All-In: Software Stocks Implode, Claude’s Hit List, State of the Union Reactions, Trump’s Tariff Pivot

概览

This episode centers on AI as a market-moving force: Anthropic announcements, coding agents, and a viral AI-doom Substack are discussed as catalysts for sharp repricing in software, financial, legal, and legacy IT stocks.

The hosts debate whether AI will destroy SaaS cash flows or expand demand through Jevons Paradox. Chamath emphasizes a market shift from asking when cash flows degrade to whether they survive; Sacks pushes back with software-engineer hiring data and argues that cheaper software creation could increase total software demand.

The second half moves from AI infrastructure to politics: data center opposition, Trump’s State of the Union, tariff legal strategy after a Supreme Court ruling, and a Science Corner segment on Yamanaka factors and aging reversal.

分段落总结

[00:00] Opening and docket

[事实] The hosts joke about conspiracy topics, solo projects, and the All-In AI bot selecting “Dario versus Hegseth” as the top topic. [事实] Jason says the episode will cover Claude’s expanding “kill list,” an AI fan-fiction Substack that moved markets, State of the Union reactions, tariffs, and Science Corner. [推测] The opening frames the episode as a mix of market analysis, AI speculation, and political argument rather than a single-topic discussion.

[01:29] Claude’s “kill list” and software-stock repricing

[事实] Jason lists Anthropic-related announcements that coincided with selloffs in legal software/data companies, security companies, and IBM after Claude’s COBOL modernization announcement. [事实] Chamath says hedge funds appear to be “de-grossing,” reducing long and short exposure and putting downward pressure on risk assets. [事实] Chamath argues the bigger structural shift is from asking when cash flows weaken to asking if AI could make those cash flows disappear at all. [推测] The group treats AI announcements as catalysts that expose fragile valuation assumptions, even when the business impact is not yet proven.

[07:40] Viral AI-doom Substack and market fallout

[事实] Jason describes a fictional 2028 “global intelligence crisis” post that received 28 million views on X and predicted an AI-driven economic death spiral. [事实] The scenario claimed companies would cut staff with AI, boost margins, shrink consumers’ spending power, and keep cutting until unemployment and equity-market losses surged. [事实] Jason says financial stocks fell after the post speculated that AI agents and stablecoins could eliminate interchange fees. [事实] Sacks questions whether the post’s virality was organic after attribution was reportedly amended to include a short fund connected to names mentioned in the report.

[10:22] AI analysis versus science fiction narratives

[事实] Sacks cites Derek Thompson’s “Nobody Knows Anything” response, arguing that serious AI macro debates often resemble competing science-fiction narratives because real-time evidence is thin. [事实] A prediction market cited in the conversation put the Citrini scenario at roughly 12%. [推测] The hosts use the episode’s first AI debate to separate plausible uncertainty from overconfident forecasting.

[12:02] Why SaaS became newly uncertain

[事实] Sacks explains that SaaS used to be easy to model through ARR, net dollar retention, RPO, renewals, seat expansion, and upsells. [事实] He says investors treated SaaS like a growth annuity, but AI now raises questions about growth, pricing models, and category durability. [事实] Sacks says he does not believe AI will eliminate Salesforce, but it could reduce growth opportunity or change pricing. [推测] The key investor problem is not only lost revenue, but the loss of confidence in SaaS as a predictable financial category.

[14:23] Productivity, consumption, and knowledge work

[事实] Friedberg says AI increases leverage on human time and capital, but asks whether productive capacity could exceed consumptive capacity. [事实] He suggests SaaS may have been a transitory business form between the internet and AI, and knowledge work may also be transitory between computing tools and AI. [事实] Sacks challenges this as another speculative narrative and asks for data. [推测] This section poses the episode’s deepest macro question: whether AI creates abundance absorbed by demand, or output that the economy cannot distribute.

[18:05] Counter-data and Jevons Paradox

[事实] Sacks points to an Anthropic software-engineer job listing at $570,000 as evidence that software engineers are not obviously obsolete. [事实] He cites a Citadel Securities report saying software-engineer postings are rising by roughly 10% year over year and company formation is expanding. [事实] Sacks and Aaron Levy’s argument is that lowering the cost of a constrained input can increase demand for that input. [推测] The optimistic case is that AI makes more people and companies capable of building software instead of simply reducing headcount.

[22:20] AI agents inside Jason’s firm

[事实] Jason says his firm trained 15 of 20 people over six or seven hours to build and use “open claw” agents. [事实] He describes agents that analyze podcast advertisers, connect to Gmail/calendar/Zoom/Notion/Slack, summarize employee work, create clips, and improve YouTube thumbnails. [事实] Jason says the firm is not adding people, but existing employees are becoming 10% to 20% more efficient each week. [推测] His examples make the AI impact concrete: agents are automating unfinished internal software projects and tasks that might previously have justified hiring.

