What happened
WIRED reports that OpenAI said it used tens of thousands of AI agents to solve the 90-year-old Navier-Stokes existence and smoothness problem, which carries a $1 million prize. The solution still requires independent verification. Mathematician Steven Strogatz says the development could transform pure mathematics, while also warning that corporate competition may be encouraging labs to prioritize public breakthroughs and headlines.
WIRED reports that OpenAI said on Tuesday it had used tens of thousands of agents to solve the Navier-Stokes existence and smoothness problem, one of the Clay Mathematics Institute’s Millennium Prize Problems. The source says the proposed solution builds on a strategy developed by Spanish mathematicians Diego Córdoba and Luis Martínez-Zoroa, and that the result still needs independent verification.
The report also describes a credit dispute. New York University mathematician Tristan Buckmaster says OpenAI moved quickly after learning about work he had done with Anthropic researcher Levent Alpöge on three closely related problems. Those claims, including the suggestion that OpenAI tried to influence attribution, are presented as Buckmaster’s claims and are not independently confirmed in the source.
WIRED places the development alongside Anthropic’s reported use of Claude to prove 29,500 small theorems while formalizing an existing proof of Fermat’s Last Theorem, and OpenAI’s reported August advances on 10 other long-standing mathematical problems. The source does not provide independent test results or technical details sufficient to assess these claims directly.
Why it matters
If independently verified, the reported result would mark a consequential advance in AI-assisted mathematical research. It also exposes difficult questions about how mathematical credit should be assigned when AI systems build on human work, whether human experts can continue to understand machine-generated proofs, and how research funding and careers may change. WIRED’s account makes clear that the significance remains unsettled because the solution has not yet been independently verified and the dispute over contribution is unresolved.
The reported activity suggests that AI systems are moving from assisting with routine mathematical work toward contributing to difficult research problems. Strogatz says AI may make breakthrough mathematics inaccessible to researchers who lack such tools, while his collaborator Alex Townsend describes feeling threatened by systems that can surpass his peak research abilities.
The implications extend beyond speed. Human mathematicians may increasingly be responsible for explaining and validating machine-generated proofs, but the source questions how long that role will remain distinct. It also raises concerns about whether AI-generated results will preserve the human judgment used to decide which mathematical questions are valuable or aesthetically meaningful.
The source presents possible benefits as well as risks: AI could broaden participation in mathematics and make previously impractical research feasible. However, those benefits depend on reliable verification, transparent attribution, and continued funding for human experts.
What to watch next
The key next steps are independent verification of OpenAI’s claimed solution, clarification of how the work relates to earlier contributions, and any decision on the $1 million prize. Further details about the agents, proof, and review process would help establish whether this is a reliable research breakthrough or primarily a demonstration of AI capability. No public user access, product availability, or pricing is documented in the source.
Independent mathematicians’ assessment of the Navier-Stokes solution is the most important unresolved issue. The source does not say when that review will be completed or identify a completed public verification.
The handling of the $1 million prize and recognition for Córdoba, Martínez-Zoroa, Buckmaster, Alpöge, and OpenAI would establish how institutions assign credit when AI systems contribute to a proof.
More technical disclosure about the agents’ methods, the proof’s formal status, and the division between human and machine work would help distinguish a validated mathematical advance from a capability demonstration.
The source does not document a product launch or general access to the systems involved. It also provides no pricing, availability, independent benchmark, or evidence that the reported methods can be reproduced by researchers outside the companies.