Overview
- OpenAI announced on Sept. 8 that an unreleased internal model produced a proposed proof showing a finite-time singularity for the three-dimensional Navier–Stokes equations after about 88 hours of work by roughly 10,000 coordinating AI agents.
- The company released a technical paper and a machine-checked Lean formalisation of the argument and said it does not intend to claim the Clay Institute’s US$1 million prize if the result is confirmed.
- Mathematicians emphasize the claim still needs independent verification to determine whether the argument meets the exact Millennium Problem formulation and to survive peer review and the Clay Institute’s two-year community acceptance window.
- A credit dispute has emerged after related AI-assisted work by Tristan Buckmaster and Levent Alpöge was posted days earlier; OpenAI denies seeing their unpublished drafts but concedes it cannot fully rule out indirect effects from de-identified training data.
- The episode highlights a new research model that uses large agent fleets, huge compute costs and automated formal verification, which could speed discoveries but raise questions about attribution, data provenance and how humans validate machine-generated math.