin of Safety）：过去 40 倍市盈率的股票被下杀到 20 倍，20 倍的被砍到 10 倍；10 倍市销率被无情打至 3 倍；折现率直接从 6% 激增至 12% 乃至 13%。Chamath 强调，这种市场重构还将引发严重的连锁反应，特别是直接冲击科技企业的股权激励（Stock-based Compensation）体系，从而全面破坏其吸引与留存顶尖人才的根基。 ### [07:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=460s) Citrini 爆火科幻文与做空者操纵疑云 随后话题转向周末引发金融圈地震的另一催化剂：Citrini Research 在 Substack 上发布的一篇以虚构的“2028 年全球智力危机”（2028 Global Intelligence Crisis）为背景的推演文章。该文在社交平台 X 上的阅读量迅速突破 2800 万次，引发了广泛恐慌并在周一进一步加速了美股抛售潮。 该文构想了一个由 AI 催生的“经济通缩死循环”：企业为提振利润而全面引入 AI 裁员（效仿大型科技公司的精简模式），效率大幅提升，净利润率激增；然而由于大规模失业，工薪阶层失去可支配收入，导致终端消费市场崩盘；面对需求暴跌，企业被迫进一步加码 AI 以维持利润指标并实施更残酷的裁员，最终拖垮整个宏观经济。文中断言失业率将飙升至 10%，标普 500 指数自高点暴跌 38%。 此外，文中设想自主 AI Agent 将直接消灭 3% 的信用卡交易刷卡手续费（Interchange Fees），将支付结算全盘迁移至稳定币体系。这导致周一消费金融板块应声大跌： - American Express（Amex）重挫 8% - Capital One 下跌 8% - Mastercard 下跌 6% - Visa 下跌 4% Jason 补充道，文中甚至断言 AI Agent 能够仅凭“氛围编程”（Vibe Coding）就颠覆 DoorDash 这类重度依赖双边网络效应与地面交付履约的巨头，这一推断显然缺乏对真实线下网络商业模式的常识理解。 Sacks 随即对这篇文章的“自发病毒式传播”提出了严厉质疑。最新披露的信息显示，该文在引发市场巨震后悄然修改了作者署名，追加的联合作者实际上是一家管理着 2.62 亿美元、在 SEC 注册的对冲基金的管理合伙人，且该基金已被证实直接做空了文章中重点点名做空的公司。Sacks 指出，这究竟是一场真正的思想病毒式传播，还是做空机构自导自演、借助舆论放大器的精准做空操纵，目前充满巨大问号。即便抛开其动机不谈，单就内容逻辑而言，也引来了如著名作家 Derek Thompson 等人的尖锐反驳。 ### [10:26](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=626s) 科学幻想之争：德里克·汤普森与“没人真正知晓未来” 大卫·萨克斯（David Sacks）提到了著名大西洋月刊专栏作家德里克·汤普森（Derek Thompson）撰写的一篇题为《没人知道任何事》（Nobody Knows Anything）的文章。这个标题脱胎于好莱坞传奇编剧威廉·高德曼（William Goldman）的名言。文章直言不讳地指出，当下根本没有人真正知道 AI 在两年后、更不用说二十年后会走向何方；正因为无人知晓，大家便纷纷诉诸“伪装成严谨分析的科幻小说写作”。汤普森将当下关于 AI 的各种宏大辩论形容为一个“彼此竞争的科幻叙事市场”。他写道：“这并不是说我认为这项技术是无聊的把戏，而是因为其不确定性如此之高，关于 AI 宏观经济效应的真实、实时数据又如此匮乏，以至于那些极其严肃的 AI 讨论往往更偏向文学创作，而非真正的分析。” 换言之，像西特里尼（Citrini）那样引发市场巨震的末日报告，本质上只是一篇引人入胜的科幻小说，缺乏任何能够自洽的扎实数据支撑。萨克斯展示了一份预测市场的数据，显示押注西特里尼末日情景成真的概率大约只有 12%。现实是，市场中充斥着完全对立的科幻剧本——一派预言全行业崩塌与大萧条，而另一派（比如汤普森本人和以斯拉·克莱因 Ezra Klein 所代表的“丰裕时代派”）则描绘了一个 AI 创造物质极大丰饶、人类再无匮乏的世界。查马斯·帕里哈皮蒂亚（Chamath Palihapitiya）对此总结道，正是这种极端的不确定性让资本市场遭到重锤：“如果买政府国债就能旱涝保收拿到 5% 的无风险收益，投资人凭什么要在这里承担过度的不确定性风险？” ### [12:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=720s) SaaS 的确定性破灭：从“增长型年金”到定价重估 萨克斯进一步深入剖析了软件股和 SaaS（软件即服务）遭受惨烈抛售的核心逻辑。他回顾了自己作为风险投资人见证的软件发展史：过去十几年里，SaaS 之所以是全市场最受青睐的商业模式，是因为它极度易于建模且极具确定性。在任何一个细分领域，经历初期的群雄逐鹿和产品试验后，市场便会走向收敛，诞生一个占据大部分市场份额和绝大多数市值的品类领导者。这些公司的财务指标如精密的机械般可预测：看年度经常性收入（ARR）、看履约义务余额（RPO），尤其是高达 120% 的净金额留存率（Net Dollar Retention, NDR）。 这种商业模式曾被资本市场视为“自带增长的年金”。120% 的留存率意味着老客户群体不仅全部续约，还会因为增加坐席、购买新功能或交叉销售而多付 20% 的钱。因此，当投资者以高达 13 倍 ARR 的估值倍数抢购这些公司时，他们确信自己买入的是坚如磐石的成长型现金流。然而，AI 的突袭彻底击碎了这种模型： - **增长上限受蚀**：即便 AI 不会彻底干掉 Salesforce 等行业巨头，也会严重蚕食它们的未来增长空间。 - **商业与定价模式颠覆**：当软件代码和功能可以通过 AI 极低成本生成，传统的按坐席收费（Per-seat pricing）体系是否面临土崩瓦解？ - **估值重估**：短短一年之内，SaaS 从最具可预测性的资产类别变成了充满巨大未知数的重灾区，市场因此出现剧烈震荡。 ```mermaid flowchart LR accTitle: SaaS 确定性破灭逻辑 accDescr: 展示传统 SaaS 的商业飞轮如何被 AI 的出现打碎并导致估值重估 subgraph Past["过去：黄金时代的增长年金"] A["品类领导者确立"] --> B["按坐席收费与高转换成本"] B --> C["120% 净金额留存率 (NDR)"] C --> D["13x+ ARR 估值溢价"] end subgraph Disruption["当下：AI 带来的冲击与未知"] E["AI 编码与智能体"] --> F["代码边际成本趋零"] F --> G["按人头收费模型受挫"] G --> H["增长空间与估值倍数双杀"] end D -.->|模式遭遇挑战| Disruption ``` ### [14:02](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=842s) 弗里德伯格的沉思：消费上限与“过渡期”的知识工作 杰森·卡拉卡尼斯（Jason Calacanis）借用《权力的游戏》名言指出“混乱是阶梯”，询问大卫·弗里德伯格（David Friedberg）：在 SaaS 承压的背景下，谁会是赢家？是开源生态，还是整个经济陷入通缩、企业利润被无限压缩？ 弗里德伯格提出了一个更为根本且具有哲学意味的经济学假设：AI 带来生产力的根本逻辑在于对“人类时间”和“资本”施加极高倍数的杠杆。从人类心理学来看，人们幸福感的底层驱动力是生活水平以每年约 10% 的速度改善——如果物价不变而收入增长 10%，买得起比去年多 10% 的新东西，人就会感到快乐；如果增长低于这个标准，人就会沮丧。这一直是驱动社会消费能力的“下限”。 然而，弗里德伯格指出，人类历史如今首次面临一个从未遇过的危机：**全球消费能力是否存在“物理上限”？** - **供给超越需求的临界点**：过去技术进步带来杠杆时，人们用更少资源制造更多产品，消费品降价让所有人享受丰裕。但现在，AI 让“制造物品与服务的能力”彻底压倒了人类自身的“消费与消化能力”。 - **知识工作的过渡属性**：不仅 SaaS 可能只是互联网兴起到 AI 普及之间的一个“过渡性商业现象”，甚至广义上的**知识工作（Knowledge Work）**本身，也可能只是计算机发明与 AI 诞生之间昙花一现的“过渡性历史现象”。 - **无处安放的产能**：如果所有脑力劳动者都被 AI 替代或赋予百倍生产力，去从事更高维、更有创意的事务，另一端真的有足够多的消费者去承接这膨胀 100 倍的产出吗？当通用经济模型与社会模型无法解释这种过剩，体系便开始崩塌。 ### [17:28](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1048s) 反思科幻末日：Anthropic 的高薪招聘与 Citadel 研报 萨克斯直接打断了弗里德伯格的宏大推想，认为这依旧是脱离现实的“科幻空谈”，并摆出具体的现实数据予以反驳。他指出，目前 AI 颠覆各行业最先落地的杀手级应用是“代码辅助生成”，正是代码编写门槛的大幅降低引发了市场对 SaaS 生态崩塌的恐慌。然而，行业现实中却出现了荒诞的自相矛盾： Anthropic 官方网站上目前赫然挂着一个软件工程师职位的招聘信息，开出了高达 57 万美元的年薪。萨克斯犀利地指出，Anthropic 等 AI 巨头一边向外界宣扬 AI 即将淘汰程序员和软件工程师，一边却自己在公开市场上砸重金争抢工程师。查马斯随即打趣说自己都想去应聘这个岗位搞点“紧缩自救”，或者入职后直接让 AI 替自己打工；他还笑称“千万别让我旗下 8090 团队看到这个薪酬包”，并透露 Anthropic 如今每年开启 50 到 60 亿美元规模的有结构二级市场交易，意味着其期权激励几乎等同于硬通货现金，要想匹配这样的薪酬，自己得开出三倍的价钱。 萨克斯进一步援引了城堡证券（Citadel Securities）近期发布的一份重磅反驳报告：该研报专门回击了西特里尼的市场崩溃论，并用扎实的客观数据表明，软件工程岗位的市场招聘需求不仅没有萎缩，反而正在迅速回升，其岗位招聘量同比增长了约 10%。数据证明，现实世界并未按照末日科幻剧本上演，AI 的崛起反而在短期内引发了对顶尖工程人才更为激烈的争夺。 ### [19:47](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1187s) 杰文斯悖论与软件工程师的“无限需求” 大卫·萨克斯（David Sacks）顺着弗里德伯格（David Friedberg）关于创业杠杆的观点指出，当下不仅企业注册成立的数量正在急剧激增，而且背后的驱动力正是人工智能让创业变得前所未有的容易。萨克斯特别引用了 Box 首席执行官亚伦·莱维（Aaron Levie）对于这一反直觉现象的解释——其核心正是“杰文斯悖论”（Jevons Paradox）：当一种此前受限于供给短缺的资源或服务成本大幅下降时，对其的实际需求并不会萎缩，反而会呈爆发式增长。 以软件工程师为例，即便在硅谷那些对顶尖工程师最具吸引力的初创公司中，技术人才也始终处于长期短缺状态；而对于《财富》500 强中的传统非科技企业而言，招募优秀技术人员更是难上加难，全社会存在着巨大的未被满足的软件开发缺口。如今，AI 正在赋予每位工程师数倍甚至数十倍的杠杆，这绝不意味着他们会被大批裁员，而是意味着整个经济体中将涌现出大量“十倍级工程师”（10x software engineers）。萨克斯提到，目前传统《财富》500 强企业的成本结构中，全部 IT 开支仅占大约 5%，纯软件投入仅有 1% 到 2%；借用埃隆·马斯克（Elon Musk）将现代企业比作“半人半软件的赛博有机体”（cybernetic organisms）的概念，未来软件在企业成本结构中的占比完全可能飙升至 50%。即便工程师生产力提升十倍甚至百倍，庞大的潜在需求也能迅速将新增供给吸收殆尽，从而带来生产力的大爆发而非大规模失业。 查马斯·帕里哈皮蒂亚（Chamath Palihapitiya）对此表示赞同并补充道，从财务模型来看，未来企业运营支出（OPEX）占总营收的比例大概率会“断崖式下跌”；但在这大幅压缩后的 OPEX 预算盘子内部，分配给技术及相关软硬件的占比将会远超当下。 ```mermaid flowchart TD accTitle: Jevons Paradox in Software Engineering accDescr: Diagram showing how AI efficiency leads to higher software demand instead of job cuts. subgraph SupplySide["技术与生产力供给"] A["生成式 AI 工具普及"] --> B["工程师生产力提升 10x-100x"] B --> C["单位软件开发成本大幅骤降"] end subgraph DemandSide["经济体潜在需求消化"] C --> D["激活财富 500 强未满足的 IT 需求"] D --> E["企业内部软件占比从 1%-2% 跃升至 50%"] E --> F["高杠杆工程师被广泛吸纳（无大规模失业）"] end ``` - **IT 成本比例倒挂**：目前一般《财富》500 强企业的 IT 总成本仅占其整体成本结构的 5% 左右，其中软件仅占 1% 到 2%，未来提升空间巨大。 - **财务结构剧变**：企业整体的 OPEX 占营收比重将大幅收缩，但技术相关支出在剩余预算中的相对份额将大幅飙升。 --- ### [22:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1339s) 知识工作者全面智能体化：Launch 团队的 30 天自动化实验 查马斯向负责孵化器 Launch 的杰森·卡拉卡尼斯（Jason Calacanis）发问，探寻现在申请孵化器的创业项目是否已经不再做传统 SaaS、转而用更低成本重构旧工具。杰森直言，现在的项目清一色都是“AI 原生”（AI-first）企业。就像此前讨论过的，有的团队致力于训练世界上最好的飞行员，Waymo 致力于打造世界上最好的自动驾驶司机，现在的创业者则在争先恐后地打造世界上最好的销售开发代表（SDR）、顶尖销售员或高管教练。 杰森随后分享了自己机构内部刚刚发生的剧烈变革。他认为未来驾驭并构建这一波软件的并非传统程序员，而是广大的知识工作者（Knowledge Workers）。上周末，杰森将公司 20 名员工中的 15 人召集到办公室，进行了长达六七个小时的集中训练，让每个人配置并拥有自己的 OpenClaw 智能体，直接动手搭建流程。在短短 30 天内，团队把过去 10 年里想买却舍不得买、想做却无暇开发的内部软件全都写了出来： - **播客赞助商线索挖掘 SDR 智能体**：自动抓取全网前 100 名播客的音频文字稿，识别赞助商身份，同步在 CRM 系统 Pipedrive 中检索过往跟进记录，并推送到 Slack 销售频道。这一原本由全职初级销售代表负责的工作，如今由智能体通宵达旦、零差错地运转。 - **代号“奥创”（Ultron）的管理智能体**：获得了 Google 邮箱、日历、Zoom、Notion 和 Slack 的底层访问权限（Root Access），每周自动统计各员工的工作成果、发出邮件、会议记录及参与的核心讨论线程，协助管理者开展精细化考核。 - **音视频切片智能体**：能自动翻看《This Week in Startups》十年前的历史视频，挑选出最精彩的 3 段切片，自动压制字幕并推送至 Slack，直接省去了一个全职视频剪辑岗位。 - **自我进化（Recursive）的封面设计智能体**：智能体配备了“灵魂文件”（Soul file）与“技能文件”（Skills file）。程序每周六自动扫描 YouTube 上的标题与封面运营讨论，甚至自主挖出了 MrBeast 团队关于热力图运用的独家内部分析文章并吸收进技能库，现在不仅能针对每次上传的缩略图给出优化修改意见，还开始亲自生成封面。 面对查马斯的调侃——“按你这说法，你业务在增长而且生产力大涨，看来你非但不是悲观派，反倒要招更多人了？”，杰森斩钉截铁地回答：“不，我们绝对不会再招新人。”他解释道，现有员工的综合效率正以每周 10% 到 20% 的幅度递增。杰森甚至将这套工作流移植到了家庭生活中，让智能体抓取全家在 Instacart 上的最近 10 次订单记录，分析高频回购商品并自动生成采购购物车。杰森兴奋地感慨道：“这是我自万维网诞生以来，在互联网上度过的最激动人心的时刻！” --- ### [26:48](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1608s) 科技企业的“买时间”考验与股权激励（SBC）危机 查马斯将讨论引向了一个更为严肃宏观的资本结构命题。无论是小团队的激进实验，还是大企业迟缓却必然的跟进，所有公司在摸索出 AI 时代的清晰商业模式之前，都必须“加固舱门”（batten down the hatches），为自己争取足够长的转型窗口期。而想要买到五六年的探索时间，企业账上就必须趴着极其充沛的现金流。 查马斯尖锐地指出，当今科技企业最严重“焚烧现金”的无底洞，恰恰是其薪酬体系——尤其是股权激励费用（SBC, Stock-Based Compensation）。绝大多数所谓的科技巨头，为了在公开市场上回购股票以抵消因发放巨额员工期权导致的股权稀释，几乎耗尽了所有的自由现金流（Free Cash Flow）。查马斯预言，行业的“第二只靴子”即将落地：科技企业将被迫向内动刀，彻底清算臃肿的 SBC 支出，转而奉行极度保守的现金储备策略，只有具备极强造血能力的现金流生成机器，才能在这场拉锯战中活到终局。 萨克斯和杰森则从人才与创业生态的角度予以回应。杰森坚信，“那些主动拥抱 AI 的人，其身价将比不愿接纳的人高出五到十倍”，经济将迎来真正的繁荣大爆发（Boom town）。如今创立一家软件公司的门槛已从过去的 300 万至 400 万美元骤降至微不足道的水平，两三个人配置几组智能体，就能同时包揽软件研发、客户销售和公关传播。杰森分享了一个让自己彻底警醒的瞬间：当团队向智能体提出希望从 Slack 中获取某项特定功能、而 Slack 本身并不支持时，智能体不仅直接给出了替代方案，还主动提议道：“Slack 确实没有，但你考虑过 Mattermost 吗？这是一个开源项目，我马上就可以为你部署一套。”这种从想法到落地的即时性，正在重塑整个软件行业的底层逻辑。 ### [29:02](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1742s) SaaS 模式的黄昏与开源平替的冲击 随着 AI 编程与 Agent 能力的爆发，企业自研定制软件的门槛被彻底打碎。Jason 分享了自己团队的现状：技术人员周末跃跃欲试，甚至想把公司使用的 Slack 导出并在本地自建一套平替。尽管目前公司在 Slack 上每年仅花费六千到一万美元，但这揭示了一个趋势——AI 正在帮企业自建 CRM、自研专属 Agent，并有接管整套企业软件栈的势头。在未来与 Slack、HubSpot 等主流 SaaS 厂商谈判或续约时，企业完全拥有了“我们可以内部自研，不再需要被增销（Upsell）新功能”的议价底牌。 ```mermaid flowchart TD accTitle: AI 时代软件价值链重构 accDescr: 模型层、应用层与硬件层之间的价值分配博弈 subgraph Infrastructure["底层基础设施"] Chips["芯片与算力硬件 (NVIDIA 等)"] FMs["底层基础模型 (OpenAI / Anthropic / 开源模型)"] end subgraph Applications["上层应用生态"] VerticalApps["垂直行业专有 SaaS (Harvey 等)"] InHouse["企业自研 Agent / 内部平替系统"] end Chips --> FMs FMs --> VerticalApps FMs --> InHouse ``` 与此同时，SaaS 的震荡正在由上市公司向一级市场的独角兽蔓延。Flexport 创始人 Ryan Petersen 在社交平台 X 上直言，Anthropic 最新发布的 Claude 法律插件效果已经不逊于估值高昂的明星法律 AI 初创公司 Harvey。这再次引发了关于技术栈中究竟哪一层能够捕获核心价值的深层讨论：究竟是底层大模型公司、上层垂直应用，还是最底层的芯片硬件巨头？如果底层基础模型和开源模型持续进化，垂直应用若没有极深的护城河，其生存价值将极其脆弱。 - 开源模型的性价比跃升正加速这一进程，例如团队近期部署了开源模型 Kimi 2.5，足以完成 80% 至 85% 的日常工作流，直接大幅压降了调用商业 API 的 Token 账单。 ### [30:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1840s) 为什么“AI 末日论”更深入人心：认知偏差与固定馅饼谬误 面对当前科技界围绕 AI 展开的各类争论，David Sacks 指出，这些论调本质上是两种截然对立的科幻叙事在交锋，但“AI 毁灭世界”或“AI 导致大萧条”的末日论叙事（Doomer Narrative）天然比乐观主义更容易获得大众共鸣。其背后有几大根深蒂固的人性认知偏差： 1. **反乌托邦文化的吸引力**：绝大多数主流科幻影视作品都倾向于描绘反乌托邦与灾难，毁灭往往比繁荣更具戏剧张力。 2. **看得见与看不见的偏差（The Seen vs. The Unseen）**：人们很容易看清眼前那些可能被 AI 自动化的既有工作岗位，却极难想象那些尚未被发明的新职业与全新商业模式。构想未来的创造性破坏需要天才与卓越的创新者，而想象既有事物的毁灭却不需要任何想象力。 3. **固定馅饼谬误（Fixed Pie Fallacy）**：公众习惯性将经济规模视作一块大小恒定的馅饼，误以为某人依靠新技术致富就必定剥夺是他人的利益。 > “经济从来都不是一块固定大小的馅饼，它是一座花园，而技术就是灌溉它的雨水。” 技术的创新不仅不会缩减总体机会，反而会加快整个经济花园的生长速度；某个产业生产力的跃升，并不意味着其他领域的就业就会必然消亡。 ### [32:21](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1941s) “Agent 指挥官”与算力基础设施的现实物理枷锁 谈到未来必然诞生的新型高价值岗位，Jason 指出，当前最紧缺、最有前景的角色正是“Agent 编排大师”（Agent Maestro / Agent Manager）。这类人才往往不是传统的底层软件工程师，而是深谙业务运营、能将复杂的业务逻辑精准拆解、并训练管理各类 AI Agent 的运营操盘手。Sacks 对此表示赞同，他强调任何颠覆性技术落地时，企业都面临着极其沉重的“变革管理”（Change Management）成本，能够带领组织完成智能化流程重构的人将迎来极佳的职业机遇。 但与此同时，物理现实的客观阻力也注定让极端激进的预测难以在短时间内兑现： - **Token 供需剪刀差**：虽然预计到年底单个输出 Token 的成本有望暴跌 90%，激发出超过 10 倍的需求井喷，但算力扩张已逼近现实物理极限。 - **重工业瓶颈阻滞**：哪怕芯片算力在单位能耗与成本维度（Tokens/Watt, Tokens/Dollar）持续飙升，基础设施的扩建依然严重受制于土地规划、电网容量、机房外壳（Shell）建设周期以及芯片制造产能。 - **企业用工成本重构**：Chamath 透露，其团队在 8090 的成本核算模型中，已经正式把算力 Token 消耗列为了员工全口径成本（Fully Burdened Cost）的固定科目，部分资深工程师跑模型产生的算力账单已直逼甚至超过薪资水平。 - **边缘算力新可能**：市场传闻苹果即将发布搭载 M5 芯片的 Mac Studio，专门针对大语言模型本地运行进行了硬件级优化，桌面级本地算力可能成为缓解云端能源压力的一个关键变量。 ### [35:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2140s) 数据中心地方抵制潮与千亿美元的算力经济损失 阻碍算力落地的一大突出现实矛盾，正在于全美各地对数据中心建设的政治反弹与社区抵制。Chamath 结合自己在 X 平台发布的调查数据，剖析了算力基建面临的严峻逆风： | 统计指标 | 2025 年实际情况 | 2026 年预估风险 | |:--- |:--- |:--- | | **受阻/抵制项目规模** | 约 25 个项目（仅第二季度就达 20 个） | 目前有上百个数据中心项目遭遇地方阻力 | | **损失电力容量 (GW)** | 约 5 GW 规划装机被取消 | 按比例推算恐将损失约 7 GW | | **行业潜在营收损失** | 约 500 亿美元 | 约 700 亿美元 | 根据 OpenAI 首席财务官 Sarah Friar 的公开测算，每 1 个吉瓦（GW）的电力容量能够为 AI 行业支撑并带来大约 100 亿美元的年化收入。照此计算，仅 2025 至 2026 两年间，因项目搁浅而蒸发的行业远期潜在产值就高达 1200 亿至 1300 亿美元。 这一现状令人唏嘘：五到十年前，全美各地市镇政府还在为了争夺数据中心入驻而争相开出税收减免与用地优惠；如今，数据中心却在多地面临环保、噪音及用电抢占的激烈指责，从香饽饽变成了被抵制的对象。如果这股政治与社区阻力继续蔓延，将成为锁死美国科技增长空间的最大瓶颈。 ### [37:46](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2266s) All-In Liquidity 大会预告与重量级嘉宾阵容 在切入国情咨文（State of the Union）讨论之前，节目公布了 All-In 团队 2026 年的重磅线下活动安排。团队将于 5 月 31 日至 6 月 3 日在加利福尼亚州知名酒乡扬特维尔（Yountville）举办“Liquidity”闭门峰会。 主持本届峰会内容筹备的 Chamath 调侃道，自己这次“严把门槛，绝不允许平庸之辈（Mids）插手议程”。他现场剧透了首批确认出席的主讲嘉宾： - **Dan Loeb**：华尔街著名对冲基金 Third Point 创始人，横跨私募股权、公开市场、不良资产与科技投资的传奇领袖。Loeb 极少接受公开演讲邀请，本次将在峰会带来罕见且极具分量的重磅 Keynote。 - **Sarah Friar**：OpenAI 现任首席财务官，曾掌舵多家顶级科技巨头的资深高管，她将深度解析 AI 前沿的商业化逻辑与万亿级算力投资版图。 ### [38:50](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2330s) 峰会预热与比尔·格利新书插曲 在正式进入严肃议题前，主持群以轻松的调侃热场。Jason Calacanis 与 David Sacks 谈及即将到来的 All-In 峰会，透露届时将在台上深入拆解 OpenAI 的整个商业模式。今年已是 All-In 峰会的第五个年头（9 月 13 日至 15 日），回想起往年“回到未来”和“银翼杀手”主题派对的盛况，现场气氛格外热烈。 随后 Jason 借机为播客挚友、传奇风险投资人比尔·格利（Bill Gurley）打 call，大力推荐其新书《奔跑追梦》（*Running Down a Dream*），甚至当场呼吁听众暂停播客立刻买三本——两本送给年轻人，一本给家长。Chamath 顺势起哄让 Jason 模仿他和格利合开播客的样子。Jason 随即用浓重德州口音惟妙惟肖地来了一段段子：“欢迎收听 JCBG 播客，我们正在德州 Terry Black's 吃烤牛胸排，今天我们要聊聊平台型市场投资，还有我的新书《奔跑追梦》——教你的孩子如何别搞砸人生；如果他们搞砸了，你还可以用德州特色拿这本书狠狠抽他们的后脑勺。”这番即兴模仿引得众人大笑。 --- ### [40:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2419s) 特朗普的“用电户保护承诺”与“BANANA”阻碍 幽默过后，话题转入白宫正在推动的一项重大政策——特朗普总统在国情咨文演讲中宣布支持的“用电户保护承诺”（Ratepayer Protection Pledge）。Sacks 指出，地方民众对新建 AI 数据中心的主要抗拒，根源在于害怕自己的居民电费账单暴涨。这项承诺的核心诉求极为直接：要求大型科技巨头（Hyperscalers）必须自筹 AI 数据中心所需的电力，或者允许其“表后”（behind the meter）自建发电设备独立供电，从而保障居民用电价格绝不上涨。 ```mermaid flowchart TD accTitle: Ratepayer Protection Pledge Mechanism accDescr: Diagram illustrating how hyperscalers self-generate power behind the meter to decouple from the residential grid. subgraph Hyperscalers["科技巨头与数据中心"] DC["AI 数据中心"] GEN["表后自建发电设备<br/>(Behind-the-Meter)"] GEN -->|独立供应能源| DC end subgraph GridImpact["对公共电网与消费者的影响"] GRID["公共电网"] GEN -.->|多余电量反哺| GRID SCALE["分摊固定基础设施成本"] SCALE --> PRICE["降低居民终端电价"] end DC -.->|不抢占公共负荷| GRID ``` Sacks 解释道，特朗普早在一年多前就提出过，美国最大的 AI 公司未来也将成为大型能源公司。当科技公司被允许在表后自建发电设施、开展同址部署（colocation）时，数据中心甚至无需接入公共电网；而一旦它们选择并网，不仅能把富余电量反哺电网，还会投入巨资扩建电力基础设施。由于发电成本中包含大量庞大的固定成本，电网规模效应的提升反而有助于拉低整体分摊后的阶梯费率。 Sacks 尖锐批评了以伯尼·桑德斯（Bernie Sanders）为代表的极左翼全面叫停建设的论调，并戏称如今阻碍基建的力量已经从“邻避主义”（NIMBY）升级成了“BANANA”——即“绝不在任何人附近的任何地方建任何东西”（Build Absolutely Nothing Anywhere Near Anyone）。相比之下，白宫的策略是“在拥抱进步的同时，让大科技公司自己掏钱买单，不转嫁给普通家庭”。 --- ### [42:47](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2567s) 数据中心全球竞速与公用事业垄断病灶 David Friedberg 站在宏观地缘经济的角度，阐述了为什么阻挠数据中心建设极其愚蠢。他强调，数据进出数据中心是以光速在光纤网络中传播的，因此数据中心理论上可以设立在世界任何地方。数据中心就像是当代的“石油钻井、铁路线、电报网和工厂”，是全新底层工业体系的核心。全球共有 196 个国家，全世界所有数据中心的物理占地面积缩小到世界地图上看不过是一枚大头针的针尖，却承载着数千亿美元的经济附加值与数十万的高薪岗位。如果美国因为政策阻挠而叫停，不仅无法遏制 AI 的发展，资本与计算需求还会直接流向其他国家。 Jason 与 Chamath 随即指出，中东的沙特、阿联酋和卡塔尔等国极其精明，他们正依托能源优势将数据中心建设计划翻十倍（10x），以极快速度审批落地并扫清官僚繁文缛节。此外，针对某些环保人士声称数据中心耗水的说法，Jason 直斥这是彻头彻尾的“骗局”（hoax），因为现代冷却水完全是闭环内循环利用的。 然而，Chamath 也提出了一个至关重要的警示：即使科技巨头全额承担了自建电力，**居民电价依然存在持续飙升的风险**。这其中的根源在于传统公用事业公司（Utilities）扭曲的商业模式： - **垄断特许经营**：公用事业公司在特定区域拥有天然垄断特权。 - **资本开支驱动回报机制**：为获取利润，电力公司会向公用事业委员会（PUC）提交庞大的资本开支预算（Capex plan），声称需要升级电网与发电机组。由于电力公司的合法回报率直接与其投资资本总额挂钩，它们天然具有借机做大基础设施投资以赚取资本回报的内在动力。 - **问责缺失**：居民日常耗电量本就在随电动汽车、家用电器增多而上升，若不严格监管公用事业公司的 Capex 转嫁机制，居民电费上涨的黑锅最终很可能会被错误地扣在数据中心头上。 最后，Jason 补充了一个令人沮丧的实例：在纽约等深蓝州，美光科技（Micron）规划投资高达 1000 亿美元的超级晶圆厂，竟然仅仅因为 5 到 6 名当地居民的诉讼就被卡住，从项目宣布到真正动工整整耗费了 1200 天，光是审批与程序拉扯就消耗了 612 天，充分揭示了繁琐监管和恶性诉讼对国家先进制造业发展的严重阻碍。 ### [48:24](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=2904s) 环保诉讼与基建阻力：少数人何以阻断百亿美元项目 讨论由环境影响评估引发的基建延误展开。大家对比了不同州对待重大项目的态度：在得克萨斯州，埃隆·马斯克仅用不到 18 个月就建成了特斯拉超级工厂（Gigafactory），展现了极高的审批与建设效率；而在以纽约为代表的蓝州，企业往往面临层层阻碍。嘉宾们感叹蓝州的这种自我限制令人费解，仿佛是在为了满足极少数人的诉求而拖累整个国家的未来。纽约当地绝大多数民众其实极度渴望获得工作机会，该项目并非普通仅需少数运维人员的数据中心，而是一座能创造大量高薪制造岗位的芯片制造厂（Chip Fab）。然而令人难以置信的是，在经历了长达两年的环境审查程序之后，仅仅凭借六个人的诉讼，就能让总投资高达千亿美元的宏大项目陷入停滞。 查马斯指出，这种阻碍机制的根源不能简单归咎于红州与蓝州的党派差异，而是非营利环保组织滥用法律机制所致。他回顾了自己在内华达州调研一项旨在推进美国锂资源本土化开采的巨额投资案例。当时内华达拥有极具开采价值的大型锂矿床，但就在即将获得环保审批的关键节点，当地环保组织以保护某种地表杂草/植被为由发起诉讼，硬生生阻断了关乎国家安全的关键供应链项目。环保非营利组织可以零成本、零风险地发起这些阻挠行动，甚至借此大肆募捐。不过，司法体系也开始出现反弹，例如北达科他州法官针对绿色和平组织（Greenpeace）阻挠石油管道项目的混乱破坏行为作出裁决，责令其赔偿近 3.5 亿美元的损失。 弗里德伯格从更深层的社会情绪切入，认为抗议背后往往缺乏理性逻辑，却充斥着强烈的负面情绪。公众对大型科技公司（Big Tech）、极少数精英聚敛财富以及缺乏普惠性质的经济增长抱有天然的敌意与被时代抛弃的失落感。如果未来的价值创造机制无法让更广泛的普通大众参与其中并切实体会到收益，社会力量就会动用手头的否决权去阻止项目落地。哪怕这六名诉讼发起人提出了住房成本、工人接触有毒化学品、空气与水污染、温室气体排放、能耗及湿地受损等一系列理论上完全可被工程缓解的问题，核心依然是公众需要成为利益的共享者。 ```mermaid flowchart TD accTitle: 项目建设面临的环保诉讼循环 accDescr: 展示巨额投资在遭遇非营利组织诉讼与公众排斥后陷入停滞的逻辑闭环 subgraph 投资与审批 A["重大基建/芯片制造厂立项"] --> B["耗费数年完成环境影响评估"] end subgraph 外部阻力介入 C["公众对大科技与财富差距的抵触情绪"] --> D["环保非营利组织零成本发起诉讼"] B --> D end subgraph 结果与代价 D --> E["极少数人阻断千亿美元级投资"] D -.-> F["司法判赔追责（如北达科他州判决）"] end ``` - **超级工厂建设速度**：马斯克在得克萨斯州仅耗时不到 18 个月即落成 Gigafactory。 - **环境评估周期**：纽约芯片制造厂在完成长达 2 年的环境评估后，仍因 6 名个体的诉讼而搁浅。 - **司法巨额赔偿**：北达科他州法官裁定绿色和平组织因阻挠输油管道需支付近 3.5 亿美元的损害赔偿金。 --- ### [52:13](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3133s) 国情咨文长达 108 分钟：两党民调与现场亮点 全美瞩目的国情咨文（State of the Union）演讲时长达到 108 分钟，创下自开始统计该数据 60 年以来的最长纪录。本次特朗普总统演讲的主题定为“250 年的美国：强大、繁荣与受人尊敬”（America at 250, strong, prosperous, and respected）。演讲中特朗普大篇幅展示了自己在通胀控制、就业增长以及边境管控等关键领域的政绩并进行“胜利巡礼”。然而，这场声势浩大的演讲背后，折射出的却是民调数据的剧烈波动与严峻挑战。 在民调走势方面，特朗普的净支持率从其开启本届任期伊始的 +11.7% 急剧下滑至 -14.3%，出现了高达 26 个百分点的剧烈震荡；经济支持率从 +3.4% 暴跌至 -18.2%；在贸易维度的支持率更是从 5.9% 骤降至 -22.7%。主持人和嘉宾们随即就这场长达 1 小时 45 分钟的演讲展开即兴点评与亮点回顾。 查马斯分享了他心目中最具冲击力的几幕：首先是众议员伊尔汗·奥马尔（Ilhan Omar）与拉希达·特莱布（Rashida Tlaib）在现场投出的愤怒“死亡凝视”（death stare）以及当众失控呐喊的场景，他评价这种举动在庄重场合显得非常不合时宜；其次是特朗普呼吁维护法律与秩序，要求将美国公民的利益置于非法移民之上时，全场民主党议员竟无一人起立鼓掌；此外，全场起立为美国冰球队鼓掌展现了难得的团结；最后则是好友布拉德·格斯特纳（Brad Gerstner）意外获得了总统的点名致敬，引爆了四人平时的私人讨论群聊。 - **演讲时长**：108 分钟，成为 60 年有记录以来最长的国情咨文。 - **支持率大跌**：特朗普综合净支持率下降 26 个百分点（从 +11.7% 跌至 -14.3%）。 - **经济与贸易民调**：经济评价滑落至 -18.2%，贸易评价暴跌至 -22.7%。 --- ### [54:20](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3260s) 优先保护本国公民：起立之争与政治极化交锋 现场随后回放了国情咨文中最具争议性的一段视频。特朗普在讲台上直接向全体国会议员喊话：“如果你认同这句话，请起立表明你的支持：美国政府的首要职责是保护美国公民，而不是非法移民。”（The first duty of the American government is to protect American citizens, not illegal aliens.）然而镜头扫过，民主党议员全员端坐不动、面无表情，场面极其尴尬。 四位主持人在演播室中对此展开了激烈交锋。主持人发问，优先保护本国公民是再清晰不过的常识底线，为何民主党人连站起来都做不到。杰森（JCal）回应称，即便自己可能对移民局（ICE）的某些激进手段持批评态度，但在支持美国公民权益和理性的移民改革上立场坚定。绝大多数国民都认同优先照顾本国公民并遣返暴力罪犯，同时也希望为长期居留者保留合法化路径。他认为民主党全员枯坐的根源在于当前体制陷入了极度政治极化，两党彻底丧失了跨党派合作的意愿，无论是关税议题还是移民治理，双方都不愿像克林顿或布什时代那样坐下来协商对话，他指责两党共同导致了当前政治机能的全面失灵。 萨克斯（David Sacks）则对此进行了坚决反驳，认为每当民主党人的极端主义立场暴露无遗时，人们总喜欢各打五十大板来和稀泥。萨克斯指出，特朗普这一提问实际上是一次极其基础的政治测试，甚至对特朗普本人而言还承担着政治风险，因为民主党人原本只需顺水推舟站起来鼓掌，就能向全美选民展示他们是讲求常识、秉持善意且愿意跨党派合作的理性温和派。然而，民主党议员们全盘拒绝。这种抵制不仅限于口号，更体现在对待具体受害者家属的冷酷态度上——当场上介绍被非法移民暴力杀害的无辜女性受害者家属（包括遇害女孩伊琳娜·扎鲁茨卡 Iryna Zarutska 的母亲）时，民主党阵营同样冷漠以对、拒绝鼓掌致意，将党派立场凌驾于基本的人性同情与国家安全之上。 ### [58:03](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3483s) 国情咨文中的党派极化与共识议题之争 大卫·萨克斯（David Sacks）继续剖析特朗普国情咨文演说的现场反响。他指出，民主党人在诸多显而易见的常识议题上均拒绝鼓掌，包括制止暴力罪犯、杀人犯与贩毒集团非法入境，以及谴责针对查理·柯克（Charlie Kirk）暗杀未遂事件后的政治暴力——甚至当柯克的妻子埃丽卡·柯克（Erica Kirk）出席现场时，民主党席位连礼貌性的掌声都没有给予。萨克斯认为，演说之所以反响极高（CNN 民调显示约三分之二的观众认为非常有效，CBS 观众中这一比例更高达四分之三），是因为特朗普列出了一连串支持率达到 80/20、90/10 乃至 95/5 的压倒性共识议题。民主党人一概拒绝鼓掌，并非仅是党派偏见，而是其已滑向激进与极端主义，以至于特朗普在台上近乎哀叹地评价道：“这些人疯了（These people are crazy）”。 杰森·卡拉卡尼斯（Jason Calacanis）则提出不同视角，认为这种对立不能单怪一方，特朗普长期以来也是“总分裂者（divider-in-chief）”，习惯于嘲弄和痛击对手，导致民主党人根本不想配合。杰森主张两党制衡与协作应回归理性与倾听，但也承认民主党在上届大选落败的根源正是脱离民众常识，比如边境管控本就是超高支持率的共识，卡玛拉·哈里斯（Kamala Harris）却未能果断承认并解决。萨克斯随即抓住杰森的让步，指出这正说明问题不在于表面的政治站队，而在于实际政策立场——特朗普站在绝大多数美国人一边，比如绝大多数选民对明尼苏达州索马里日托中心诈骗案深恶痛绝，而民主党议员伊尔汗·奥马尔（Ilhan Omar）却只会在场下大声抗议。 - **民意调查数据**：CNN 调查中有三分之二的受访观众认为特朗普的国情咨文演说非常成功，CBS 民调中的认可比例更达到近 75%。 - **核心争议案例**：包括严打非法移民重罪犯、打击明尼苏达州大规模日托骗保案、降低处方药价格以及反对政治暴力。 - **政治合作悖论**：杰森呼吁两党重新寻求妥协与制衡，但萨克斯强调当一方连 90% 赞同率的底线议题都不支持时，分歧实质是政纲理念的根本断层。 --- ### [01:02:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3720s) 佩洛西内幕交易梗与两党唯一的共识：债务膨胀 辩论过程中，杰森指出民主党并非全盘抵触，例如伊丽莎白·沃伦（Elizabeth Warren）就在特朗普呼吁禁止国会议员炒股时起立鼓掌，同时也在阻止伊朗获得核武器的议题上表达支持。特朗普当即在演说中抓住了这一互动，现场即兴调侃：“看，伊丽莎白·沃伦都站起来了，我真不敢相信南希·佩洛西竟然没站起来”，引发全场轰动。萨克斯盛赞这一时刻充分体现了特朗普不依赖提词器的现场抓梗与应变能力，同时也反驳了杰森关于纯粹党派极化的论调——沃伦之所以起立，是因为她真心赞成禁止议员内幕交易的法案，而她在边境议题上坐着不动，纯粹是因为政见不同。 随后，杰森向戴维·弗里德伯格（David Friedberg）提问：在充满戏剧表演的政治秀场背后，两党面对特朗普首年未能遏制且新增达 2.5 万亿美元的联邦财政赤字，究竟能否在核心要务上达成任何实际合作？弗里德伯格讽刺地给出了残酷的结论：华盛顿两党唯一能保持高度默契并起立鼓掌的事情，就是“继续烧钱，让我们背负更多债务（burn more capital and put us more in debt）”。 ```mermaid flowchart TD accTitle: 国情咨文两党分歧与共识议题图 accDescr: 展示国情咨文中争议议题、个别交叉议题与两党实际唯一共识的关系 subgraph 极化对立议题["80/20 绝大多数共识（民主党拒绝鼓掌）"] A1["边境重罪非法移民遣返"] A2["反对政治暴力 / 枪击受害者表彰"] A3["降低处方药价格与严惩暴力罪犯"] end subgraph 个别交叉点["偶发交叉议题"] B1["禁止国会议员内幕交易法案<br/>(沃伦起立，佩洛西坐地招致调侃)"] B2["遏制伊朗发展核武计划"] end subgraph 两党真正共识["弗里德伯格总结的两党唯一默契"] C1["无节制发债与扩大财政赤字<br/>(首年预估新增 2.5 万亿美元债务)"] end ``` --- ### [01:03:58](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3838s) 科学角：山中因子人体试验获批，逆转失明的抗衰曙光 在全场针对萨克斯离场去洗手间的插科打诨中，弗里德伯格正式开启“科学角”（Science Corner）。他带来了长寿与细胞重编程领域的重磅突破：诺贝尔奖得主山中伸弥（Shinya Yamanaka）发现的四种转录因子——“山中因子”（Yamanaka factors），不仅能重置哺乳动物细胞的表观基因组（epigenome）与生物钟状态，让细胞重回年轻活力，如今更迎来了首个人体临床里程碑。 哈佛大学著名抗衰学者大卫·辛克莱（David Sinclair）联合创立的生物技术公司 Life Biosciences，已与美国食品药品监督管理局（FDA）达成重大监管协议，将成为全球首家在人体上使用山中因子进行治疗的公司。尽管弗里德伯格坦承辛克莱在学界颇具争议（此前曾以 7.2 亿美元将宣称白藜芦醇能延缓衰老的公司售予葛兰素史克 GSK，最终该靶点未能成行，常被指责存在学术炒作），但这项临床试验的设计极为具体且令人振奋：研究人员计划将山中因子蛋白递送注射到患者眼球的玻璃体液中，直接作用于视网膜，旨在逆转因青光眼或眼部缺血性“中风样”病变所导致的失明。 - **科学机制**：山中因子能够擦除累积在 DNA 表面的有害表观遗传标记，将成熟体细胞的时钟拨回至年轻状态。 - **临床管线定位**：Life Biosciences 首批获准的人体试验聚焦于眼科适应症，通过向玻璃体腔内注射递送因子，修复受损的视网膜神经节细胞以恢复视力。 - **商业与学术背景**：联合创始人辛克莱虽因早年葛兰素史克收购案及保健品推广在学界背负争议，但此次推进到 FDA 批准的人体临床阶段，仍是整个细胞抗衰与再生医学领域的标志性时刻。 ### [01:06:35](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3995s) 科学角：山中因子人体试验与表观遗传重置 在讨论这项针对失明的临床一期试验时，弗里德伯格（David Friedberg）详细拆解了其革命性的技术路径。该试验的核心预期是将山中因子（Yamanaka factors）导入眼部，使视网膜细胞焕发生机、重返年轻态，进而恢复视力。如果这项试验取得成功——鉴于其在动物模型中展现的显著效果，这种成功极具确定性——它不仅将攻克特定失明难题，更将成为人类历史上首次将山中因子应用于逆转衰老的人体临床试验。在具体递送方式上，科研人员将合成这些蛋白质的 DNA 编码打包进腺相关病毒（AAV 病毒）中，注入眼球内部。病毒进入视网膜细胞后释放载荷，促使眼细胞自行表达这些蛋白质。 尤为惊艳的是，研究人员在其中设计了一套精密的“分子开关”控制机制。山中因子的表达并非持续失控进行，而是能通过患者口服一种名为多西环素（Doxycycline）的常见抗生素实现精准的开启与关闭。患者接受药物递送后，只需按需服用该抗生素，即可激活体内山中因子的表达，促使眼部细胞在理论上实现“逆生长”，进而修复视力。弗里德伯格指出，这项临床一期试验标志着人类首次将山中因子递送到人体内，但这仅仅是冰山一角。目前已有十多家初创公司正致力于通过递送山中因子或其他类似蛋白质，重置细胞的表观基因组（epigenome），开启新一轮生物医药革命，这甚至被不少人视作寻觅已久的“不老泉”。 当被问及这是否仅是一项毒理学评估机制试验时，弗里德伯格解释称，虽然出于安全考量初始剂量会控制在较低水平，但研究团队预期能观测到实际疗效。其底层生物学逻辑在于：衰老过程中遭受侵蚀的往往不是 DNA 序列本身，而是附着在 DNA 上负责控制基因开闭的表观遗传标记。随着年龄增长，这些标记发生漂移并错位；而当山中因子这类重编程蛋白进入细胞后，能驱动整个系统将错位的化学标记重新归位，让细胞按年轻时的程序恢复运转。 ```mermaid flowchart TD accTitle: 山中因子眼部抗衰机制 accDescr: 展示从 AAV 病毒递送、多西环素分子开关到表观基因组重置与视力恢复的完整生物机制流程。 subgraph Delivery["递送与激活"] A["AAV 病毒载体<br/>(携带山中因子 DNA)"] --> B["注射进入眼部视网膜细胞"] C["口服多西环素<br/>(Doxycycline 开关)"] --> D["触发山中因子蛋白表达"] B -. 待命.-> D end subgraph Epigenetics["表观重编程"] D --> E["复位 DNA 表层修饰标记"] E --> F["修复受损表观基因组"] end subgraph Outcome["临床效果"] F --> G["视网膜细胞重返年轻状态"] G --> H["恢复视力 / 逆转局部器官衰老"] end ``` - **分子开关调控**：借助口服抗生素多西环素，医生可以随心所欲地控制体内重编程蛋白的合成节奏，大幅规避细胞过度去分化所引发的潜在致癌风险。 - **潜在应用外推**：除了视网膜修复，灵长类动物实验显示该技术经皮递送后能显著抚平皱纹。未来该技术极有希望进一步扩展至关节炎治疗、软骨磨损修复以及膝关节年轻化。 - **历史跨度**：山中伸弥团队于 2006 年首次发现诱导多能干细胞相关因子，历经近二十年探索，这项科学突破终于迈入真正的人体临床阶段。 现场随即爆发了一阵幽默互动。杰森（Jason Calacanis）打趣道，这个播客节目眼下最迫切需要的其实是“发际线逆龄重置”。萨克斯（David Sacks）立刻反驳：“你替你自己说吧！我今年 50 岁，我的发际线简直完美无瑕，你可别乱套用。”杰森则笑着调侃道：“得了吧兄弟，你那额角明明已经开始冒出小尖尖了，不过以 50 岁的标准来说确实维持得不错。” --- ### [01:10:16](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4216s) 最高法院裁定关税越权，特朗普启动紧急关税预案 话题随后切入震撼全美的重大司法与经济裁决。最高法院以 6 票对 3 票的裁定，正式否决了特朗普总统依据紧急状态权力所推行的关税政策。在由彭博社报道的这起裁决中，三位保守派法官（罗伯茨、巴雷特、戈萨奇）罕见地与三位自由派法官站在了一起。这是自 1935 年最高法院推翻富兰克林·罗斯福总统“第一新政”（First New Deal）立法以来，长达 91 年历史中最高法院对现行行政权力的最严厉驳斥。 宾夕法尼亚大学沃顿商学院的分析数据显示，截至裁决生效时，这批紧急关税累计已征收约 1750 亿美元，其中高达 50% 的税款可能面临退款风险。目前已有两千多家进口商向法院正式提交退税申请，相关司法清算预计将经历漫长的诉讼程序。然而，特朗普立即宣布自己绝不受挫，随即依据《1974 年贸易法》第 122 条款（Section 122），重新针对全球商品祭出 15% 的全方位临时关税。预测市场 Polymarket 上的动态随即剧烈分化：最高法院裁决出台后，法院强制特朗普政府退税的概率从 18% 飙升至 40%；而对于国会能否在 3 月 31 日前自行立法通过任何关税法案，市场押注的概率仅有可怜的 3%。 --- ### [01:11:52](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4312s) 萨克斯与查马斯复盘关税大棋：替代性法律武器与制造业回流 萨克斯对此发表了极具深度的法律与地缘经济研判，直言“关税绝不会就此烟消云散”。他指出，大法官卡瓦诺（Brett Kavanaugh）长达 70 页的反对意见书实际上为白宫后续的关税推行提供了一份详尽的“行动路线图”。现行法律框架内本就包含大量可替代的法定依据： - **《1974 年贸易法》第 122 条款**：允许行政部门为纠正国际收支失衡，征收为期最长 150 天、最高 15% 的临时关税，白宫目前已迅速启动此条款。 - **《1974 年贸易法》第 301 条款**：赋予总统对外国不公平贸易行为采取报复性关税的法定权力。 - **《1930 年关税法》第 338 条款**：专门针对歧视美国商业贸易的国家设立惩罚性关税。 萨克斯强调，150 天的临时关税期正是为政府争取时间，通过各部门审查与专项研究夯实证据链，进而顺利转入第 301 或第 338 条款的长期关税通道。最高法院多数派意见书与协同意见书中均未要求行政部门退还已征收的税款，这表明法官们心知肚明关税政策仍会延续。“美国背负着数万亿美元的国债，凭什么要把上千亿美元退给一群进口商？”萨克斯直言，此前关于关税会导致经济崩溃的末日预言已被事实全部击碎，无论未来是共和党还是民主党执政，关税都将演化为一项获得跨党派民意支持的长期基石政策。 查马斯（Chamath Palihapitiya）对此表示高度认同，断言这场针对全球贸易格局的“压力测试实验”已经取得了决定性胜利。这项政策戳破了所谓的全球主义神话，彻底暴露了长期以来美国与主要贸易伙伴之间不合理且缺乏经济理性的结构性失衡。正是过去几十年的错误政策导致了美国制造业被掏空、中产阶级被边缘化，而关税壁垒最终将为美国工人群体重新赢回公平与尊严。他呼吁国会应当顺应大势，将这些关税措施以立法形式固定为长效制度；即使国会陷入僵局，总统手中依然握有充足的行政工具，因为关税对于重塑美国经济基本盘而言是无比明智的选择。面对弗里德伯格对“最高法院打破党派界限判决”的提问，对话转向了关于司法独立与制衡原则的深入探讨。 ### [01:15:50](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4550s) 最高法院裁决捍卫分权制衡：司法独立与行政权力的边界 弗里德伯格（David Friedberg）指出，外界许多人一直认为联邦最高法院具有极高的党派偏见，但本次关税裁决却给全美民众注入了一剂强心针。尽管最高法院目前由保守派法官占据多数——其中不少人还是特朗普亲自提名的政治盟友，法院依然做出了完全违背总统意愿的裁决。这恰恰向所有人证明了建国先贤们设立的宪政体系运转良好：当行政部门的行动超越宪法与法律授权时，司法部门确实有能力予以纠偏。弗里德伯格强调，抛开关税的经济后果、国家安全影响以及贸易结构等具体政策争论不谈，最高法院在法律审理上表现出的非党派性独立姿态，本身就极具制度价值与历史意义。 ```mermaid flowchart TD accTitle: 三权分权制衡机制 accDescr: 展示行政权力受司法裁决纠偏以及国会应当协同立法的权力制衡流程 subgraph ExecutiveBranch["行政分支（白宫）"] A["单边行政权力扩张<br/>(如特朗普关税 / 拜登学生贷款赦免)"] end subgraph JudicialBranch["司法分支（最高法院）"] B["司法独立审查与纠偏<br/>判定行政越权违宪/违法"] end subgraph LegislativeBranch["立法分支（国会）"] C["重掌宪法税收与贸易授权<br/>追认存量关税并设定2026年清晰参数"] end A -->|"遭遇司法诉讼"| B B -->|"迫使权力回归宪政正轨"| C C -.->|"提供合法工具与明确边界"| A ``` 卡拉卡尼斯（Jason Calacanis）对弗里德伯格的观点表示完全赞同，并呼吁公众高度警惕行政权力的单边膨胀。无论是拜登试图单方面赦免学生贷款，还是特朗普单方面大举征收关税，都暴露出行政部门试图碾压其他权力分支的危险倾向。卡拉卡尼斯打趣道，这种分权制衡虽然常常导致华盛顿陷入政治僵局，让人感到沮丧，但正如他们四个合伙人每周四准时录制播客一样，“必须学会在吵架中协作共事”。他直言特朗普在推行关税时手段过于混乱，让全美企业主陷入无所适从的恐慌之中。国会理应主动承担宪法责任，与白宫并肩坐下来谈判，明确贸易失衡的痛点，给总统提供合法工具，而不是任由行政权肆意乱撞。 - **三权分权实证**：保守派占多数的最高法院裁定特朗普越权，证明了司法独立并非虚设。 - **关于关税退税与立法的妥协方案**：卡拉卡尼斯建议国会应当明确表态，立法追认特朗普此前实施的各项关税以避免陷入强制退款（refunds）的财政混乱，同时为2026年及未来的关税政策划定明确的法定参数与协商机制。 --- ### [01:18:38](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4718s) “政治战法”之争与跨党派合作的呼吁（节目收尾） 萨克斯（David Sacks）随即向卡拉卡尼斯提出尖锐质疑，打破了他描绘的“围坐篝火、和睦共处（Kumbaya）”的温情愿景。萨克斯援引了苏珊·赖斯（Susan Rice）近期的公开抨击言论，指出赖斯不仅威胁要对共和党政客实施清算，甚至声称只要与现任政府合作过的科技公司未来都将面临司法起诉。在萨克斯看来，民主党人已经在毫不掩饰地宣告，一旦重掌大权，他们将立即启动激进的“法律战（lawfare）”进行报复。 卡拉卡尼斯反驳称，法律战绝对是两党共同犯下的罪孽——特朗普重新上台后同样对前联邦调查局局长詹姆斯·科米（James Comey）和美联储主席杰罗姆·鲍威尔（Jerome Powell）展开政治追查。他大声疾呼，两党必须彻底停止这种恶性循环的司法清算，取消荒谬的总统特赦特权，并在华盛顿特区重建久违的“团队合作精神与集体凝聚力（esprit de corps）”。卡拉卡尼斯强调，中期选举与2028年大选中，选民应当将选票投给愿意跨党派合作的温和派候选人，唯有摆脱极化混乱，美国社会才能真正恢复稳定。 在节目的最后，卡拉卡尼斯以标志性的幽默语调向戴维·萨克斯、戴维·弗里德伯格和查马斯·帕里哈皮蒂亚（Chamath Palihapitiya）三位好友表达了感谢，自封“世界最强主持人 JCal”并向听众致意道别。随后，节目在经典的标志性彩蛋音频合辑（包括“让赢家继续狂奔”、粉丝共创梗以及关于团队“性张力”的荒诞玩笑）中画上句号。 - **两党法律战风险**：苏珊·赖斯针对科技企业与保守派的言论，与特朗普针对鲍威尔、科米的施压，展现了司法工具武器化的恶性循环。 - **政治诉求**：呼吁在即将到来的中期选举与2028年大选中支持中间温和派，遏制白宫行政扩权与党派清算。 ## Notable quotes > [10:26](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=626s) 这并不是说我认为这项技术是无聊的把戏，而是因为其不确定性如此之高，关于 AI 宏观经济效应的真实、实时数据又如此匮乏，以至于那些极其严肃的 AI 讨论往往更偏向文学创作，而非真正的分析。 — Derek Thompson (由 David Sacks 引用) > [10:26](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=626s) 如果买政府国债就能旱涝保收拿到 5% 的无风险收益，投资人凭什么要在这里承担过度的不确定性风险？ — Chamath Palihapitiya > [22:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1339s) 不，我们绝对不会再招新人。 — Jason Calacanis > [22:19](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1339s) 这是我自万维网诞生以来，在互联网上度过的最激动人心的时刻！ — Jason Calacanis > [30:40](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=1840s) 经济从来都不是一块固定大小的馅饼，它是一座花园，而技术就是灌溉它的雨水。 — David Sacks > [54:20](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3260s) 如果你认同这句话，请起立表明你的支持：美国政府的首要职责是保护美国公民，而不是非法移民。 — Donald Trump > [58:03](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3483s) 这些人疯了（These people are crazy） — Donald Trump (由 David Sacks 引用) > [01:02:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3720s) 看，伊丽莎白·沃伦都站起来了，我真不敢相信南希·佩洛西竟然没站起来 — Donald Trump > [01:02:00](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=3720s) 继续烧钱，让我们背负更多债务（burn more capital and put us more in debt） — David Friedberg > [01:11:52](https://www.youtube.com/watch?v=kzWbCF_IkHY&t=4312s) 美国背负着数万亿美元的国债，凭什么要把上千亿美元退给一群进口商？ — David Sacks ## Mentioned Alex Jones, Anthropic, Dario Amodei, Pete Hegseth, Thomson Reuters, RELX, LexisNexis, LegalZoom, CrowdStrike, Cloudflare, Okta, IBM, COBOL, Citrini Research, American Express, Capital One, Mastercard, Visa, DoorDash, Derek Thompson, Ezra Klein, Salesforce, Citadel Securities, Box, Aaron Levie, Elon Musk, Launch, Waymo, OpenClaw, Pipedrive, Notion, Slack, MrBeast, Instacart, Mattermost, HubSpot, Flexport, Ryan Petersen, Harvey, Kimi 2.5, NVIDIA, Sarah Friar, OpenAI, Apple, Mac Studio, Dan Loeb, Third Point, Bill Gurley, Bernie Sanders, Micron Technology, Greenpeace, Ilhan Omar, Rashida Tlaib, Brad Gerstner, Iryna Zarutska, Charlie Kirk, Erica Kirk, Kamala Harris, Elizabeth Warren, Nancy Pelosi, Shinya Yamanaka, David Sinclair, Life Biosciences, FDA, GlaxoSmithKline, John Roberts, Amy Coney Barrett, Neil Gorsuch, Brett Kavanaugh, Franklin D. Roosevelt, Polymarket, Susan Rice, James Comey, Jerome Powell, Jason Calacanis, David Sacks, Chamath Palihapitiya, David Friedberg ## Video description (0:00) Bestie intros (1:22) Claude's hit list, SaaS crash, and Citrini's AI letter (30:39) Why Doomer narratives are more popular, valuable new AI jobs (40:19) Understanding the Rate Payer Protection Pledge, what's behind datacenter opposition? (52:13) State of the Union reactions (1:03:58) Science Corner: Cure for blindness via Yamanaka Factors? (1:10:17) SCOTUS strikes down tariffs, Trump pivots Apply for Liquidity: https://allinliquidity.com Follow the besties: https://x.com/chamath https://x.com/Jason https://x.com/DavidSacks https://x.com/friedberg Follow on X: https://x.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@theallinpod Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://x.com/yung_spielburg Intro Video Credit: https://x.com/TheZachEffect Referenced in the show: https://www.investing.com/news/stock-market-news/wolters-kluwer-relx-shares-slip-after-anthropic-unveils-aienhanced-legal-tool-4481124 https://www.barrons.com/articles/ibm-stock-had-worst-day-in-25-years-ai-disruption-fears-5f632d6c https://www.forbes.com/sites/daniellechemtob/2026/02/24/forbes-daily-ibm-suffers-its-worst-day-since-the-dot-com-bubble https://x.com/chamath/status/2027077786503164260 https://www.citriniresearch.com/p/2028gic https://thedefiant.io/news/tradfi-and-fintech/credit-card-stocks-fall-after-citrini-ai-report https://x.com/TurnerNovak/status/2026332990914101699 https://x.com/anistotle_/status/2026306126674108788 https://www.notyourtypicalfinancebro.com/p/vibe-laundering-pt-2-citrini-the https://www.goodreads.com/quotes/457097-nobody-knows-anything-not-one-person-in-the-entire-motion https://www.derekthompson.org/p/nobody-knows-anything https://x.com/kalshi/status/2027040345419129166 https://x.com/StockMarketNerd/status/2019837212515528730 https://www.citadelsecurities.com/news-and-insights/2026-global-intelligence-crisis/ https://x.com/DavidSacks/status/2027087693327237251 https://x.com/levie/status/2026885050411745491 https://x.com/typesfast/status/2026998028222013679 https://x.com/cboyack/status/2021647373571862952 https://x.com/chamath/status/2025369318696124859 https://x.com/pat_hedger/status/2026742424471560636 https://x.com/SemiAnalysis_/status/2026719180284666046 https://x.com/WesternLensman/status/2024661247296172486 https://www.bloomberg.com/opinion/articles/2026-02-20/supreme-court-s-tariffs-ruling-finally-holds-trump-accountable https://polymarket.com/event/will-the-court-force-trump-to-refund-tariffs-2026-06-30 https://polymarket.com/event/will-congress-pass-any-tariffs-by-march-31 #allin #tech #news [Source](https://www.youtube.com/watch?v=kzWbCF_IkHY)

[10:26] Derek Thompson (由 David Sacks 引用): 这并不是说我认为这项技术是无聊的把戏，而是因为其不确定性如此之高，关于 AI 宏观经济效应的真实、实时数据又如此匮乏，以至于那些极其严肃的 AI 讨论往往更偏向文学创作，而非真正的分析。 [10:26] Chamath Palihapitiya: 如果买政府国债就能旱涝保收拿到 5% 的无风险收益，投资人凭什么要在这里承担过度的不确定性风险？ [22:19] Jason Calacanis: 不，我们绝对不会再招新人。 [22:19] Jason Calacanis: 这是我自万维网诞生以来，在互联网上度过的最激动人心的时刻！ [30:40] David Sacks: 经济从来都不是一块固定大小的馅饼，它是一座花园，而技术就是灌溉它的雨水。 [54:20] Donald Trump: 如果你认同这句话，请起立表明你的支持：美国政府的首要职责是保护美国公民，而不是非法移民。 [58:03] Donald Trump (由 David Sacks 引用): 这些人疯了（These people are crazy） [62:00] Donald Trump: 看，伊丽莎白·沃伦都站起来了，我真不敢相信南希·佩洛西竟然没站起来 [62:00] David Friedberg: 继续烧钱，让我们背负更多债务（burn more capital and put us more in debt） [71:52] David Sacks: 美国背负着数万亿美元的国债，凭什么要把上千亿美元退给一群进口商？

[20:06] David Sacks: When you lower the cost of something that was previously supply constrained, demand for that thing goes up... the market for software and software engineers was so constrained by the lack of availability that even if we 10x or 100x the productivity of software engineers, the demand will be there to absorb this new supply. [30:39] David Sacks: The economy is not a pie, it's a garden and technology is rain. All of this technological innovation is going to increase the growth rate of the garden. It's not a fixed pie. [32:18] Chamath Palihapitiya: When we think about fully burdened cost of employees, we now factor that in because we're at a place where some of our engineers are just racking up ginormous bills... [35:42] Chamath Palihapitiya: Last year we lost almost 5 gigawatts in terms of cancelled projects... You're talking about 130 billion of lost revenue over these two years that'll go forward in time that we miss out on. [40:20] David Sacks: The president's approach finds a very good balance here, which is look, we can have progress. Just don't make residential consumers pay for it. Let the big tech companies pay for it themselves. [42:47] David Friedberg: All the data centers in the world fit under the tip of a pin... If we're going to give up hundreds of thousands of jobs and many billions of dollars of economic value creation, we're being pretty silly and pretty obtuse in our view of the world. [52:13] Donald Trump: The first duty of the American government is to protect American citizens, not illegal aliens. [50:17] David Sacks: I don't think that the tariffs are going away... Why would we want to give back hundreds of billions of dollars to a bunch of importers when we're trillions of dollars in debt?

