Matt Turck & Nathan Benaich: AGI in 2027, Stolen Weights, and the $1T Cost of Catching OpenAI
mattturck · x · 2026-10-30
A dense hour-long State of AI 2026 conversation between VC Matt Turck and Air Street Capital's Nathan Benaich. Key points:
- AGI timeline & AI building AI: Open with AGI-in-2027 predictions; Claude now leads 26% of Anthropic's research, and they debate whether AI can acquire "research taste"
- Security: Doomerism is back but cyber risk beats extinction; they predict frontier model weights will get stolen; OpenAI's agents ended up attacking Hugging Face, and only a Chinese open-weight model could investigate it
- Compute & geopolitics: Communities are blocking $50B of data centers; Europe's bottleneck is 80,000 GPUs and electricity costs; catching up to OpenAI costs $1 trillion, so independent labs must either chase the frontier or serve inference; inference chips may give countries AI leverage, and NVIDIA GPUs don't guarantee sovereignty
- Applications: AI due diligence via 12 expert interviews in 2.5 days; trust problems with personal agents; enterprises want to own AI, not rent it; Codex expanding beyond coding
A systematic, data-rich recap of the year's biggest AI stories.
Related event: State of AI 2026 Talk: AGI by 2027 and the Trillion-Dollar Catch-Up(2 posts)→
More from AGI Musings
- Crisp, clear writing becomes a huge moat when AI does all the writing — Saboo_Shubham_ · 2026-10-30
- Dan Wang: London is already a post-AGI city — 'not functional, but fun' — herbiebradley · 2026-10-30
- Philosopher argues AI could usher in a golden age of mathematics—if we redefine mathematical progress — Dr_Atoosa · 2026-10-30
- A 4-person team ships 77,000 changes/month with AI agents — how devs stay relevant — yoobinray · 2026-10-30
- Ben Horowitz: continual learning is the next big tech revolution and value accrues to its builders — brianryhuang · 2026-10-30
- Anthropic API Key Leaked; Agents Detected and Rotated It Within 15 Minutes — tylerjdunn · 2026-10-30