OpenAI Announces Navier–Stokes Solved and Creates a Mathematicians' Committee
In a single post: an internal model claimed to have cracked more than 100 open problems, and an independent panel of experts to manage the shockwave.
In brief
OpenAI claims that an internal model, trained starting August 28, solved the Millennium Prize problem on Navier–Stokes equations along with more than 100 long-standing open problems in mathematics. In the same post, the company announces the creation of an independent advisory group made up of nine leading mathematicians, tasked with evaluating and orchestrating the communication of these results. The committee, however, will have no say in the pace of OpenAI's internal progress.
🍺 Bar-stool version
OpenAI casually mentions that one of its in-house models solved one of the seven Millennium Prize problems and about a hundred other questions left open for decades, then pivots straight into announcing a committee. It's a bit like announcing you've found a universal vaccine in the same press release where you unveil your new PR service. The committee, for its part, brings together very serious names — a Fields Medalist, a Witten — but its mandate explicitly excludes telling OpenAI to slow down. If the results hold up, this is one of the most important moments in the history of mathematics; and for now, there's nothing to actually read.
Key takeaways
- 1
OpenAI says it began training a new internal model on August 28, and claims it has since solved the Navier–Stokes Millennium Prize problem.
- 2
The same model reportedly solved 'more than 100' long-standing open problems across most areas of mathematics, at a pace that surprised the company's own mathematicians.
- 3
No proof, publication, or technical detail accompanies the announcement: the post is about governance, not results.
- 4
An independent advisory group brings together François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood.
- 5
Its members are not paid by OpenAI, can make their opinions public, can issue them unprompted, and can modify the group's composition themselves.
- 6
The mandate excludes one specific thing: advising OpenAI on the pace of its internal progress in mathematics.
- 7
The announcement partly responds to the open letter 'A Severe Misalignment of AI in Mathematics,' which criticizes the use of open problems as a benchmark.
What OpenAI Claims
The post opens with a sentence that, if confirmed, would alone justify a decade of press coverage: a new internal model, whose training began on August 28, reportedly solved the Millennium Prize problem on Navier–Stokes equations. This is one of the seven problems carrying a million-dollar prize from the Clay Mathematics Institute.
OpenAI adds that this same model has since solved 'more than 100' long-standing open problems, spanning most branches of mathematics. The company states that the pace of this progress surprised its own mathematicians.
Nothing in the post substantiates these claims: no proof, no preprint, no list of the problems involved, no reviewer names. This is precisely the gap the committee announced later in the text is supposed to fill.
The Committee, Its Mandate, and Its Limits
Nine mathematicians make up this independent advisory group: François Charles (ENS-PSL), Camillo De Lellis (IAS, GSSI), Timothy Gowers (Collège de France, Cambridge), Martin Hairer (EPFL, Imperial College London), Nikhil Srivastava (Berkeley, Simons Institute), Ulrike Tillmann (Oxford, INI), Ravi Vakil (Stanford), Edward Witten (IAS), and Melanie Matchett Wood (Harvard). The lineup is heavyweight: at least two Fields Medalists, and central figures in analysis, algebraic geometry, and theoretical physics.
Their role, as described: help OpenAI assess the significance of emerging results, advise on coordinating their dissemination, and uphold the academic and professional standards of mathematical research. The group is also meant to guide how OpenAI's tools can serve research and education.
The independence guarantees are explicit and unusually precise: no payment from OpenAI, freedom to make opinions public, freedom to issue unsolicited advice, freedom to comment on OpenAI's impact on mathematics, and control over its own composition.
One exclusion is stated in black and white: the group will not advise OpenAI on the pace of its internal progress. In other words, it can speak on how to communicate, but not on the decision to accelerate or not.
The Open Letter in the Background
OpenAI itself cites the open letter 'A Severe Misalignment of AI in Mathematics,' in which mathematicians denounce the negative externalities of a practice that has become common: using open problems as a validation benchmark for new AI systems.
The criticism is not corporate posturing. An open problem is not a unit test: its value lies in collective construction, peer verification, and deliberate slowness. Turning it into a leaderboard score changes the nature of the activity, and can flood the community with alleged proofs that are impossible to review at scale.
OpenAI acknowledges that these criticisms 'highlight the need for thoughtful engagement from AI companies with the mathematical community.' The committee is presented as the response — and as 'a first step.'
What This Says About OpenAI's Strategy
Structurally, the post is strange: a potentially historic scientific announcement serves as the introduction to a governance announcement. One could read this as caution — not publishing results without community validation — as a way of circulating the claim while deferring its verification to later.
Setting up an outside committee, unpaid and free to publish, is a real concession. It transfers to recognized mathematicians a share of the credibility the company cannot claim on its own in this domain. It's also a gamble: if the committee publicly disputes the results, OpenAI has itself armed them to do so.
What remains is the red line on pace. A group that can comment on everything except the speed at which things move is as much a communications device as a counterweight. That's the axis on which, in the coming months, we'll need to judge what this 'first step' is really worth.
“In addition to resolving the Navier–Stokes Millennium Prize problem, this model has now resolved more than 100 long-standing open problems across most areas of mathematics.”
“Its value depends on its members being able to exercise their own judgement and challenge ours.”
“Importantly, the group will not be responsible for advising us on how to pace our internal progress on mathematics.”
Why it matters
If the claims hold up, we're talking about the first Millennium Prize problem solved by a machine and about a hundred open results falling in a matter of months — a shift for a discipline whose unit of time is the decade. But nothing is verifiable at this stage, and that's the whole issue: OpenAI is announcing an extraordinary result without proof, while simultaneously creating the body meant to arbitrate its credibility. The setup is clever and its independence guarantees are real (no payment, right to publish, right to criticize). Still, the scope excludes the one question that actually matters, the pace, and the mathematical community is being offered a seat at the communications table rather than the decision-making one. Worth watching closely: the committee's first public statement will be worth far more than the post announcing it.
Read next
Jev: the model built to be called by code, not by humans
Diogo Almeida, ex-OpenAI, explains why he created a new class of "system one" models — and why he rejects public benchmarks, refusals, and pre-training.
Source · Latent Space · Why We Made Jev — Diogo Almeida, TypeSafe Co-founder & CEO
99% or 0%: Four Experts Bet on the End of the World
Steven Bartlett gets four specialists to write their extinction probability on an envelope. The answers range from 99% to zero, and the debate that follows is the most useful we've heard on the topic.
Source · The Diary Of A CEO · AI Emergency: The AI Labs Are Lying To Everyone, He Says 99% Chance Of Extinction | Roman Yampolskiy
#agiTodayOpenAI Wants Global Standards Before AI Self-Improvement Takes Off
The lab is calling for an international technical framework to govern the moment AI starts doing AI research on its own — and wants the US to hold the pen.
Source · OpenAI · Building standards for the next phase of AI