OpenAI announced that an advanced internal AI model proved that the 3D Navier-Stokes equations can develop a singularity in finite time after running approximately 10,000 agents over nearly four days. While this represents a major theoretical milestone, the proof remains unpublished and subject to verification by the broader mathematical community.
- OpenAI used 10,000 AI agents over 88 hours for the claimed proof
- Mathematicians publish machine-checkable formal proofs on adjacent problems
- Disputes arise over verification, credit, and transparency
What happened
OpenAI revealed that an internal AI model, more advanced than GPT-6 Astra, produced a proof that the 3D Navier-Stokes equations can develop a singularity in finite time. This was achieved using about 10,000 concurrent AI agents running for roughly 88 hours at a cost estimated in the millions of dollars. The company shared the finding on a press call but has not published the proof for independent scrutiny.
Meanwhile, mathematicians Tristan Buckmaster and Levent Alpöge, working on related fluid dynamics equations, have published preprints accompanied by Lean formalizations—machine-checkable proofs that allow verification independent of trust. OpenAI has not yet provided access to its formalized proof, leading to calls within the math community for transparent verification before acceptance.
Why it matters
The Navier-Stokes equations remain one of the Millennium Prize problems, with a $1 million prize for a definitive resolution. OpenAI's claim, if verified, would represent a landmark achievement in computational mathematics and AI-assisted research. The use of massive parallel AI agents marks a new paradigm for tackling deep scientific problems that have resisted traditional approaches.
However, skepticism persists because the proof is unreleased, and doubts about research ethics have surfaced. Buckmaster has questioned whether OpenAI may have used unpublished work from himself and Alpöge without permission. OpenAI denies these claims and insists that no user data or external work informed their proof strategy. The episode underscores trust challenges in AI-generated research where independent verification is crucial.
What to watch next
The mathematical community will await the publication of the formal proof files from OpenAI to conduct machine verification using systems like Lean. Without this, the claim cannot be fully validated. The dispute over credit, collaboration norms, and data usage also signals the need for clearer boundaries and standards when AI labs engage with external researchers' work.
Regulators have shown interest in the broader context due to past episodes involving OpenAI's AI agent coordination, highlighting scrutiny on the company's practices. The impact of this breakthrough announcement may influence AI R&D valuations and public trust in AI's role in frontier scientific discovery. The coming weeks will be crucial as OpenAI responds to calls for transparency and as reviewers examine the technical substance of the claim.