Mathematicians and experts are questioning whether OpenAI truly understands its own 166-page potential solution to the Navier-Stokes problem, which was presented in early September. Several groups of mathematicians told Scientific American that the proof actually takes a different approach than the one claimed by OpenAI, with EPFL mathematician Gonzalo Cao-Labora noting the similarities are only surface-level. The potential proof was written in Lean, a formal verification language, and has been under review by the Clay Mathematics Institute since its release.
Before publication, NYU mathematician Tristan Buckmaster accused OpenAI of building on work he and Anthropic's Levent Alpöge did with public models, an allegation OpenAI denied. Camillo De Lellis of the Institute for Advanced Study stated that a definitive answer is likely impossible because language models remain black boxes. In early October, OpenAI further angered mathematicians by publishing roughly 400 additional loose results on GitHub that the community can barely keep up with.
OpenAI also admitted uncertainty regarding the role of its massive compute deployment, which used about 10,000 model instances for roughly 88 hours and 130 billion tokens. OpenAI researcher Noam Brown stated that nobody actually knows what the agent swarm contributed, as running comparison tests at that scale is too expensive. Brown attributes less than 10 percent of the success to the multi-agent setup, emphasizing that the base model's ability to grind through tasks in parallel is what truly matters. Why it matters: This controversy highlights the growing friction between rapid AI-generated mathematical outputs and the rigorous validation standards required by the human scientific community.