[0:00] All right, everybody. Welcome back to [0:02] your favorite podcast, the AllIn [0:04] podcast. Today we have a conspiracy [0:08] corner episode for you. We're going to [0:10] go over the 9/11 inside job. We're going [0:12] over flat earth, JFK assassination. It's [0:16] going to be all conspiracy all the time [0:18] after our amazing blockbuster episode [0:21] during Ski Week. We're going all [0:22] conspiracy here. Our guest today, Alex [0:24] Jones. [0:25] >> How many views did it get? Nine views. I [0:28] mean, it's tough when you have one out [0:30] of four besties. It doesn't Michael [0:32] Tracy is on standby. [0:34] >> Not true. I can carry an episode for at [0:36] least 400,000 views. [0:37] >> I mean, you might. You might. I like [0:39] your Hey, for people who don't know, [0:40] Jimoth has his own YouTube channel. He's [0:42] got his escape hatch for when this train [0:44] wreck burns to the ground. He started [0:46] his own YouTube channel. He's hedging [0:48] his bets. Freyberg's working on his solo [0:50] project. Everybody's doing solo project. [0:52] The band's got a lot of solo projects. [0:54] >> The Beatles are experimenting. The [0:55] Beatles. [0:55] >> They're experimenting. We had a little [0:57] Yokoono situation going on here. You [0:59] know what the number one topic for this [1:00] show was by the all-in AI bot sacks. [1:04] >> What's up? [1:05] >> The number one was Daario versus Hegs, [1:08] the uh Department of War versus [1:10] anthropic was the number one topic [1:12] selected by our AI bot as a programming [1:14] note for folks. That decision will be [1:16] made end of the day Friday when this [1:17] podcast comes out. So, we will talk [1:18] about it next week. But, let's get to [1:20] work. We've got a full docket. The [1:22] clawed kill list has expanded and an AI [1:25] fanfiction substack tanked your 401k on [1:28] Monday. Let's get into it. Anthropics [1:31] generational run continues. They're now [1:33] three for three in tanking different [1:36] market sectors in February. [1:38] Congratulations. This is like they they [1:41] took the the mantle from Brad Gersonner [1:43] uh tanking the market. [1:44] >> The Enthropic list. [1:46] >> It is February 3rd. Anthropic announces, [1:48] hey, we got a legal plugin for Claude [1:50] Co-work, Thompson Reuters, Lexus, Nexus, [1:53] Legal Zoom, all down at least 10% since [1:56] February 3rd. Then on February 20th, [1:58] Claude Code Security is announced in a [2:01] limited research preview. Stocks tank [2:03] again. Crowd Strike, Cloud for Octa, all [2:06] down. Then February 23rd, Anthropic [2:10] announces Claude can modernize Cobalt [2:12] databases. If you don't know Cobalt, [2:14] that's the like oldest coding language [2:16] in the world. That's where Sachs learned [2:18] code when he was in college in the 70s. [2:19] It's used for banking, payroll, [2:21] government, [2:22] >> healthare. [2:23] >> Healthcare runs 95% of ATMs in the US [2:27] and it powers social security payments. [2:29] 85% of all coal code runs on IBM [2:32] machines. So IBM decided they would tank [2:34] 13% on Monday, their worst day since [2:38] 2000. 31 billion in market cap losses. [2:43] So let's stop here before I get into the [2:45] fanfiction piece. [2:47] What's your take here of what's [2:49] happening in the market, Shmath? Is this [2:52] simply people are looking for an excuse [2:54] to trim their positions because things [2:55] have been top ticking all-time highs and [2:57] people are just looking for an excuse or [2:59] is this reality? Is this the go forward [3:01] reality that AI is going to compress [3:04] these kind of stocks because it solves a [3:06] lot of problems? [3:08] >> I'm going to give you two explanations. [3:11] I don't know what percentage I would [3:14] allocate [3:15] across the two, but I think one is [3:18] tactical and one is much more strategic, [3:20] but I think both are happening. The [3:23] tactical one is that we are at a moment [3:25] in time where a lot of the smart money [3:29] hedge funds are starting to massively [3:32] degross. And what that means is they're [3:36] trimming a lot of positions and they're [3:38] just taking on a lot less risk. Why? I [3:42] don't exactly know. It could be [3:44] motivated by the second thing that I'm [3:45] going to talk about. But the point is in [3:48] a deg [3:51] risk and making your position sizes much [3:54] smaller. So the longs become less long, [3:56] the shorts become less short and you [3:58] just shrink. And so there's just general [4:00] downward pressure. That is a clear [4:02] behavior right now. But I think the [4:05] structural change is the more important [4:07] one. And this is sort of what I talked [4:09] about this morning. In a normal [4:12] functioning market, [4:14] what we are always debating is when a [4:18] set of cash flows go from becoming [4:21] highly confident to less highly [4:24] confident. It's a when conversation. So [4:27] when will Coca-Cola's cash flows be [4:30] impacted? When will Eli Lily's cash [4:33] flows be impacted? When will Meta's cash [4:36] flows be impacted? And the answer to the [4:38] when [4:40] gets translated by the public markets [4:42] into three things. Your price to [4:44] earnings multiple where if you invert [4:47] that number what that is equivalent to [4:50] is the yield on the money that you get. [4:52] Okay. So if you you know 20 times PE [4:55] that's a 5% yield. [4:58] The second is a revenue multiple. And [5:00] the third is what's called your weighted [5:03] average cost of capital. Which is to [5:05] say, if you look at the next 20 to 30 [5:08] years of earnings and you want to figure [5:10] out what that is worth today, [5:13] you have to discount all of these back [5:15] and you have to assume [5:18] a percentage of interest effectively [5:21] that it takes to get there. And the [5:23] basic math of this is that when you have [5:25] a high whack, it's called, you're [5:27] massively discounting these cash flows. [5:29] When you have a low whack, you're [5:31] assuming that these things are very [5:32] durable. Okay. So, what is happening? [5:36] We used to debate when this is no longer [5:38] a when moment. The market is very much [5:41] in an if mode. [5:43] Are these cash flows durable at all? [5:46] >> Could they fall off a cliff in year [5:48] three? Is there some AI model that's [5:50] going to come around the corner and [5:52] obliterate this business without me [5:54] knowing it? And because they've shifted [5:57] into this if mindset, [6:00] your risk becomes totally different. You [6:03] have this event risk that you don't know [6:06] how to price. And whenever the market [6:08] shifts into that mode, what you see are [6:12] that the holders of those equities want [6:14] a massive margin of safety. What does [6:16] that mean? They have to take pees way [6:19] down. If you used to trade at 40, you [6:21] should trade at 20. If you used to trade [6:23] at 20, you should trade at 10. [6:25] They take revenue multiples down. You [6:27] used to trade at 10 times revenue. Now [6:29] you're going to trade it three times. [6:31] You take the whack way up. Used to be a [6:33] 6% discounted weighted average cost of [6:35] capital. You know what? I'm taking you [6:37] to 12 or 13. That's the market's way of [6:40] saying, I'm now debating [6:43] if these things will even exist and so I [6:47] need to give myself a huge buffer to own [6:49] this stuff. [6:50] >> That's what's happening right now. It [6:53] has a lot of ripple effects that we can [6:55] talk about. Freeberg and I have talked [6:56] about this a lot. The most obvious [6:58] impact is how these tech companies [7:00] recruit and retain talent because the [7:02] biggest thing that it starts to eat into [7:05] are the cash flows of a business which [7:07] really directly tied to stockbased comp [7:09] and all this other stuff. But let me [7:10] just stop there. So we are we have moved [7:12] away from a when [7:14] >> to now an if and I think that that is a [7:17] very smart question to be asking. The [7:19] answer may be from any of these [7:21] companies that they will survive, but we [7:23] don't know how long. And until that [7:24] becomes clearer, you have to give [7:26] yourself room to be wrong. [7:28] >> You said when, then if. Did you mean if? [7:30] >> No, no, no. [7:31] >> To when? [7:32] >> We we've always debated when. When will [7:34] these cash flows disappear? Now it's [7:35] like [7:36] >> will they even exist? [7:37] >> Got it. Okay. So the second part of the [7:40] story Friedberg and Sachs is that a [7:43] Substack post fanfiction [7:46] uh taking place in the fictional 2028 uh [7:51] global intelligence crisis went mega [7:54] viral. 28 million views on X. It was [7:56] posted Sunday night. It made the market [7:58] tank on Monday. In this fictional [8:02] Substack post, the author said there's [8:05] going to be essentially a death spiral [8:07] that happens because of AI. How does [8:09] that work? Well, first, companies [8:11] embrace AI. Everything goes right. [8:13] They're able to cut staff. Their margins [8:15] go up, similar to how Amazon has, you [8:18] know, trimmed their white collar staff. [8:21] Then they're so successful at this that [8:23] they lose their customer base because [8:25] consumers don't have discretionary [8:26] funding to spend. Then it creates a [8:29] death spiral where the companies keep [8:30] deploying AI to try to hit the margins, [8:32] cutting staff, and the entire economy [8:34] collapses. Dr. Doom level stuff. [8:36] Unemployment's at 10%, S&P goes down [8:38] from 38% highs. After this piece came [8:41] out which speculated that agents would [8:43] get rid of all the 3% interchange fees [8:46] and move everybody to settle [8:47] transactions on stable coins. All the [8:49] financial stocks got hit on Monday. Amex [8:51] down 8%, Capital 1 down 8%, Mastercard [8:54] 6%, Visa 4%, yada yada. Finally, this [8:58] piece got a lot of push back. There was [9:00] a silly piece in it or a section in it [9:02] where they said AI agents would vibe [9:06] code their way to displacing Door Dash. [9:09] And uh that's kind of silly if anybody's [9:11] run a networkbased business knows. Sax, [9:14] I assume you read this piece or at least [9:17] saw the fallout from it. What's your [9:19] take? And then we'll go to you Freeberg. [9:21] >> Yeah. Well, I I know that this uh [9:23] Catrini article got passed around like a [9:25] join at a Grateful Dead concert, but I'm [9:29] starting to question how legitimately [9:31] viral it really was. There's some [9:33] information that just came out that the [9:36] attribution of the article has been [9:38] amended, meaning the co-authors have [9:39] been amended to include a short fund [9:43] that was shorting some of the names [9:45] mentioned in the article. This is [9:47] according to another post that just came [9:49] out. According to this post, the [9:51] authorship attribution attributed to [9:55] market moving was changed after [9:57] publication of the co-author is the [9:59] managing partner of a $262 million SEC [10:02] registered hedge fund who confirmed [10:04] short positions in the companies the [10:06] report named. So I think that's point [10:08] number one is I just wonder did this [10:11] article truly go viral or did the [10:13] authors do anything to kind of amplify [10:15] it and we just don't know the answer to [10:17] that question. But regardless of that, [10:19] let's just take the arguments on their [10:20] face. I think one of the best responses [10:23] to it was by another writer named Derek [10:26] Thompson who wrote a article called [10:28] Nobody Knows Anything, which I think is [10:30] a reference to a famous take by [10:33] legendary Hollywood writer William [10:35] Goldman. [10:36] >> Yeah. [10:36] >> In any event, what the article says is [10:39] no one really knows what's going to [10:40] happen with AI in two years, never mind [10:42] 20 years. And so they resort to science [10:45] fiction writing masquerading as [10:47] analysis. And the author here, Derek [10:50] Thompson, says that the conversation [10:52] about AI is really just a marketplace of [10:55] competing science fiction narratives. [10:57] And he says, "That's not to say I think [10:58] the technology is a parlor trick, but [11:00] rather that the level of uncertainty is [11:02] so high and the quality and supply of [11:05] real world real-time information about [11:07] AI's macroeconomic effects so poultry [11:10] that very serious conversations about AI [11:12] are often more literary than genuinely [11:15] analytical." So in other words, what [11:17] he's saying is, look, this guy is [11:18] writing very compelling science fiction, [11:21] but there's no real analytics behind it [11:25] to defend it. And yes, this could [11:26] happen. Here's a prediction market on [11:29] whether people believe the Catrini [11:30] report's going to come true. Something [11:32] like 12% believe the Catrini scenario is [11:35] going to happen. But the truth is, no [11:36] one really knows. I mean, there's other [11:38] dueling science fiction narratives where [11:40] AI is going to create such a world of [11:42] abundance that we're not going to need [11:45] for anything. And just by the way, Derek [11:47] Thompson is one of the abundance guys [11:49] with Ezra Klein. [11:51] >> This is why the market's getting [11:52] whacked. I think that you're right, [11:53] Saxs, nobody knows. So, if you can get [11:56] 5% for owning government bonds, why are [11:58] we taking excessive risk here? [12:00] >> Yeah. Let me just build on your point [12:01] about SAS. So the reason why there's so [12:03] much uncertainty around SAS is that SAS [12:06] used to be such a easily modeled and [12:10] predictable category and you know I saw [12:13] as a VC we saw the same story play out [12:15] across many many different categories of [12:17] software. You'd have this initial period [12:19] where there'd be this experimentation [12:21] phase you have a bunch of different [12:22] products that come to market. There'd be [12:24] a battle and then the market would [12:25] eventually settle and they'd be a [12:27] category leader and they would capture [12:29] most of the market share and the vast [12:31] majority of the market capitalization [12:33] and they would have very very [12:34] predictable metrics. It was very easy to [12:37] grade a SAS business. You look at ARR, [12:39] you say annually recurring revenue. You [12:41] look at the net dollar retention you [12:43] want to see depending on the phase [12:45] >> RPO RPO revenue under performance. [12:48] >> And so you know these things began to be [12:50] seen as like a annuity with growth right [12:52] because [12:53] >> they were rock solid. Yeah. [12:54] >> Yeah. Because a good net dollar [12:57] retention would be something like 120%. [12:59] Which means that your sort of cohort of [13:01] existing customers on balance would all [13:03] renew the next year and actually they [13:05] would renew at 120% of the previous [13:08] year's contract values. And the reason [13:10] you got that extra 20% is they would buy [13:12] more seats or there'd be additional [13:15] products or features they would upsell. [13:17] It got to be very very predictable. And [13:19] so when people were buying software [13:21] companies at I don't know 13 times ARR, [13:24] they thought they were buying a growth [13:25] annuity. And now all of a sudden you got [13:28] to factor into that. Well, wait a [13:29] second. What if AI disrupts the whole [13:33] market? What if it doesn't eliminate? I [13:35] don't think AI is going to get rid of [13:36] Salesforce, but it could eat into their [13:38] growth opportunity. We just don't know [13:40] what if it changes the pricing model. I [13:42] mean, it just creates a whole lot of [13:43] unknowns. And I I actually don't believe [13:45] in the Catrini or or the doomer take on [13:48] this, but I can see why the market would [13:52] feel this level of uncertainty [13:54] >> 100%. [13:55] >> Given how predictable a category SAS [13:59] used to be just say a year ago. [14:02] >> Yeah. Well, chaos is a ladder, Freeberg, [14:05] and this means opportunity. So, if we [14:08] look at this and SAS has headwinds, then [14:11] is there a winner? Is open source the [14:13] winner? Or is this all deflationary in [14:14] your mind, Freeberg? And we just make [14:16] less money and the earnings of these [14:19] companies get compressed, the size of [14:21] them gets compressed. How do you think [14:22] about it? I think fundamentally if [14:25] you're driving productivity with AI, [14:29] you're driving leverage on human time [14:32] and leverage on capital. [14:34] The question is how quickly can you [14:36] drive that up? And that's a function of [14:38] how much consumption