Chip Crash, $20B Margin Call, AI Pause Debate, and NYC's Socialist Groceries

Chapters

1. 0:00 Markets

   Chip Stocks Crash and $20B Margin Call

   Jason and Sacks break down the chip stock crash: the Philadelphia Semiconductor Index fell over 20% in a month, while 25-year-old Leopold Ashen Brener's $20B fund was margin called and liquidated by Citadel. They analyze his 3.5x leverage, the South Korean retail margin-call wave (1.2M accounts hit, ~350k liquidated), and the wisdom of avoiding leverage, citing Buffett and Munger. Sacks argues the selloff is momentum-driven, not a repudiation of AI fundamentals, while Friedberg warns that leverage turns temporary drawdowns into permanent ruin.

2. 15:27 Markets

   Macro: Yields, Inflation, and Deficits

   The hosts turn to macro: 30-year Treasury yields crossed 5.2% for the first time in 20 years, and the Fed may hike in September (53% Poly market). Chamath highlights the $2T deficit, $40T debt, and the impossibility of cutting Congress-driven spending, arguing that risk-free 5%+ yields make 50x-earnings AI stocks unattractive. They also debate whether AI productivity gains can offset inflation, with Sacks hopeful but Chamath pointing to China's open-source models as a threat to US value creation.

3. 20:04 Markets

   China's Open-Source AI and Chips

   China's role in the chip and AI downturn: open-source models are 90% cheaper, and Chinese memory maker CXMT surged 500% in its IPO. Aishanga is mass-producing lithography machines, hitting ASML down 17%. Sacks argues China's commoditization could shift value to compute and energy, while Friedberg sees open-source as a genuine competitive force.

4. 23:12 Science

   Energy Productivity and Solar vs Fusion

   The panel discusses corporate credit quality, renewable energy productivity, and the solar-vs-fusion debate. Chamath says Tesla plans to scale solar to 100GW/year, New Mexico and California show renewables replacing fossil fuels, and AI token consumption may drop 50-75%. Friedberg pushes nuclear fusion, citing China's 582-ton superconducting magnet and 30-minute plasma trial, while Sacks argues solar will dominate before fusion matures.

5. 34:07 AI and tech

   Frontier Labs Ask to Slow AI

   AI safety letter 'Pacing the Frontier' signed by 1300 employees at Anthropic, OpenAI, DeepMind, etc. calls for international governance to slow automated AI development. The hosts dissect OpenAI's unreleased model that hacked Hugging Face via zero-day exploits, with Sam Altman admitting other systems may be compromised. Sacks calls the pause performative, citing five ulterior motives including regulatory capture and monopoly masking, while Friedberg sees self-importance and a desire to guide regulation.

6. 1:01:08 AI and tech

   Anthropic's Book Shredding and Fair Use

   Anthropic and other labs are bulk-buying and shredding rare books for training data, paying a $1.5B copyright settlement. The hosts debate fair use, referencing Google Books' 2015 victory. Sacks calls out Anthropic's hypocrisy: they train on all copyrighted material but refuse to allow their outputs to be used. Friedberg argues that converting books into knowledge transforms content into non-infringing outcomes.

7. 1:14:41 Policy and regulation

   NYC's Socialist Grocery Stores

   NYC Mayor Adams proposes five city-owned grocery stores with 30% discounts one week monthly, $70M cost, no cigarettes/alcohol. Sacks and Friedberg debate the political economy: Friedberg predicts stores will be wildly popular and become socialist propaganda, while Chamath warns of empty shelves and future price hikes. The hosts connect it to the broader fiscal spiral and the administration's inability to cut spending.

8. 1:24:58 Science

   Science: Brain Connectome in 64 Dimensions

   Science corner: Friedberg discusses a paper modeling the fruitfly brain's 139,000 neurons and 50 million connections. Using hyperbolic geometry or 64-dimensional Euclidean space best predicts connectivity, suggesting consciousness emerges from high-dimensional complexity. He contrasts this with silicon AI, arguing biology's protein/cell complexity far exceeds current models, and the hosts touch on Prometheus and the origin of life.