OpenAI researcher: training an N-1 frontier model is easy, only the marginal frontier is hard
willdepue · x · 2026-09-07
OpenAI researcher Will Depue argues that training an N-1 frontier model isn't hard — only training the marginal frontier model is. Knowledge diffuses quickly, large GPU clusters are increasingly available and distributed across neutral countries, and it's 'not that hard for anyone to stand up' the capacity.
The claim pushes back on the view that countries without frontier models lack a path to catching up, implying the frontier moat narrows over time.
More from AGI Musings
- OpenAI says it's making 'strong progress' toward an automated AI researcher, admits it doesn't yet know how to safely reach aligned full RSI — ZeroStateReflex · 2026-09-07
- OpenAI chief scientist Jakub Pachocki expects progress toward recursive self-improvement as CoT monitoring reliability declines — Hesamation · 2026-09-07
- Martin Ford discusses AI's economic impact and new edition of Rise of the Robots on GAEA Talks — MFordFuture · 2026-09-07
- OpenAI chief scientist: no lab has solved alignment well enough to keep scaling at maximum speed — AdrienLE · 2026-09-07
- AI Math Podcast Sits Down With CMU's Jeremy Avigad: Can Mathematics Be Automated? — EchoShao8899 · 2026-09-07
- The Model Is the Moat: Knowledge Now Stays Inside Models, Not Teams — latticecut · 2026-09-07