there is, how much [14:40] capacity there is for consumption. So on [14:43] the one hand, I'll just speak broadly. I [14:46] think like humans have this desire to [14:48] improve their livelihoods by roughly 10% [14:50] every year. Meaning like your income and [14:53] your ability to purchase stuff that's [14:54] new relative to where you were last year [14:56] has to go up by 10% for you to be happy. [14:58] If it's less than 10%, you're probably [14:59] unhappy. [15:00] >> Is that that's your anecdote or that's [15:01] like [15:02] >> that's just like an anecdote. Like I [15:03] think I think that's sort of like my [15:05] rubric for thinking about like why are [15:07] people unhappy? You're happy. So if your [15:08] earnings are the same but things are [15:09] getting more expensive, you're not [15:11] happy. If your earnings go up by 10% and [15:13] things stay the same price, you got 10% [15:16] more than you had last year. You're [15:17] you're going to be happy. I just think [15:18] like all humans are driven by this need [15:19] to consume more each year than they did [15:22] last year. So I think for me that's like [15:25] the lower limit on consumptive capacity [15:28] in the world. [15:31] The question that we're now facing which [15:33] we've never faced in human history [15:34] before is there a upper limit [15:38] >> on consumptive capacity because AI [15:41] creates such a profound shift in [15:44] productivity and in leverage that [15:47] normally you would say hey when we get a [15:49] new tool or we get new leverage in a [15:50] system we build a new technology we can [15:52] make more with less. Therefore, everyone [15:55] gets access to more things for the same [15:57] price or the cost of things that they [15:59] consume come down by a certain price. [16:01] But there may be a situation now where [16:04] the ability to make stuff exceeds the [16:08] capacity to consume stuff. And that is [16:11] something that I don't think we've faced [16:12] before. And I think that's sort of where [16:14] a lot of the models start to break. Just [16:16] general economic models, just general [16:18] productivity models, and general social [16:20] models. And this goes to the point about [16:23] like what is everyone going to do? In [16:25] the same way that I think we've argued [16:26] that maybe SAS was a transitory business [16:29] phenomenon that existed between the [16:32] foundation of the internet and the era [16:35] of AI. It may be the case that knowledge [16:38] work in general is also a transitory [16:41] phenomenon that only existed between the [16:43] foundation of the computer or computing [16:46] tools and the existence of AI generally [16:48] speaking. And if all of that goes away [16:50] very quickly and all of those people can [16:53] be redistributed and recast into doing [16:55] other higher level, more creative [16:57] things, their productivity goes up by [16:59] 100x. [17:00] Is there really a consumer on the other [17:02] end of all of that productivity? Is [17:04] there really enough consumptive [17:06] capacity? And I think that's the [17:08] profound question that we all face. I [17:09] don't think that there's any limit. [17:11] >> Is that your way of saying that SAS goes [17:13] to zero or that's your way of saying [17:16] these companies go to zero? I'm just [17:18] saying knowledge work in general like [17:20] the the like what is the implication [17:22] >> is this just another dueling science [17:24] fiction take I mean or what's your [17:26] evidence for this [17:28] >> I think it's fine to have a sci-fi take [17:30] intuition I think is data because I can [17:33] show you some data that I think [17:35] contradicts what you're saying [17:36] >> and I have some firstirhand sort of [17:38] >> in the sense that there's more leverage [17:40] stacks that people are able to actually [17:42] >> well Jake I want to hear what you have [17:43] to say because I know you're [17:44] experimenting with this but let me just [17:45] show you a few data points real quick um [17:47] because I think this is relevant. So [17:49] we're really talking about the [17:50] disruption caused by coding assistance, [17:53] right? This is like the first big killer [17:55] app of AI. I mean I guess after writing [17:57] and research for chat bots and we'll [18:00] have agents later, but really it's all [18:02] about coding assistance, right? And the [18:04] ability to more easily create code. [18:05] That's what's creating the disruption to [18:07] the SAS category. Well, so let's just [18:09] focus on the data we see right now [18:11] around that. And there are a lot of [18:14] people who are pointing this out that [18:17] Enthropic right now has a job listing [18:20] for a software engineer on their website [18:23] right now for $570,000. [18:27] And a lot of people are kind of pointing [18:28] out, okay, so wait, so what Anthropic is [18:30] saying is they're still trying to hire [18:32] software engineers at a very high wage, [18:34] but somehow they think these jobs are [18:36] going to be eliminated. [18:38] >> Timoth might apply for that job. is. [18:40] >> Yeah. Austerity measures sounds pretty [18:42] good to me. [18:43] >> That's a lot of money. [18:44] >> Jim might take that job and then just [18:46] have AI do it for him. [18:47] >> No, I'm I'm I'm worried like I hope my [18:48] 8090 team doesn't see that offer. That's [18:50] a [18:51] >> that's a big number. [18:53] >> Our equity is way higher, but our [18:54] salaries are not that high. I mean, you [18:57] look you put these things together, it's [18:58] like that equity is money good. The [19:00] reality is like those guys are doing 5 [19:02] to6 billion structured secondaries every [19:05] year now or they're starting which means [19:06] that they will. That's like cash [19:08] compensation. So for for me to match [19:11] that, I need to be 3x higher than that. [19:13] >> Right? And I think a lot of people are [19:14] kind of pointing out, well, this is a [19:15] contradiction. Anthropic doesn't really [19:17] seem to be practicing what they're [19:18] preaching if they're paying enormous [19:20] amounts still for software engineers [19:22] even as they claim they're obsoleting [19:23] the entire category. Something doesn't [19:25] quite add up. Citadel Securities did a [19:28] new report that rebutts that Catrini [19:32] report and they show a couple of stats [19:34] here which I think are really [19:35] interesting. So, job postings for [19:37] software engineers are rapidly rising. [19:40] They're showing, I think it was roughly [19:42] a 10% year-over-year increase in the [19:45] demand for software engineers. On a [19:47] related note, they also show that [19:51] company formation is also rapidly [19:54] expanding and that may have something to [19:56] do with AI making it easier to start a [19:58] business or to get leverage to your [20:00] point, Freeberg. So look, there's a [20:02] couple of competing effects going on [20:04] here. And I think Aaron Levy had a [20:06] really good explanation of why you might [20:10] see something very counterintuitive [20:11] happening. And again, it all goes back [20:13] to Jeban's paradox. But what Aaron says [20:16] is that when you lower the cost of [20:18] something that was previously supply [20:20] constrained, demand for that thing goes [20:22] up. Software engineering is just one of [20:24] the easiest examples to contemplate, but [20:26] there are going to be many other jobs [20:27] like that. But think about software [20:29] engineering. Even among startups in [20:31] Silicon Valley, which I think are [20:33] probably some of the most attractive [20:34] places for software engineers to work, [20:37] there's always been a chronic shortage [20:39] of them. Then you've got the Fortune 500 [20:41] companies, non- tech companies, which [20:43] have always had an even harder time [20:45] hiring technical talent. So, you have [20:47] this massive unfilled need for software [20:51] engineers across the entire economy. [20:53] Now, you're going to be able to get a [20:55] lot more leverage out of software [20:56] engineers. It doesn't mean they're going [20:58] to get fired. It just means that now [21:00] maybe you can have a lot more 10x [21:03] software engineers and they're getting [21:05] those jobs are now being spread [21:07] throughout the whole economy. You know, [21:09] I also think just to put some numbers on [21:11] this, I think the cost structure of the [21:12] average Fortune 500 business is [21:15] something like 5% it includes all of [21:18] their IT, not just their software. You [21:21] know, what should it be? What should the [21:23] percentage of software be in an [21:26] enterprise cost structure? Elon [21:28] describes companies as cybernetic [21:30] organisms that are part software, part [21:32] human. [21:34] >> If you think about the current Fortune [21:35] 500 company being one or two percent [21:37] software, maybe they should be 50% [21:39] software. I think what Aaron is saying [21:41] here is the market for software and [21:44] software engineers was so constrained by [21:46] the lack of availability [21:48] that even if we 10x or 100x the [21:51] productivity of software engineers, the [21:52] demand will be there to absorb this new [21:55] supply. And so it could lead to this [21:57] explosion in productivity without the [21:59] massive job loss. [22:00] >> I think you're right. I think the the [22:02] thing that I would look at is I would [22:05] expect OPEX as a percentage of revenue [22:07] to fall off of a cliff but within that [22:10] opex the percentage of it that you [22:13] allocate to technology and technology [22:15] related things probably goes way way up [22:17] than what it is today. Okay, Jason, the [22:19] batch of people that are applying for [22:21] launch, [22:23] has SAS stopped? Has software stopped? [22:25] >> It's AI first companies obviously and [22:28] people [22:29] >> are they rebuilding traditional SAS [22:31] tools just cheaper? [22:32] >> Basically, everybody's building the [22:35] great, you know, as we talked about at [22:36] the all-in summit like some of these [22:38] companies are trying to build the best [22:40] pilot in the world or Whimo's trying to [22:42] build the best driver in the world. [22:43] People are now trying to build the best [22:45] SDR in the world, the best salesperson, [22:47] the best executive coach. And so we have [22:50] been like obsessed with claude co-work, [22:53] but mainly OpenClaw. And so what we did [22:56] was and and I think it's not developers [22:59] that are going to do all this work, it's [23:01] knowledge workers. So we had we have 20 [23:04] people in our firm. We had 15 of them [23:05] come in this weekend and they all got [23:08] trained over like six or seven hours had [23:10] to have their own openclaw agent and we [23:12] started building it. Every piece of [23:14] software that we wanted to buy or build [23:16] over the last 10 years that we never got [23:19] to my people are building in the last 30 [23:22] days. As an example, you know, when [23:24] you're selling ads for a podcast, you [23:27] want to check all the other podcasts and [23:28] what advertisers they have. We trained [23:30] an agent to go take the top 100 [23:33] podcasts, look through the transcripts, [23:35] figure out who the advertisers are, [23:36] check those advertisers in pipe drive, [23:38] tell us when the last time we contacted [23:40] them and put it into the sales room. [23:41] That was an SDR job that we wanted to [23:43] fill and software we wanted to build. [23:45] Then we wanted people to have [23:47] >> hold on that was a that was a human that [23:49] you were paying money and now you've [23:50] replaced with software or that human [23:52] still exists but now they just do it in [23:54] a better way. [23:54] >> Redeploying that human. We had a human [23:56] doing it. we're going to redeploy them [23:58] to do other things and the consistency [24:00] of this chimoth and the accuracy and [24:02] then it's doing it all night long. So we [24:05] have like seven of these agents in these [24:07] kind of roles. The next piece we did was [24:09] we gave my agent which is like the [24:10] ultron root access to Gmail, calendar, [24:14] zoom, notion, slack. And what it's doing [24:18] is it's giving each person here's what [24:20] you got done this week with their [24:21] manager. Here's the emails you sent. [24:23] Here's the meetings you took. Here's the [24:24] contacts. here the threads you were [24:26] involved in and then it's helping manage [24:28] those people and so we are [24:31] >> okay but all that all that to me says [24:33] you Jason despite all your dumerism [24:36] seems like you're growing and you're [24:38] going to be hiring more people and [24:39] you're more productive am I getting this [24:41] wrong [24:41] >> I'm not dumerous what I think's going to [24:43] happen in this position is [24:44] >> but you're growing and you're going to [24:45] be hiring more people [24:46] >> no no we're not going to add more people [24:48] definitely not adding people we have are [24:51] becoming 10 or 20% more efficient every [24:54] week Because the software we would have [24:55] paid for or built from another vendor if [24:57] we had the time or we wanted to build [25:00] custom software we had 10 engineers it's [25:02] being built by our openclaw agents. As [25:05] an example when we make clips for this [25:07] podcast other podcasts we have it go and [25:10] look at like uh this weekend startups [25:12] episode from 10 years tell us the three [25:14] best moments and it makes the clip it [25:16] puts the subtitles on it and then it [25:17] puts the clip into the Slack room. That [25:19] was something that was going to be a [25:21] full-time job. So, we're getting 10 20% [25:24] more efficient. Then I started doing it [25:26] at home. So, I had to take our [25:28] Instacart, pull out the last 10 orders [25:30] we did, and then tell us what we order [25:32] most of time. And then it's going to [25:34] automatically build a car for us. Every [25:36] single knowledge work job is being [25:38] automated right now. And you can take it [25:41] and if you're a business process head, [25:43] where you know how to like do a business [25:45] process and you can structure it and [25:47] write it with an agent, it'll just run [25:49] it every day, every week. We did another [25:51] agent, how do you make better [25:53] thumbnails? And we said, every Saturday [25:57] in your skills, so when you build an [25:59] open claw, it has like a soul file and [26:01] it has a skills file. In the skills [26:03] file, we told it sax every week go out [26:05] and look for people discussing how to [26:08] make better thumbnails on YouTube, how [26:10] to make better titles. It found this [26:12] week, Chimath, somebody at Mr. Beast [26:15] company talk about how they're using [26:18] heat maps. It was an article I would [26:20] have never known. It added it to its [26:22] skill and now whenever we post a [26:24] thumbnail, it tell tells us based on its [26:26] skill that it refineses every week how [26:28] to make that thumbnail better. And it's [26:30] starting to make the thumbnails. This is [26:32] becoming recursive. So you keep the same [26:34] number of people, but they get 10 or 20% [26:36] more efficient. I don't know what this [26:38] means for the larger economy. All I know [26:39] is it's the most exciting time I've had [26:42] online since the web came out, since the [26:44] internet came out. It is so much fun to [26:47] automate all this stuff. The big [26:48] question that I am thinking about that I [26:52] haven't gotten a good answer about so I [26:54] don't I don't know what you guys think [26:55] is [26:56] all these businesses are going to need [26:58] to batten down the hatches and give [27:00] themselves room to figure this all out [27:02] right like if you take Sax's point like [27:04] if you take your point Jal which is the [27:07] young nimble companies like yours are [27:09] going to be rapidly experimenting the [27:12] bigger larger companies are going to [27:14] slowly onboard themselves to start [27:16] experimenting All of that means we're [27:18] going to get much clearer answers to all [27:20] of this. But what it also means is that [27:22] you're going to have to have time so [27:25] that you can figure this all out. [27:27] >> And [27:28] if you want to buy yourself time, you're [27:31] going to need a ton of cash. And if [27:33] you're going to think about saving