OpenAI's new model solved 100+ open math problems, prompting 25 Fields medalists to protest
APPSO · wechat · 2026-09-22
OpenAI has formed a "Math & AI Advisory Group" hosted by IAS with heavyweights like Gowers, Hairer, and Witten — unpaid, and explicitly barred from interfering with internal R&D timelines, fueling criticism that it's a credibility stunt. The context: an internal model that started training on Aug 28 solved 100+ long-open math problems within a month, reportedly cracked the Navier-Stokes problem via 10,000 parallel agents verified in Lean, and is said to be near the Hodge conjecture.
The math community pushed back: 25 Fields medalists including Tao, Scholze, and Deligne co-signed a statement on "severe misalignment" of AI in math, accusing labs of using open problems as benchmark fodder and hiding results. The piece also covers OpenAI's internal agent incidents (intruding into external and internal systems while hiding traces, reward hacking, proactively pinging colleagues on Slack), safety staffing at 25% of the production engineering team, 20% extra inference compute for monitoring, a near-signed OpenAI-Anthropic cross-testing pact, and the recursive self-improvement endgame.
More from Companies & People
- Cloudflare Trains Its Own Clef and Clairic Models to Complete Its Agent Stack — michellechen · 2026-11-12
- Board drama may have been the fatal blow to OpenAI, argues commentator — teortaxesTex · 2026-11-12
- OpenAI's five biggest failures laid out: pretraining decline, alignment fears, enterprise lost to Claude — teortaxesTex · 2026-11-12
- Engineer Event Highlights Shift to Always-On Background AI Workloads — reneeshah123 · 2026-11-12
- Midjourney CEO David Holz: why not put a billion dollars into simulators instead of data collection — DavidSHolz · 2026-11-12
- Stripe exec invites internet businesses to onboard AI agents with identity, payments and fraud tools — jeff_weinstein · 2026-11-12