In early September 2026 OpenAI posted a 166‑page draft that it said could solve the Navier‑Stokes problem, one of the Clay Mathematics Institute’s seven million‑dollar Millennium problems. The draft is written in Lean, a formal verification language that checks each logical step automatically. OpenAI said the proof followed the method of Spanish mathematicians Diego Córdoba and Luis Martínez‑Zoroa.
Mathematicians see a mismatch
Several mathematicians, including Gonzalo Cao‑Labor of EPFL, have examined the draft and say the reasoning does not match the Córdoba‑Martínez‑Zoroa approach. They describe the similarities as only surface‑level. The Clay Mathematics Institute has been reviewing the draft since it appeared.
Massive compute, unclear impact
OpenAI ran about 10,000 model instances for roughly 88 hours, processing 130 billion tokens to generate the draft. Researcher Noam Brown estimates that less than 10 % of the result can be attributed to the multi‑agent setup, suggesting most of the work came from the base model itself. The exact role of the agent swarm remains unknown because systematic tests at that scale have not been done.
Further releases draw criticism
In early October OpenAI added roughly 400 more AI‑generated results to a loose GitHub collection, which many mathematicians said cheapened the field. The rapid output has left the math community struggling to keep up.
Why it matters
For a regular reader, the episode shows that even massive AI resources can produce work that experts still dispute. It also highlights that AI‑generated math proofs are not yet a shortcut for human verification, so the prize‑winning problem remains unsolved.