cash, [27:35] the one place tech companies literally [27:38] incinerate cash is how they do [27:41] compensation. [27:43] And so I kind of think like at some [27:45] point the next shoe will drop and all of [27:47] these tech companies have to really look [27:48] at stockbased comp because they [27:51] literally incinerate most if not all of [27:53] their free cash flow fighting the [27:55] delilution from stockbased compensation. [27:57] So if you want five or six years to just [27:58] be in the arena on the field figuring [28:00] this out, you're going to want to kind [28:03] of be very cash flow generative and [28:05] really conservative in how you spend [28:07] your money. Yeah, Saxs, the people who [28:09] embrace this, I think, become five or [28:11] ten times more valuable than the people [28:13] who are not. That's where I think the [28:14] opportunity in the economy is. So, [28:16] unless you think humanity is going to [28:18] run out of problems to solve, I think [28:20] it's going to be boom. It's going to be [28:21] boom town. And I think people are going [28:23] to start more companies because the the [28:26] barrier to start a company is no longer [28:27] three or four million dollars. You can [28:29] just have two or three people and you [28:31] start setting up these agents and man [28:32] you can [28:34] >> make software, you can do sales, you can [28:36] do PR, everything is getting faster and [28:38] faster and faster. So the time between [28:41] like conceiving of a product and [28:43] publishing it and finding a developer, [28:45] you don't even need a developer, you can [28:46] just publish software, the the wakeup [28:49] moment for me was we were talking to our [28:51] agent about, hey, we want to get this [28:55] functionality out of Slack. And it's [28:56] like, "Yeah, Slack doesn't have that, [28:58] but have you considered Matterpost?" I'm [29:00] like, "What's Matter Post?" Like, "Oh, [29:01] it's an open source project. I can spin [29:02] it up this weekend. Export your Slack [29:04] instance and put it there." And I was [29:05] like, "Oh, don't do that. We're only [29:07] spending 6K a year on Slack or 10K a [29:10] year." But the software is building CRM [29:13] systems for us. It's building agents for [29:15] us, and it wants to just build all the [29:18] software stack. So you could when you [29:20] renegotiate with Slack or HubSpot or [29:24] whatever company you're working with, [29:27] you're going to be able to say to them, [29:28] hey, we could roll our own and uh when [29:30] you want to upsell us on this latest [29:32] thing like you talked about SAS, [29:33] upselling is such a big part of SAS. [29:35] You're like, I can actually build that [29:37] software myself internally. I don't need [29:38] you to do it. [29:39] >> Ryan Peterson just posts on XC Claude [29:42] for legal seems to work just as well as [29:45] Harvey by the way. Now it's the SAS [29:48] apocalypse is going after private [29:50] companies now too. [29:51] >> Well, I think for a while now there has [29:53] been a question of which layer of the [29:54] stack is going to capture all the value. [29:57] >> So is it going to be the model companies [30:00] or could it be the applications that are [30:01] built on top of the models or you know [30:04] if there's a lot of competition at both [30:05] those layers of the stack did the chip [30:06] companies get it all? I think it's a [30:09] unclear question but [30:10] >> totally [30:11] >> yeah I think you know for any given [30:12] vertical application you do have to [30:16] defend why you think your value prop [30:19] will be sustainable as the underlying [30:22] foundation models get better themselves [30:25] >> and it's open source like this week we [30:27] put up Kimmy 2.5 it can do about 80 85% [30:31] of the jobs so we lowered our token [30:33] bills massively when we stood that up [30:35] all right listen this is TBD We got a [30:38] lot more to think about on this topic. [30:40] >> Just on this point of a lot of these [30:42] debates about AI are dueling science [30:46] fiction narratives. I just think that [30:48] the doomer narratives are inherently [30:50] more appealing to people. I mean, I [30:52] think it's partly just you look at most [30:54] sci-fi movies are dystopian, not [30:56] utopian. In addition to that, I think we [30:59] have a bunch of heristic biases in favor [31:01] of the dumer narrative. So, one of them [31:03] is the scene versus the unseen. It's a [31:05] lot easier to see the jobs that already [31:08] exist that could be obsoleted than it is [31:11] to imagine the new jobs and the new [31:13] business models that haven't been [31:15] created yet. And that will likely take [31:17] some great innovator or a genius to [31:19] think of in order to create. So we have [31:22] that huge heristic bias of not being [31:24] able to see the creation that's coming. [31:26] It takes way less creativity to think [31:28] about the potential destruction. And [31:30] then finally I think you know the other [31:31] heristic is just the whole fixed pie [31:34] fallacy. Most people do tend to think of [31:37] the economy as a fixed pie. This is why [31:39] you see so much anger against you know [31:41] millionaires and billionaires is because [31:44] of this idea that if someone's getting [31:45] rich it must be at the expense of [31:46] someone else. That's not actually the [31:48] case. The economy itself could be [31:50] growing larger as a result of someone [31:52] inventing something new that increases [31:54] production. A really good line from [31:56] another article that was written just a [31:57] couple weeks ago was the economy is not [32:00] a a pie, it's a garden and technology is [32:03] rain. So again, you know, all of this [32:05] technological innovation is going to [32:08] increase the growth rate of the garden. [32:10] It's not a fixed pie. And just because [32:13] you see an expansion and productivity in [32:17] one part of the economy does not mean [32:18] that you're going to see job loss in [32:20] another part of the economy. [32:21] >> Yeah. I think the job people are not [32:23] seeing but I'm seeing right now is the [32:25] person who creates agents, manages them [32:28] and is the maestro of the agents. The [32:31] person who can take the business [32:34] process, explain it and train the agent [32:36] to do it. And there are certain people [32:37] in business who are just really good at [32:39] operations. You were one of them Sachs [32:41] running companies and like that person [32:43] who can fire up an agent, train the [32:45] agent and figure out how to manage them [32:47] and figure out how to increase [32:49] >> with any new technology. Great job. and [32:51] it's not a developer. [32:52] >> Look, with any new technology, there's [32:54] always a huge change management aspect [32:56] with enterprises because it's hard for [32:59] them to adapt and change. And the people [33:01] in the organization who can lead that [33:04] change management are the ones who are [33:06] going to create an amazing career [33:07] opportunity for themselves. But it's [33:09] hard to do and that's going to slow down [33:10] the rate of change just the amount of [33:12] inertia in the economy. And also one [33:14] other constraint is going to be that at [33:16] some point here we may be token [33:18] constrained, right? I mean, we may not [33:19] have enough energy like we've talked [33:21] about, even though the chips are getting [33:23] so much better, that tokens per second, [33:26] tokens per watt, and tokens per dollar [33:29] are all increasing very fast, but we're [33:31] still going to probably be constrained [33:33] in the next couple of years on some [33:35] dimension, whether it's land, power, [33:36] shell, or just energy production or [33:38] maybe chip production. There are real [33:40] world constraints on just how fast we [33:43] can scale the infrastructure and that [33:45] will mean that these like hyper utopian [33:48] or hyperdystopian narratives will be [33:50] wrong. I don't think there's time in the [33:52] next few years for the whole economy to [33:54] change in the way that the extremes [33:56] would present. [33:57] >> I think you're right. I think you're [33:59] going to see a 10xing in [34:03] the demand for tokens, but I also think [34:05] you're going to see a 90% price [34:08] reduction in the cost of an output token [34:10] probably by the end of this year. So, I [34:12] think that to your point, like it's it's [34:15] going to just create an enormous upswell [34:17] of demand because we're going to be able [34:19] to cut the prices of an output token so [34:21] dramatically. And I think by the way [34:23] >> that discussion we had Shamath last week [34:25] when we talked about co the tokens [34:26] outpacing the employee salary and just [34:29] where are these tokens all going to come [34:30] from that was our most viewed clip or [34:33] one of the most viewed clips in the [34:34] history of this podcast. So people are [34:36] actually really focused on this. [34:38] >> I had my team at 89 we we redid our cost [34:40] model and now we have that as a line [34:43] item. When we think about fully burdened [34:46] cost of employees, we now factor that in [34:49] because we're we're at a place where [34:51] some of our engineers are just racking [34:53] up ginormous bills and then separately [34:55] just general runs that we do for general [34:59] purpose stuff that we need to just run [35:00] our product is it's so expensive. So I [35:03] am waiting with baited breath for what [35:05] Zach said which is like we need an [35:06] explosion in the capacity that's [35:10] available because I do think that the [35:12] the silicon solutions are coming that [35:14] will cut the cost but we need a large [35:17] block of land power shell ready to then [35:20] turn all of this stuff on so that we can [35:21] actually take advantage of it. [35:23] >> Rumors the new Mac Studio is coming will [35:26] have an M5 chip in it and will be [35:28] language model ready. So that's the [35:30] rumor is that they're building it for [35:32] models. So that could be an incredible [35:34] turn of events. Everybody's desktop [35:36] running the local model. Sax, you want [35:38] to have the final word here? Uh Free or [35:39] Freeberg before we [35:40] >> just to go back to what Jamath was [35:42] saying there. I mean, you've got [35:44] political forces that want to stop the [35:46] construction of all data centers in the [35:47] United States. So [35:49] >> if that gains steam, then that's going [35:50] to be a huge constraint on any change [35:52] whatsoever. [35:53] >> Can I tee this up for you, Jal? So I I [35:55] went back this weekend [35:57] and I looked at the number of data [36:00] centers that have faced local opposition [36:03] and whether there were patterns. And I I [36:05] posted it on X. So Nick, maybe you can [36:07] put this up, but [36:09] it was really a a very small [36:14] behavior which was pushing back on data [36:16] centers and getting them cancelled. We [36:18] had about 25 projects total of which [36:22] 20 were just in Q2 alone. There are 100 [36:25] data center projects right now that are [36:27] facing some form of local opposition. [36:29] >> So interesting. [36:30] >> If you take that 40% number and you [36:34] apply this and then you multiply by the [36:37] number of megawatts that they have [36:38] announced. Last year we lost almost 5 [36:42] gawatt [36:44] in terms of cancel projects. [36:46] This year coming in 26 we have about [36:49] seven that could be cancelled if you use [36:51] this math. If then you flow that through [36:55] open AI Sarah Frier said this that every [36:58] gigawatt for her for open AAI is about [37:00] 10 billion of revenue. So if you if you [37:03] assume that that's roughly accurate plus [37:04] or minus a billion here or there. What [37:07] that means is that 2025 the industry as [37:11] a whole lost [37:13] 50 billion of revenue [37:15] and this year if 7 gawatt gets canceled [37:18] it's about 70 billion. Now you're [37:20] talking about 130 billion of lost [37:21] revenue over these two years that'll go [37:23] forward in time that we miss out on. I [37:27] think that that's really bad. We need to [37:28] figure out a way to nip this in the bud. [37:30] >> So confounding because we were sitting [37:33] here 5 years ago, 10 years ago, local [37:36] municipalities were fighting and giving [37:38] discounts to try to get these data [37:39] centers open to get the jobs and get the [37:41] revenue. And now we've got people trying [37:43] to stop them. This is a perfect [37:45] transition for the State of the Union. [37:46] Before we get there, two important [37:49] programming notes. All-In is going to [37:51] host two events in 2026. One of them, [37:53] liquidity, May 31st to June 3rd, uh, in [37:56] Yonville, uh, up in Wine Country. [37:58] Chimath has taken control of the event [38:01] and he has set a standard for who gets [38:03] on stage. [38:04] >> None of you mids can control the [38:05] programming. I just [38:06] >> That's it. Chimath came in and he [38:09] dropped the hammer. Who do you got so [38:10] far? You want to tease a couple of [38:12] people who you invited to come speak? [38:14] >> I'll tease two. The first is an [38:17] incredibly dear friend of mine, [38:19] >> the axe of axes, [38:22] Dan Loe, who founded Third Point, who is [38:26] an unbelievable investor in literally [38:30] every domain, private credit, [38:33] public equities, private tech. [38:36] He's he's just a he's [38:39] >> a beast. So, he'll be doing a really [38:41] important keynote. He has not done one [38:43] of these public speaking slots in a very [38:45] long time. And then the second is the [38:47] CFO of OpenAI, Sarah Frier. [38:50] >> Oh wow. [38:50] >> Unbelievable start. And we're going to [38:54] double click into the entire business [38:55] model of OpenAI on stage in front of [38:57] everybody. [38:57] >> So uh go to allin.com and then for those [39:00] of you who plan ahead for travel, [39:01] >> more coming more coming every week. [39:04] >> If you are an All-In Summit fan, I can't [39:07] believe it. Freedber, we're going to be [39:08] in our fifth year September 13th to [39:10] 15th. only gets better. Gets better [39:13] every year. I could have some good [39:14] parties, too. I mean, that Back to the [39:16] Future and the Bladeunner parties, those [39:17] were epic. [39:18] >> allin.com/events. Hey, can I give a plug [39:21] to friend of the pod, [39:23] >> Bill Gurly? He's got an amazing new [39:24] book, Running Down a Dream. Please, [39:26] >> wait, wait, wait, wait. Before we start, [39:28] >> before I start, [39:29] >> there it is. Running down a Dream, I [39:30] just want everybody to just stop the [39:32] pause the podcast. I want you to buy [39:33] three copies, give it to two young [39:35] people and a parent. You know, [39:36] >> this book is incredible. [39:37] >> It's a great book. It's a great book. It [39:40] It really is. inspiring for kids and uh [39:42] Bill Gurly, friend of the pot. He always [39:44] shows up for us. [39:44] >> Jal, do an impression for us of what it [39:46] would be like if you and Bill Gurley [39:48] started a podcast together. [39:49] >> All right, everybody. Welcome to the [39:50] JCBG podcast. I'm your host, Jason [39:53] Calakanis. and I'm Bill Gurly and we're [39:56] here in Texas at Terry Blacks where [39:58] we're getting some beef ribs and we're [40:00] going to discuss investing in [40:03] marketplaces [40:04] as well as my new book running down a [40:08] dream which will teach your kids how to [40:10] not be fuckups and if your kids are ups [40:14] you can hit them in the back of the head [40:16] with the book [40:18] Texas style. [40:20] One of the big topics uh and I think [40:22] something you're working on with [40:24] President Trump's acts is uh this energy [40:28] pledge. I've been seeing uh rumblings [40:30] about this. Explain what's going on in [40:31] terms of getting the country in sync [40:34] around these data centers and energy. [40:37] Well, the president announced in the [40:38] State of the Union last night that he [40:40] supports a rateayer protection pledge [40:44] which requires uh the major tech [40:47] companies to provide for their own power [40:49] needs for AI data centers so that [40:52] residential consumers do not see their [40:53] rates going up. I think this makes total [40:55] sense. I think Chamas, to your point, [40:58] this is the reason behind a lot of