Andrej Karpathy tweeted about buying a Mac Mini to tinker with Claws, calling them "an awesome, exciting new layer of the AI stack" — a new orchestration layer on top of LLM agents handling scheduling, context, and persistence. Simon Willison notes "Claw" is becoming a term of art for the entire category of personal-hardware AI agents that communicate via messaging protocols. #1 on Hacker News with 178+ points. The naming matters: when Karpathy gives something a name, the industry tends to adopt it.

Analysis of how AI assistant companies are quietly pivoting to advertising as their core business model. With 158 points on HN, the piece strikes a nerve: if your AI assistant is funded by advertisers, whose interests is it actually serving? The concern isn't hypothetical — it's the logical endpoint of "free" AI services that need to monetize. The assistant that recommends restaurants may soon be the assistant that recommends sponsored restaurants.

The New York Times reports on AI's growing role in the 2026 midterm elections. Anthropic has put $20 million into a super PAC to counter OpenAI's political spending. The AI labs are no longer just building technology — they're directly funding the political infrastructure that will regulate them. This is a new phase: the companies building the most powerful technology in history are now active combatants in the political arena that decides how (or whether) that technology gets constrained.

Dr. Alex Wissner-Gross publishes a new essay looking one year ahead at what will exist by February 2027. From one of the highest-signal AI forecasting newsletters, this is an invitation to think concretely about the near future — not decades out, but twelve months. Given the pace of the last year (Opus 4.5, autonomous agents at scale, the Pentagon vs. Anthropic), the exercise of projecting just one year forward is both tractable and terrifying.

Lean 4, the formal theorem prover, is gaining serious traction as a competitive tool in AI research. Formal verification and mathematical reasoning are becoming increasingly relevant as AI systems tackle more complex proofs and as labs race to demonstrate genuine reasoning capability. The convergence of theorem proving and AI isn't just academic — it's becoming a differentiator for labs that can prove their models actually reason rather than pattern-match.

Jordan Schneider's ChinaTalk investigates the explosive rise of gray market peptides from China's pharmaceutical ecosystem — now a market worth hundreds of millions of dollars. American consumers are injecting research chemicals for weight loss and muscle recovery, sourced from Chinese labs operating in regulatory gray zones. It's a fascinating collision of Chinese manufacturing capability, American demand for biohacking, and the regulatory gaps that emerge when supply chains outpace oversight.