OpenAI's internal model reportedly cracked 100+ open math problems; IAS advisory panel formed
AI寒武纪 · wechat · 2026-09-22
Per a report by AI Cambrian (unverified leak), OpenAI's internal new model has solved over 100 long-standing open math problems, following its earlier reported resolution of the Navier-Stokes equations in early September. The model began training on August 28, and its mathematical reasoning stunned OpenAI's own researchers.
Partnership with the Institute for Advanced Study (Princeton)
- OpenAI and IAS announced an independent Math & AI advisory panel to assess the academic value of these results and set publication norms
- Members include Fields medalists Edward Witten, Timothy Gowers and Martin Hairer, plus mathematicians from IAS, ENS Paris, UC Berkeley, Oxford, Stanford and Harvard
- Panelists take no OpenAI salary, may publicly criticize the company and control their own membership — but cannot restrict OpenAI's internal R&D pace
Context: Terence Tao and other Fields medalists recently published an open letter on "severe misalignment of AI in mathematics," criticizing AI labs for treating open problems as benchmark competitions. The author predicts math and code will be the first domains to reach superhuman capability, iterating on each other.
More from Models
- OpenAI's five biggest failures laid out: pretraining decline, alignment fears, enterprise lost to Claude — teortaxesTex · 2026-11-12
- Fable 5.2 Hits 41.3% on Remote Labor Index, Up From 2.5% a Year Ago — alliekmiller · 2026-11-12
- Beaker automates MIT's weight-space programming: one-neuron edit lifts MATH-500 from 7% to 37% for under $4 — VictorSanh · 2026-11-12
- Researcher says Anthropic dodged questions on Mythos false positive rate; 37,966 findings, 2 exploited — OwariDa · 2026-11-12
- Grok spotted building meeting "Listen to Conversation" feature with live transcript — nima_owji · 2026-11-12
- MoonMath ships uncensored 763B DeepSeek-V4.1 via closed-form weight editing, claims zero capability loss — udmrzn · 2026-11-12