the [41:00] opposition to new data centers is that [41:02] the local residents fear that their [41:03] electricity prices are going to go up [41:05] and that shouldn't be the case. And so [41:07] the president has said that he's [41:09] committed to not allowing residential [41:11] rates to go up as a result of data [41:13] centers. It's pretty straightforward. [41:15] You get the big tech companies, the [41:16] hyperscalers to pay for the increase in [41:18] the electricity cost or you let them set [41:22] up their own power behind the meter. The [41:23] president's been talking about this for [41:25] over a year that our biggest AI [41:27] companies would also become big power [41:29] companies because we would let them uh [41:31] stand up their own power generation [41:32] behind the meter. So these data centers [41:36] don't even have to connect to the grid. [41:38] They could just do collocation [41:40] themselves. But also, I think that with [41:42] this rateayer protection pledge, what [41:45] you're going to see is that it could [41:46] actually bring down consumer prices [41:49] because what happens is that when these [41:51] data centers then set up their own power [41:53] and connect to the grid, they can give [41:55] back the excess to the grid. Also, they [41:58] will make investments in scaling the [42:00] infrastructure. So, although electricity [42:03] is priced at a metered rate, the costs [42:05] to generate it are not all variable. [42:07] There's a lot of huge fixed costs in [42:09] there. So when you increase scale then [42:12] you can actually reduce the the metered [42:14] rate. So again you know this is really I [42:17] think the rebuttal to Bernie Sanders who [42:19] just wants to stop all progress [42:21] whatsoever. [42:22] I saw a funny post calling it bananas [42:25] which is build absolutely nothing [42:26] anywhere near anyone. So this is the [42:30] >> bananas is replacing the new nimi. So [42:32] you just can build absolutely nothing. I [42:34] think the president's approach finds a [42:36] very good balance here, which is look, [42:38] >> we can have progress. Just don't make [42:40] residential consumers pay for it. Let [42:41] the big tech companies pay for it [42:43] themselves. And I think you'll see more [42:44] coming out about this from the White [42:46] House next week. [42:47] >> Quite a deaf move. Freeberg, how should [42:50] America be thinking about this great [42:53] data center buildout, energy usage, you [42:56] know, if you expand it out over the [42:58] coming decade? And how do you sell that [43:01] to the backdrop that you talk about the [43:03] socialist movement? You got a great [43:04] interview coming out with Ray Dalio and [43:06] the all-in interview program next week. [43:09] How do you think about those competing [43:11] forces? You've got the socialists saying [43:13] bananas, nimi, slow down, del, and then [43:18] you've got this incredible race we're in [43:20] for efficiency and this opportunity and [43:22] abundance. How would you sell it to kind [43:25] of bring these two sides together? Or is [43:27] it just [43:30] impossible. The data coming in and out [43:32] of data centers moves at roughly the [43:34] speed of light. So you could put them [43:37] anywhere. And I think that our policy [43:40] makers need to be very cognizant of that [43:43] fact. You have and we do connect the [43:47] internet using high-speed cable, [43:50] high-speed [43:52] fiber optic throughout the world. And so [43:54] theoretically if we don't embrace and [43:58] allow the economic development of the [44:01] data center industry and it will [44:02] fundamentally be an industry because it [44:04] is almost like the new sort of oil. [44:07] Where are the oil rigs going to go? [44:09] Where are the railroads going to go? [44:11] Where are the telegraph lines going to [44:12] go? Where are the factories going to go? [44:14] If we don't put them here, someone else [44:16] will put them on their shores. Someone [44:18] else will put them in their country. [44:19] Someone else will put them in their [44:20] jurisdiction. And a lot of the economic [44:22] value that arises from the people that [44:25] will build those facilities, the energy [44:28] that will be installed to produce power [44:30] for those facilities, and then all of [44:33] the second and third order industries [44:35] that emerge as a result of those [44:36] installations, that value will acrue [44:39] elsewhere. [44:40] >> Such a good point. [44:40] >> So, [44:41] >> yeah, [44:42] >> it's not going to like just go away. The [44:44] the demand is there. The economy is [44:45] moving forward. AI is moving forward. We [44:49] live in a world with 196 countries and [44:52] data centers do not take up a lot of [44:54] space. They're very small relative to [44:56] the economic value that they produce. If [44:58] you zoom out on the map of the world, [45:00] all the data centers in the world fit [45:02] under the tip of a pin. And so this is a [45:06] very small footprint and if we're going [45:08] to give up hundreds of thousands of jobs [45:11] and many billions of dollars of economic [45:14] value creation, we're being pretty silly [45:15] and pretty obtuse in our view of the [45:17] world. I would just like encourage the [45:19] system that I think is the right system [45:21] and we talked about this last time where [45:23] provided data centers are producing [45:25] their own electricity. That means that [45:27] you're taking electricity consumption [45:29] off the grid because they otherwise are [45:32] not being used on the grid and that will [45:34] reduce the cost of electricity for other [45:36] residential and industrial users. So, [45:39] it's silly to think that we need to put [45:40] a moratorium on data centers. As soon as [45:42] you do that, the companies that use data [45:44] centers are not going to slow down. [45:45] they're going to go put them somewhere [45:46] else and we're going to miss out. [45:47] >> And it's such a good point, Chimamoth, [45:49] because you were recently in the Middle [45:50] East and I've been there a bunch in [45:53] Saudi, UAE. These are the folks who [45:56] built a large portion of those oil [45:58] refineries. And they are savvy to this. [46:01] And what are they doing in Saudi, UAE, [46:04] Qatar, all of these regions? They're [46:06] doubling down. They're 10xing their data [46:08] center builds. So to your point, [46:11] Freedberg, either we build them or [46:13] they're going to go somewhere else. And [46:14] there are people who are willing to [46:16] underwrite these and they're willing to [46:18] take out the red tape [46:21] from the process here and move quicker [46:23] than us. So we I think this is a pretty [46:24] deaf move by President Trump to say, [46:27] "Hey, you guys should all just guarantee [46:29] that consumers don't get impacted." The [46:31] water thing is a total hoax. Like the [46:33] water is recirculated. That's a hoax. I [46:35] think this is really smart. I think that [46:37] what the president's doing and what Sax [46:39] is doing is really smart. The thing to [46:41] keep in mind is that there's still a [46:44] risk that prices go up and it has [46:45] nothing to do with these data centers [46:47] and it has everything to do with the [46:48] business model of being a utility [46:51] because what happens is in order to get [46:54] a license a monopoly license in an area [46:58] to provide energy to generate energy for [47:01] a community. [47:03] The exchange works in the following way. [47:05] You go and you present a capex plan to [47:08] the public utilities commission. That's [47:10] effectively your budget that says here [47:12] are the lines I'm going to upgrade. Here [47:14] are the generators I'm going to upgrade. [47:16] Independent of data centers. The reality [47:18] is the draw the electricity consumption [47:21] of individual Americans is going up [47:23] because we have more devices, we have [47:25] cars, we have all of these other things. [47:27] So what we also have to do is we have to [47:29] look at how utilities's business model [47:31] actually incentivizes them to increase [47:34] prices by making all kinds of [47:37] investments. So we have to do a good job [47:39] of making sure we hold everybody [47:40] accountable because otherwise what you [47:42] could see is that the data centers [47:45] taking on the burden for themselves but [47:47] price is still continuing to escalate [47:49] because a utility says I need to spend a [47:51] billion dollars this year to upgrade my [47:53] infrastructure. And what that allows [47:55] them to do is take that billion dollars [47:56] and essentially invest it for a return. [47:58] That's the business model of utility. [48:00] And this is really happening in blue [48:03] states. Micron has a hundred billion [48:06] dollar mega fab in New York. And there's [48:08] a lawsuit by six 1 2 3 4 5 citizens. [48:13] >> That's shameful. [48:13] >> And the project has taken 12 [48:17] >> 1200 days. 1,200 days [48:20] >> between their announcement and the [48:21] groundbreaking. And they spent 612 days [48:24] on the environmental impact study. [48:27] People wake up. Just go to Texas. Elon [48:30] built his factory here, the Gigafactory, [48:33] in under like 18 months. This is the [48:36] great state of Texas. Come here. We'll [48:38] build it for you and you'll be done. [48:40] >> Yeah. I don't know why anyone bothers [48:41] with the blue states anymore. They make [48:43] it too hard to build. [48:44] >> It's It's so dumb. It's such a [48:46] self-owned, too. Like, what? Don't you [48:48] want to be part of the future? You're [48:49] literally ankling the entire country to [48:53] scratch the odd. [48:54] >> By the way, there are a lot of people in [48:55] New York who want to work. This is not a [48:56] case actually of this new fab being [49:00] unpopular. [49:02] The majority of people in the area [49:04] actually want this plant being built. [49:06] They want the jobs that are going to [49:07] come there. A lot of people say data [49:09] centers don't create a lot of jobs. This [49:10] is actually a chip fab. So, it will [49:12] create a lot of jobs. A lot of good high [49:14] paying jobs. People want it. But six [49:16] people can stop it with a lawsuit after [49:18] it's already been through a two-year [49:20] environmental review. [49:21] >> It's not blue and red states. These are [49:23] nonprofits that get organized to create [49:24] this kind of chaos. I remember looking [49:26] at a massive lithium investment in [49:31] Nevada. And the whole point was to [49:34] domesticate [49:35] lithium production. And what was [49:37] interesting is this enormous deposit [49:40] that's just sitting there ripe for [49:41] development right before they were about [49:43] to get environmental approvals or right [49:45] after there was a lawsuit by people who [49:48] wanted to protect the upper land grass. [49:50] It's seared in my mind that the upper [49:52] land grass of Nevada is the reason why [49:54] we do not have domestic national [49:56] security around lithium. [49:59] And you have to ask yourself why is this [50:01] possible? And it's possible because you [50:03] have these environmental nonprofits that [50:06] can go and create this chaos with [50:08] absolutely no risk to them. Zero. They [50:11] can fund raise around it and they can [50:12] create this chaos. I mean, Nick, to this [50:14] point, this is an example of Greenpeace. [50:17] And specifically here, they were pushing [50:20] back on an oil pipeline to such a degree [50:22] and they created so much chaos that they [50:24] were sued. And a North Dakota judge just [50:27] said that he's going to order Greenpeace [50:28] to pay damages. That should total almost [50:30] $350 million in connection to those [50:33] protests. [50:34] >> And it should not be the case that six [50:36] people can slow down a hundred billion [50:38] investment package. That's not right. [50:40] Well, I think there's and I just want to [50:42] highlight this important point. There's [50:44] not a lot of logic and reason. [50:47] You guys are right. But I do think [50:49] there's a lot of emotion and there's a [50:52] huge aversion to big tech, a huge [50:54] aversion to wealth creation by select [50:56] individuals, select companies, a huge [50:59] aversion to economic growth that doesn't [51:01] benefit everyone. There's a fundamental [51:03] kind of underlying left behind emotion [51:06] that drives a lot of this. And I've said [51:08] it before, but I think unless there's [51:10] systems or mechanisms that get folks to [51:12] come along with the value creation ahead [51:14] and and help them connect their own [51:17] lives to the value creation that's being [51:18] realized, they're not going to be [51:20] supportive because there is this kind of [51:22] diametric opposition towards big tech, [51:25] towards the wealth gap, towards value [51:27] acrruel to a select few companies or [51:29] select few individuals and this fuels [51:31] and feeds that. So I think fundamentally [51:33] maybe it's not just about giving the [51:35] data centers their own power capacity [51:38] but there's got to be mechanisms and [51:40] tools that helps the broader population [51:42] understand or recognize or get some [51:44] benefit from it as well where they're an [51:46] owner in it or participant in it because [51:49] they have the power as we're seeing they [51:50] have the power to stop it. Therefore [51:52] they want to have some benefit for [51:54] providing authority to do it. And these [51:56] six people are concerned about housing [51:59] costs, worker exposure to toxic [52:02] chemicals, pollution in air and water, [52:04] greenhouse gas emissions, energy [52:06] consumption, flooding of the wetlands, [52:08] all these things that obviously could be [52:10] mitigated. All right, let's keep moving [52:12] here. We got a lot more docket to get [52:13] through. State of the Union came in at [52:16] 108 minutes [52:19] and it's the longest in 60 years. [52:20] actually the longest since they started [52:22] tracking this. The theme of President [52:24] Trump's State of the Union this year, [52:26] America at 250, strong, prosperous, and [52:30] respected. Trump took a bunch of victory [52:32] laps, inflation, jobs, closing the [52:34] border, all those have gone really well. [52:36] But this comes to the backdrop of [52:38] Trump's approval rating being super [52:40] challenged. He started his first year at [52:44] plus 11.7%. Now he's negative 14 14.3% [52:49] 26 point swing. Economy started plus 3.4 [52:52] down to 18.2 and trade started at 5.9% [52:56] and we'll talk about the tariff stuff [52:57] later and went down to 22.7. [53:01] So let's call balls and strikes here, [53:03] gentlemen. Favorite moments. What were [53:05] your favorite moments from the State of [53:07] the Union and just general impressions [53:09] of 1 hour and 45 minutes of Trump going [53:14] to town? [53:16] I thought it was great. [53:18] >> Favorite moment? Favorite moment or two? [53:21] Well, I had a couple. [53:24] One was the [53:26] Elon Omar Rashida Tlay death stare and [53:31] them just like losing their minds and [53:34] and screaming. I just thought it was so [53:38] unamerican. The second was when he was [53:40] calling for law and order where and [53:43] focusing and prioritizing on American [53:45] citizens and [53:47] >> illegal uh aliens [53:49] >> and none of the Democrats stood up. I [53:52] thought that was kind of foolish. It was [53:54] like obvious things and the Democrats [53:55] wouldn't applaud but this time they did [53:57] like they did for the hockey team which [53:58] I thought was like the right thing to [53:59] do. And then the fourth thing is just a [54:01] a shout out to our friend Brad Gersonner [54:03] who got a big shout out from the [54:04] president. I don't know Sax if you [54:05] engineered that or not but that was [54:07] >> that was fantastic. [54:09] He got like a double shout out. It was [54:10] like a double tap. [54:12] >> Yeah, that was really cool. [54:13] >> That was surreal. [54:14] >> Our group chat group chat went crazy. It [54:16] was really That was really cool. [54:18] >> Those are my four highlights. [54:20] >> Great. Here's your uh here's your clip [54:22] of uh yeah, Democrats not standing for [54:26] Americans uh over illegal aliens. 