[27:56] Cash discipline and SaaS value capture

[事实] Chamath says tech companies may need years of cash runway to experiment through AI disruption and should examine stock-based compensation. [事实] Jason says agent-built internal tools could weaken SaaS vendors’ upsell leverage because customers can threaten to build features themselves. [事实] Sacks says it remains unclear whether model companies, application companies, or chip companies capture the most value. [推测] The discussion implies that vertical SaaS companies must prove defensibility as foundation models and open-source alternatives improve.

[30:39] Doom bias, new jobs, and token constraints

[事实] Sacks argues dystopian AI narratives are more appealing because people can see threatened jobs more easily than future jobs or business models. [事实] Jason identifies a new role: people who create, train, and manage agents by translating business processes into agent workflows. [事实] Sacks says change management, enterprise inertia, token supply, chips, land, power, and energy may slow both utopian and dystopian AI scenarios. [事实] Chamath expects token demand to 10x and output-token prices to fall by roughly 90%, while noting token costs are now part of employee cost modeling.

[35:45] Data centers as the AI bottleneck

[事实] Chamath says around 100 data center projects face local opposition and estimates canceled capacity could represent large lost revenue using OpenAI CFO Sarah Friar’s gigawatt-to-revenue framing. [事实] Sacks warns that political opposition to data centers could constrain AI infrastructure buildout. [推测] The hosts treat data centers as the physical layer of AI competition, where local permitting decisions can affect national economic upside.

[40:36] Ratepayer protection, energy, and permitting fights

[事实] Sacks says Trump supports a ratepayer protection pledge requiring major tech companies to provide power for AI data centers so residential electricity rates do not rise. [事实] Friedberg argues data centers can be placed globally because data moves near the speed of light, so blocking them domestically may shift economic value elsewhere. [事实] Chamath warns that utility business models can still push prices higher through capex plans, independent of data center burden-sharing. [事实] The hosts discuss Micron’s New York fab delays, environmental lawsuits, lithium development opposition, Greenpeace damages, and the need for broader public benefit-sharing.

[52:13] State of the Union reactions and polarization

[事实] Jason says Trump’s State of the Union lasted 108 minutes and cites falling approval numbers on the economy and trade. [事实] Chamath’s highlights include Ilhan Omar and Rashida Tlaib’s reactions, Democrats not standing for law-and-order or citizen-priority lines, and Brad Gerstner being mentioned. [事实] Sacks argues Democrats failed easy consensus tests by not applauding statements about citizens, border security, political violence, violent criminals, and prescription drug prices. [事实] Jason argues both parties are polarized, Trump is combative, and Congress and the executive branch should work together on issues like tariffs and immigration.

[64:42] Science Corner: Yamanaka factors

[事实] Friedberg explains that Yamanaka factors are four proteins discovered by Shinya Yamanaka that can reset epigenetic markers and make mammalian cells more youthful. [事实] He says David Sinclair’s Life Biosciences reached an FDA agreement to begin a Phase 1 human trial delivering Yamanaka factors into the eye for blindness from glaucoma or stroke-like eye disease. [事实] The method described packages DNA into an AAV virus delivered to retinal cells, with protein production switched on or off using doxycycline. [推测] If the trial works, Friedberg sees it as an early step toward broader regenerative or anti-aging therapeutics, though he acknowledges Sinclair’s controversial reputation.

[70:21] Tariff ruling and Trump’s fallback path

[事实] Jason says the Supreme Court voted 6-3 against Trump’s emergency-powers tariff approach, with Roberts, Barrett, Gorsuch, and three liberals in the majority. [事实] He says UPenn Wharton estimated around $175 billion in collected tariffs might be refunded and that 2,000 importers had filed for refunds. [事实] Jason says Trump invoked a 15% global tariff through Section 122 of the 1974 Trade Act after the ruling. [事实] Sacks argues tariffs will not disappear because Section 122, Section 301, and Section 338 provide alternate legal bases.

[75:46] Courts, executive power, and lawfare

[事实] Friedberg says the ruling should reassure Americans that the Supreme Court can act independently of perceived partisan alignment. [事实] Jason says executive power should be constrained whether the issue is Biden’s student loans or Trump’s tariffs. [事实] Jason criticizes Trump’s tariff rollout as chaotic but says Congress should approve the existing tariffs, avoid refunds, and create a more orderly process for 2026. [事实] Sacks raises Susan Rice and Democratic lawfare concerns; Jason responds that lawfare happens on both sides and calls for moderates and cooperation.

播客点评/总结

[推测] The episode’s strongest value is its concrete look at how AI changes both market narratives and operating workflows. Jason’s agent examples make abstract SaaS disruption feel practical, while Sacks’ Jevons Paradox framing gives the optimistic countercase.

[推测] The limitation is that several claims are deliberately speculative: AI macro outcomes, data center revenue-loss estimates, tariff popularity, and future legal paths are debated from participant viewpoints rather than settled evidence.

[推测] This episode is best suited for listeners interested in venture capital, public-market repricing, AI productivity, data center policy, and U.S. political economy. It is less useful for listeners seeking neutral policy analysis without partisan argument.