20 [54:30] seconds. [54:31] >> If you agree with this statement, then [54:33] stand up and show your support. The [54:36] first duty of the American government is [54:39] to protect American citizens, not [54:42] illegal aliens. [54:52] >> Why wouldn't you stand for that? That's [54:53] an easy one to stand for. Doesn't make [54:56] any sense. [54:57] >> Would you stand for it, Chico? [55:00] >> Yeah. I mean, I I'm I'm pro I can be [55:02] anti-ICE, but I'm pro-American and I'm [55:05] pro reasonable immigration like 90% of [55:08] the country is. So, it just doesn't make [55:10] any sense. Uh, do you have any [55:12] highlight? [55:12] >> Do you think American citizens should be [55:14] prioritized over illegals? [55:15] >> Well, of course, of course. Yes. Of [55:18] course. Yes. And then I I also think [55:20] there should be a path to citizenship [55:22] for people who have been here for a [55:23] while. And I think that's what the [55:24] majority of the country thinks as well. [55:25] >> Your point is I can hold two thoughts in [55:27] my head. I would have stood if he asked [55:29] me. [55:30] >> Yes, obviously we should take care of [55:32] American citizens first. Yes. And we [55:34] should deport violent criminals. We've [55:36] been over this like a million times [55:37] here. This is like consensus. [55:39] >> But what do you what do you think is [55:40] going on in everybody else's head when [55:42] they're like we got to we cannot stand [55:43] for this? [55:44] >> These two sides I mean I think it's like [55:46] the tariff thing. It's like the ice [55:48] thing. These two sides cannot work [55:49] together. It's just the most polarized [55:52] it's ever been. Trump is not like the [55:54] kind of guy to reach across the aisle. [55:56] the Democrats are now digging in. So, we [55:58] just have a dysfunctional government [56:00] where, you know, in a more functional [56:02] time period, like under Clinton, let's [56:05] say, or Bush, people would have gotten [56:07] together, and I'm I'm jumping ahead to [56:09] the tariff discussion. And they would [56:10] have said, "Yeah, of course, tariffs are [56:11] done in Congress. That's the law. [56:13] Whatever. What are your thoughts, Mr. [56:15] President? How can we support your [56:16] tariff program?" But now it's like, "Oh, [56:18] well, we don't work together. We don't [56:20] actually have discussions anymore. [56:22] There's no bipartisan, [56:24] you know, collaboration. All these [56:26] politicians are disgraceful descriat [56:28] across the board. They should be working [56:30] together for the American people. If the [56:31] president wants to do tariffs, [56:33] >> they should be reasonable about it and [56:35] give him the power to do reasonable [56:37] tariffs and he should be reasonable and [56:38] say, "Hey, I understand that's your [56:39] power. Let's get together and we'll [56:41] we'll chop it up and let's have dinner [56:42] together." But they're just too [56:43] polarized. It's just disgraceful where [56:45] this country has gotten to. I blame both [56:46] parties. [56:49] Whenever the Democrats get smoked out as [56:52] being radicals and extremists, you [56:54] always want to basically say a pox on [56:56] both your houses and blame the [56:57] Republicans and Democrats equally. The [56:59] fact of the matter is the president said [57:02] to the audience, to the members of [57:04] Congress, hey, if you agree with the [57:06] statement, stand up. And of course, [57:08] every single Democrat sat there [57:10] stonefaced and refused to applaud or [57:14] acknowledge what he was saying. This was [57:17] a very easy test for the Democrats to [57:19] pass. [57:20] >> What I just said, [57:21] >> in fact, it was what I said. [57:22] >> In fact, it was a political risk for the [57:24] president because it was so easy for the [57:27] Democrats to demonstrate that they're [57:29] operating in good faith and that they're [57:30] willing to be bipartisan and they're not [57:32] extremists and they're actually common [57:34] sensical and logical and they completely [57:37] failed the test. And by the way, it [57:39] wasn't just on that one. I mean, let me [57:40] just tell you some of the other ones [57:42] where they refused to applaud. So they [57:45] refused to applaud the grieving families [57:48] of innocent American women and children [57:50] murdered by criminal illegal aliens, [57:52] including the mother of Ireina [57:55] Zeritzkaya. [57:56] >> That was very sad. That was sad. [57:58] >> That was unbelievable. They refused to [58:00] applaud for securing our homeland and uh [58:03] ending the invasion of criminal illegal [58:06] aliens, killers, rapists, gang members, [58:08] and traffickers. They refused to applaud [58:10] for unifying against political violence. [58:13] So, the president mentioned the [58:14] assassination of Charlie Kirk. They [58:16] would not even do a polite clap for [58:19] Erica Kirk and unifying against [58:20] political violence. They refused to [58:23] applaud for keeping violent criminals [58:25] locked up. They even refused to applaud [58:28] for lower prescription drug prices for [58:30] millions of Americans because it was [58:31] President Trump who orchestrated that [58:33] policy. And there were so many other [58:35] examples like that. And you know, I [58:37] think the reason why this speech was so [58:39] effective, and by the way, it's not just [58:41] me saying it. something like twothirds [58:43] of the people that CNN pled, so [58:45] twothirds of CNN watchers said it was [58:48] highly effective and something like [58:49] threearters of CBS news viewers said it [58:52] was highly effective is because the [58:54] president laid out 8020 issue one after [58:57] another, right? Or even 9010 issues or [59:00] 955 issues. I mean, these were all [59:03] issues where the overwhelming number of [59:05] Americans, I think, agree with the [59:07] policy the president laid out. And in [59:09] every single case, the Democrats [59:11] supposed to wouldn't even give it [59:12] applied applause. And that is different [59:14] than than in the past. And you can say [59:16] that's because of hyperartisanship and [59:18] polarization, but it's also because of [59:20] another thing. It's because the [59:21] Democrats have become a party of [59:24] radicalism and extremism. And the [59:27] viewpoints that they expressed through [59:30] their aesthetics the other night, they [59:33] do express those things in policy and in [59:35] speeches all the time. So it's not just [59:38] like a oneoff or you know somehow like [59:41] we have a misconception of who these [59:43] guys are. I think you know the big line [59:45] of the night was when Trump just sort of [59:47] said these people are crazy. I mean he [59:49] said it in like almost mournful and [59:51] regretful way. He doesn't want them to [59:52] be crazy. He wants them to be rational [59:54] so he can work with them. I mean, [59:56] >> but I think that point re I'll I'll [59:57] still point the other side, which is [59:59] this has been going on for a couple of, [60:01] >> you know, state of the unions here [60:03] across this. The Republicans didn't [60:04] stand for the Democrats often and uh [60:07] it's a bit of showmanship. But the truth [60:09] is Trump is the divider and chief chief. [60:11] He always is attacking people. He's [60:13] always mocking people. So, they don't [60:15] want to play ball with him. So, I do [60:17] think you can both choice in politics. [60:19] You got to counter punch. [60:20] >> Uh no, I don't think so. That's actually [60:22] the that's actually the problem with the [60:25] that philosophy, Trump's philosophy of [60:27] we have to counter punch, we have to [60:28] attack, we never have to apologize, we [60:30] never have to be reasonable. That's part [60:32] of what's broken down in our politics [60:33] and these two sides should work [60:35] together. We should go back to a [60:36] bipartisan. [60:37] >> Are you going to work with Helen Omar [60:39] when he is when the president going to [60:40] be easy, but she is the mirror. Hold on, [60:42] I'll finish my statement. You asked a [60:44] question. I think Trump is the mirror of [60:47] that. He has been hostile towards these [60:49] Democrat. He doesn't give them an inch. [60:50] they should be more collaborative. That [60:52] that's what the balance of power between [60:55] the executive branch, you know, and and [60:58] these uh you know, congressmen and and [61:00] the Congress and the Senate. Like this [61:02] is how it's supposed to work. And these [61:04] two sides need to learn how to get back [61:06] to listening to each other, [61:08] understanding each other's positions, [61:09] and then finding a middle ground. And [61:11] that's why the Democrats lost last time [61:13] because they didn't have the common [61:14] sense to say, "Hey, everybody wants the [61:16] border closed." To your point, it's a [61:18] 90% issue. and Kla Harris was too dumb [61:21] to just say, "Yeah, we should have [61:22] closed the border. It's closed now." And [61:24] uh we've got it. Anyway, the whole thing [61:26] is a mess. I understand you got to fight [61:28] for your team. I don't like the [61:29] >> count. You just actually made the key [61:31] point. You made the key point which is [61:33] underlying the optics and the [61:34] polarization. You have issues and on [61:37] those issues, President Trump is on the [61:40] side of the American people. the issues [61:42] where 80% of the American people agree. [61:44] Some huge percentage, I don't know [61:45] exactly what it is, thinks that the [61:47] Somali daycare fraud in Minnesota was an [61:51] outrage. And the president is right to [61:54] point that out. And what's the [61:55] Democrat's reaction? You've got Ilan [61:57] Omar screaming from the audience at him. [62:00] >> Yeah, she's alone. I mean, at the end of [62:02] the Democrats are not that different. [62:03] You did have you did have Elizabeth [62:06] Warren stand for uh stopping Nancy [62:10] Pelosi from trading stocks and that gave [62:11] Trump his best oneliner of the night. [62:13] That was his best oneliner clearly and [62:15] they stood for Iran too and stopping [62:18] Iran from being a nuclear. [62:19] >> I give Elizabeth Warren credit for that. [62:21] >> There you go. You don't have to punch [62:22] her back. [62:23] >> Stop insider trading act without delay. [62:27] >> Yeah. See, that's something bipartisan. [62:29] Look at that, Zach. That's what you need [62:31] to get the country back to. But that [62:33] hold on. But this this disproves the [62:35] point you were making before. You said [62:37] that it was polarization. [62:38] >> You stood up for that. I can't believe [62:41] >> can't believe Nancy Pelosi stand up if [62:44] she sing pal. Good job. [62:47] >> That's why he's so good is he's in the [62:49] moment and he's reacting to what's [62:50] happening in the chamber. He's not good. [62:52] He's not just reading from a [62:53] teleprompter. And he nailed it. But [62:55] look, that moment disproves what you [62:57] were saying, Jal, because this is not [62:58] just about polarization. On that issue, [63:01] Elizabeth Warren was willing to stand [63:02] because she actually, to her credit, [63:04] wants to ban insider trading by members [63:06] of Congress. Yes. But on the rest of [63:08] those issues, like securing the border, [63:09] she did not stand. Why? Because she does [63:11] not agree with the president on that [63:13] issue. [63:14] >> We just have to get back to these sides [63:16] working together. That's my personal [63:18] feeling. [63:19] uh Freeberg, any thoughts on the uh [63:21] theatrics and uh Trump's first year at [63:24] large and you know the sort of back and [63:26] forth and is there any hope that these [63:29] two teams could collaborate at some [63:32] point on something like say the [63:34] ballooning deficit which Trump has not [63:36] gotten under control in his first year [63:38] and it's going to be $2.5 trillion [63:40] dollars added. What are your thoughts [63:42] here on them collaborating on anything [63:44] important? Friedberg, that's probably [63:45] one thing they can agree on is just keep [63:47] the money flowing. Got it. They'll both [63:49] give a They'll both give a a standing [63:51] ovation for burn more capital and put us [63:53] more in debt. Well said my guy David [63:57] Friedberg, sultpan of science, it is [64:00] your time to shine. The world's greatest [64:03] moderator has decided we're going [64:05] directly to science corner. This is your [64:07] time to shine. [64:09] >> Sax's time to drop a deuce. [64:10] >> Sax went immediately off camera. [64:12] >> Yeah, he has to drop a deuce. [64:13] >> What's up? [64:16] >> But what do you need me for? This is [64:17] very important for you. [64:19] >> Freeberg's going to talk. [64:20] >> Lightning round for science corner. Go [64:21] Freeberg. [64:22] >> Wait, wait, wait. Are we doing any more [64:23] topics after this or can I just leave? [64:24] >> Yes, tariffs. [64:25] >> Yes, we're doing tariffs. [64:26] >> Wait, why would we do that? [64:27] >> Because I want to get the audience to [64:30] sleep. I'm not saying Science Corner [64:32] doesn't have its audience. [64:33] >> Listen, you can go. [64:34] >> Why wouldn't we do Yeah, exactly. You're [64:37] screaming. [64:37] >> Let him do his work. [64:38] >> Let him let him cook. Let him cook. [64:40] >> Freeberg, tell us about this Harvard [64:41] scene. [64:42] >> Speaking of science, I think that [64:43] there's a very important moment [64:44] happening right now. We've talked a [64:46] number of times on the show about [64:47] Yamanaka factors. These are these four [64:50] proteins that were discovered by Shina [64:53] Yamanaka that we found later that when [64:56] applied to cells, mamalian cells can [64:59] actually reverse the age of those cells, [65:01] reset the epigenetic clock, reset the [65:04] epiggenome, which is the little markers [65:05] on top of the DNA that turn genes on and [65:07] off back to a youthful state. [65:10] extraordinary groundbreaking work that [65:12] was done that won the Nobel Prize led to [65:14] the foundation of several companies. [65:16] There's a Harvard uh scientist named [65:18] David Sinclair. He's a bit of a [65:20] controversial character. Do you guys [65:21] know him? I think you guys one one or [65:23] two of you may have met him. Chim, you [65:24] ever met him? [65:25] >> I I I followed him. I've seen his stuff. [65:28] So Sinclair is kind of bemoaned a little [65:31] bit by the scientific and academic [65:33] community for being a little too [65:34] overhypy snake oil salesman as some of [65:38] folks have claimed because years ago he [65:40] sold a company to GSK saying resveratrol [65:43] would reverse aging and you know he made [65:45] $720 million on that and it didn't end [65:47] up working and he's promoted certain [65:49] supplement companies and so on. So, I [65:51] want to preface with that before I kind [65:52] of underwrite what he's saying with this [65:54] next thing, but he's a co-founder of a [65:56] company called Life Biosciences, and [65:59] they've reached a major agreement with [66:00] the FDA to be the first company to treat [66:03] humans with Yamanaka factors. [66:06] Specifically, what they're doing is [66:07] they're going to be delivering these [66:09] Yamanaka factors, these are these [66:11] proteins [66:13] that rejuvenate cells and make them [66:15] youthful again into the eye. And so [66:18] their their first indication is to [66:20] actually inject them into the vitrial [66:22] fluid in the eyeball and they'll affect [66:24] the retina in the eye to address people [66:28] that have gotten blind from glaucoma or [66:31] one of these kind of stroke like [66:33] diseases that happen in the eye. And the [66:35] expectation with this phase one clinical [66:37] trial is that the delivery of these [66:39] Yamanaka factors into the eye will [66:41] rejuvenate the retina, make it youthful [66:44] again, and restore vision. If it works, [66:47] which it's expected to because we see [66:48] this result happen in animal models, it [66:51] could be an extraordinary breakthrough, [66:53] not just in terms of blindness, but in [66:55] terms of the first human application of [66:58] Yamanaka factors to reverse aging. The [67:01] way they're doing it is they're actually [67:03] packaging up DNA that will make these [67:06] proteins into viruses, AAV virus that is [67:12] delivered into the eye. The virus will [67:14] then go into the retinal cells and then [67:17] will deliver this payload for this DNA [67:19] to make these proteins in the eye uh [67:21] cells and it can be turned on and off. [67:24] Amazingly they've created a switch [67:26] mechanism in it where the protein [67:28] production the production of these [67:29] Yamanaka factors can be turned on and [67:31] off by taking an antibiotic called [67:34] doxycyc. So the person that gets the [67:36] delivery of this drug takes the [67:38] antibiotic turns on the production of [67:40] these Yamanaka factors and then [67:41] theoretically their eye cells will deage [67:44] will get youthful and their vision will [67:46] be restored. So phase one clinical [67:48] trials underway. First time in human [67:50] history we're seeing Yamanaka factors [67:52] being delivered into humans. Literally [67:54] the tip of the iceberg. There are now [67:56] over a dozen startups that are trying to [67:59] deliver Yamanaka factors which are these [68:01] proteins or some other sort of protein [68:03] that can actually reverse aging by [68:05] restoring the epiggenome and cells and [68:07] make them young again. So this is the [68:09] beginning of a wave of what I think will [68:11] be the most extraordinary revolution in [68:14] human therapeutics and ultimately could [68:16] lead to [68:18] you know some people would argue the [68:20] fountain of youth. Is this just a talk [68:22] study? So is it mechanism? [68:25] >> They're not. Yeah, they're they're well [68:26] they'll see results. They'll see [68:27] results, but they're going to they're [68:28] going to keep dosing low. But you will [68:30] see results. [68:31] >> God, that's going to be incredible. [68:32] >> It's going to be incredible. By the way, [68:34] the number of other folks that are [68:36] gearing up for phase one using if not [68:39] the Yamanaka factors, other factors that [68:41] they've identified or designed as an [68:44] alternative to Yamanaka factors again to [68:46] rejuvenate the cells. And just to remind [68:47] folks, the way this works is it was [68:49] discovered that these proteins when they [68:51] go into a cell, they take all of those [68:53] little markers that sit on top of your [68:55] DNA that turn genes on and off and they [68:58] create a system that causes them all to [69:00] move to the right place. So, it resets [69:02] the markers so that those cells will [69:04] start to operate like they're supposed [69:05] to when they were young again. [69:07] >> That's going to be [69:08] >> it's going to be incredible. Yeah. Do [69:10] you think like No, in all seriousness, [69:11] people's knees or joints or what where [69:14] do you think it could flow to next in [69:16] arthritis? [69:17] >> Yep. Um and and by the way, when applied [69:19] and if it's distributed in the skin, [69:21] they've seen some results in monkeys [69:22] where like wrinkles go away. [69:24] >> It like literally makes these cells all [69:26] work youthful again. And so a lot of the [69:28] damage that happens over time is not [69:31] damage to DNA. It's damage to the [69:33] epiggenome. It's the parts that sit on [69:36] top of the DNA that turn genes on and [69:38] off. And they get moved to the wrong [69:39] place as you get older. And by resetting [69:41] them and getting them back to the right [69:42] place, boom, the cell is young again, [69:44] the organ is young again, and suddenly [69:46] you look and act and feel young again. [69:48] It's an incredible technology. We're [69:50] just at the early stages, the early [69:52] innings of turning it into therapeutics. [69:54] Again, the discovery goes back to 2006 [69:57] and now we're starting to see it get [69:58] into clinic. [69:59] >> All right, let's rejuvenate. Let's [70:00] rejuvenate some hairlines on this [70:02] podcast. Uh, that would be next up. [70:04] >> Speak for yourself. [70:05] >> I don't know. You got You got a little [70:07] uh You got a little peaks going there, [70:08] my brother. A little peak. [70:09] >> What are you talking about, bro? My [70:10] hairline's incredible. I'm 50. [70:12] >> I mean, it's not bad for 50. I give you [70:14] credit. You're You're holding your own. [70:15] All right, let's talk about our final [70:16] topic. Scotas struck down Trump's [70:19] emergency powers tariffs. Last Friday, [70:22] Scotas voted 63 against President [70:25] Trump's tariffs. Six judges voted [70:28] against. three conservatives, Roberts, [70:30] Barrett, Gorsuch, uh, and three liberals [70:33] via Bloomberg. This is the biggest [70:35] rebuke of existing executive policy in [70:39] 91 years since Scottish struck down [70:41] FDR's first new deal in 1935. [70:45] UPUP Wharton Analysis says the tariffs [70:48] collected about 175 billion to date, 50% [70:51] of all tariff duties, um, might wind up [70:54] being refunded. This is going to take [70:57] some time to sort out in the courts. [70:59] 2,000 importers have already filed for [71:01] refunds. We we talked about it here. I [71:04] think the majority of people felt like [71:05] this is the way the decision would go. [71:08] And we talked about here that there were [71:10] other options for President Trump to [71:12] pursue. He immediately said he was not [71:16] deterred and invoked a 15% global tariff [71:18] across the board via section 122 of the [71:21] 1974 [71:23] trade act. Here's your poly market. Will [71:26] the court force Trump to refund tariffs? [71:29] 18% chance but spike to 40% after the [71:32] SCOTA's decision. How will Congress [71:34] react? Poly market says 3% chance. [71:37] Congress passes any tariffs by March [71:40] 31st. So again, as I referenced earlier, [71:44] these two sides just can't seem to work [71:46] together. And that would have resolved [71:48] the whole thing. Saxs, you want to um [71:51] give us your take here? [71:52] >> First of all, I don't think that the [71:53] tariffs are going away. What the court [71:56] basically indicated especially the [71:58] 70page Kavanaaugh descent is that [72:00] there's multiple alternative bases in [72:02] law for the tariffs in existing law. So [72:06] for example section 122 of the trade act [72:10] of 1974 [72:12] enables temporary 150day tariffs of up [72:16] to 15% to address balance of payments [72:19] issues and the president has already [72:21] invoked this. So we are now operating [72:23] under that. What the 150 days is going [72:25] to do is buy the administration time to [72:28] substantiate via studies and agency [72:32] reviews. What it needs to prove in order [72:34] to invoke more sweeping tariff authority [72:38] under section 301 of the trade act and [72:40] under section 338 of the tariff act. [72:45] Section 301 authorizes tariffs [72:46] responding to unfair foreign trade [72:49] practices. Section 338 of the tariff act [72:52] allows tariffs against countries [72:55] discriminating against US commerce. The [72:57] Kavanaaugh descent actually provided a [72:59] roadmap for the administration to put in [73:03] place tariffs using one of these [73:05] alternate uh bases. So, I think that one [73:08] way or another, the tariff policies of [73:09] this administration and the favorable [73:12] trade deals that they allow us to strike [73:13] with many nations, they will continue. [73:15] And I think the court seems to know that [73:18] because the majority's opinion as [73:20] concurrences collectively said nothing [73:23] about how the administration should go [73:24] about refunding the tariff revenue [73:27] already collected. I think that if they [73:29] expected this decision to end the tariff [73:31] policies altogether, they probably would [73:33] have said something about that. And I [73:36] think that brings up a really important [73:37] point just on the merits here, which is [73:40] why would we want to give back hundreds [73:42] of billions of dollars to a bunch of [73:44] importers when we're trillions of [73:46] dollars in debt? And I'll just say that [73:48] the people who originally predicted that [73:50] somehow these tariffs would be [73:51] catastrophic for the economy. Those [73:53] predictions all prove not to be true. So [73:56] I think that this is ultimately, I [73:58] think, going to be a popular policy. The [74:00] administration will figure out a [74:01] different way to do it. And I predict [74:02] that future administrations, whether [74:04] they're Republican or Democrat, will [74:07] keep some version of the tariffs in [74:08] place, but I think that they will be [74:10] ultimately popular on a long-standing [74:12] basis. Chamathy, your thoughts? [74:15] >> I think we've proven the experiment has [74:17] been successful. [74:20] >> What was the experiment? [74:22] We needed to smoke out what the right [74:25] balance of trade should be between the [74:27] United States and all of its partner [74:29] countries. I think that what we [74:31] uncovered is that for the most part they [74:33] were structural imbalances that were [74:36] made not because they made economic [74:37] sense for America, but it was just part [74:40] of a hodgepodge of globalist dril that [74:44] people just bought into. And if you [74:45] strip all that stuff away, we had a [74:48] hollowedout manufacturing class and we [74:50] have a hollowedout middle class and the [74:52] tariffs will create more equality for [74:55] the American worker in the end. So now I [74:57] think the debate should be about how to [74:59] implement these in a structural and [75:01] permanent way. [75:03] I think we talked about this before [75:04] Jason that this was sort of expected and [75:08] >> there are many other mechanisms. I think [75:10] the president activated one of them [75:12] immediately. [75:13] >> I don't think this is going away and I [75:15] don't think it should go away. So I [75:17] think now the point is [75:19] Congress really should ratify these [75:21] things because it is clear that it was [75:23] the right thing to do [75:25] and if they don't then you know the [75:28] president still has a lot of room to get [75:29] these done but these make smart economic [75:31] sense in my opinion. [75:32] >> Freeberg any thoughts on the ruling? [75:34] Does it give you [75:37] some I don't know um respect for the [75:40] courts that they made a judgment not [75:42] along party lines for once? Uh does it [75:45] Yeah. Okay. Exactly. [75:46] >> Yeah. And I think I think that all I [75:48] think all Americans [75:50] should feel assured and comforted in the [75:53] fact that I think a lot of people view [75:55] the Supreme Court as having a high [75:56] degree of partisanship. [75:59] The fact that the president despite [76:01] having a majority of what others would [76:03] think were kind of politically aligned [76:05] appointees on the court had a ruling [76:07] that he did not want. [76:10] I think should give everyone good faith [76:11] that the system that the founders set up [76:14] is working that there is a judicial [76:15] branch that adjudicates the law against [76:18] the executive branch when they think [76:19] that um it doesn't map and I think that [76:22] that was very important to see. So, you [76:25] know, I clearly you the the debate about [76:27] tariffs, the economic effect of tariffs, [76:29] the security structural trade [76:31] relationship effect of tariffs and the [76:33] importance of that is a separate [76:34] conversation, but I do think that the [76:35] read of the law being what I would say [76:37] is nonpartisan with respect to the [76:39] court's action is important and probably [76:42] very valuable. I'll reiterate that this [76:44] is a great moment, I think, for uh the [76:47] Supreme Court to make a thoughtful [76:49] decision. And I think we need to think [76:51] about executive power a whole bunch. [76:53] Whether it's Biden with student loans or [76:55] Trump with tariffs, we have this [76:58] beautiful system set up by the founding [76:59] fathers. I know it's frustrating. [77:01] Gridlock's frustrating. Having to work [77:03] together is frustrating. Trust me, we we [77:05] all come to this podcast every Thursday. [77:08] We have to work together. It's hard to [77:09] work together. You got to learn to work [77:11] together. And we don't want an executive [77:13] branch that can unilaterally just roll [77:15] over the other branches. And uh that's [77:18] going to end. I think Trump's going to [77:19] lose the midterms and we're going to get [77:21] to more chaos again and it we might as [77:24] well start this reconciliation process [77:26] of these two sides stopping their law [77:29] for against each other and working [77:31] together for the American people on the [77:33] important issues. The tariffs there are [77:36] some fundamentally important things that [77:38] Trump was doing there and they were [77:40] working. They could have been chaotic. [77:43] uh that's a reasonable um you know [77:46] criticism of them because business [77:47] owners didn't know what to do. So Trump [77:49] did it in a chaotic way. That's just a [77:50] fact. He should have done it in a more [77:52] thoughtful way and the Congress should [77:54] have been alongside him saying, "Hey, [77:56] what tools do you need? How can we help [77:58] support this? We know that there's trade [78:00] imbalances. We know that people are [78:03] being unfair. Let's work together as one [78:05] America to negotiate these things." So [78:07] both sides start having dinner together, [78:10] start playing cards together, and do [78:11] what we do here on this podcast, which [78:13] is you fight it out. You argue, but then [78:15] you come together and try to find some [78:16] resolutions for this stuff. So they [78:18] should go and tell Trump, "Hey, we'll [78:21] approve all the tariffs you did. We will [78:23] not force you to get refunds." The [78:25] Congress should come out and just say [78:26] that. And then they should say, "Hey, [78:28] and when you want to do them in 2026, [78:30] just run them by us or ask us for some [78:32] parameters that you want and let's just [78:34] be thoughtful about it. These are our [78:35] concerns. That's it. Thank you for [78:37] coming to my TEA talk. Can [78:38] >> I just do one? [78:39] >> Absolutely. I'm sure you have some [78:40] debate club points that you want to [78:42] point that out. [78:42] >> Well, I just want to make one. I mean, [78:43] do you think Susan Rice is going to [78:46] respect your call for comedy and [78:50] basically working together Kumbaya? [78:53] >> I want the Aspree to corpse. No, I [78:54] don't. I think both sides. She just had [78:57] a diet tribe where she basically said [78:58] that Republicans and actually not just [79:02] like partisan Republicans, but even tech [79:04] companies that merely were working with [79:06] the administration should expect to get [79:08] prosecuted. I mean, she was basically [79:10] outright saying she's she and the [79:12] Democrats are going to pursue lawfare as [79:14] soon as they get back in charge. [79:16] >> They're they're absolutely going to do [79:17] that. Just like when Trump got in, he [79:19] went after Comey, he went after Jerome [79:21] Pal. The lawfare is happening on both [79:22] sides. Both sides need to drop the [79:24] lawfare. We need to get rid of these [79:25] pardons. They're ridiculous. And these [79:27] team this, we have to be a team. So, [79:30] let's just get some aspa of corpse and [79:31] teamwork going in Washington DC. And [79:34] that's what we should vote for in the [79:35] midterms. We should vote for moderates [79:37] who want to work together. And in 2028, [79:40] we should have some kind of moderates [79:43] and tickets that want to work together. [79:46] That would be better for all Americans. [79:48] This kind of chaos is not good, folks. [79:50] All right. Listen, this has been another [79:51] amazing episode of the All-In podcast, [79:53] your favorite podcast. Uh, like, [79:55] subscribe, whatever the hell you want to [79:57] do on your own time for David Sachs, [80:00] David Freeberg, Chamal Pia. Love you, [80:03] boys. I am the world's greatest Monterey [80:05] J. See you next time. Byebye. Byebye. [80:09] >> We'll let your winners ride. [80:16] And it said, "We open sourced it to the [80:18] fans and they've just gone crazy with [80:20] it." [80:21] >> Queen of [80:29] >> besties are [80:32] my dog taking notice your driveways. [80:37] >> Oh man, my habitasher will meet up. [80:40] >> We should all just get a room and just [80:41] have one big huge orgy cuz they're all [80:43] just useless. It's like this like sexual [80:45] tension that we just need to release [80:46] somehow. [80:51] >> Your feet. [80:53] We need to get merch. [80:55] >> I'm going all in. [81:03] I'm going all in.