Scooped by OpenAI, Buckmaster and Anthropic's Alpöge Answer With a Navier-Stokes Blowup Proof
Two months after OpenAI got there first, N.Y.U.'s Tristan Buckmaster and Anthropic's Levent Alpöge are back with a new AI-aided proof that the fluid equations blow up - and it now faces the only test that counts.

N.Y.U. professor Tristan Buckmaster and Anthropic employee Levent Alpöge have produced a new AI-aided proof of Navier-Stokes blowup, roughly two months after OpenAI scooped them on the same problem, Scientific American reports. The result, if it survives independent scrutiny, resets the theoretical floor under every numerical fluid simulation in aerospace, weather, combustion and beyond, and hands Anthropic a reputational asset that no product launch can buy.
Two months after OpenAI got there first, Tristan Buckmaster and Levent Alpöge are back with a new AI-aided proof of Navier-Stokes blowup. Buckmaster is a professor at N.Y.U. Alpöge is an Anthropic employee. Those two job titles are why this is more than a math story: a university department and a frontier AI lab are jointly attached to a result at the outer edge of pure mathematics, and they arrived there roughly eight weeks after being beaten to the same target. Scientific American's headline on the work is blunt: "AI just solved the 'real' Navier-Stokes problem." Note the scare quotes around "real" - they signal a qualifier, that there is some adjacent version of this question in circulation that is not the full thing. Either way, the result is a second swing at a summit someone else reached first.
The Navier-Stokes equations are how physics writes down moving fluids. Air over a wing, water through a pipe, smoke off a candle, the weather over an ocean: same set of equations, different inputs. Engineers solve them numerically every day and get answers good enough to design aircraft and forecast storms. The open question sits underneath all that usefulness. Do smooth, well-behaved starting conditions always produce smooth solutions that last forever, or can the mathematics itself break down, with velocity running to infinity in finite time? That breakdown is what mathematicians call blowup. A proof of blowup says the equations permit a singularity, that the smooth world is not guaranteed.
This is one of the most heavily fortified problems in mathematics. Navier-Stokes existence and smoothness is on the Clay Mathematics Institute's list of seven Millennium Prize Problems, posted in 2000 with $1 million attached to each. Exactly one of the seven has fallen: the Poincare conjecture, solved by Grigori Perelman, who then declined the money. The other six have stayed shut for a quarter century, and Navier-Stokes is the one physicists care about most, because it is the one describing things they can see. Work at this altitude historically moves in years, not weeks.
Which makes the two-month turnaround the genuinely interesting part. Getting scooped is a specific kind of sting for a mathematician, and the standard response is to disappear for a while. Buckmaster and Alpöge instead came back with a new proof of blowup. Scientific American frames it as AI-aided, which is the phrase that will be repeated in every citation that follows. The scoreboard now reads OpenAI first, N.Y.U. and Anthropic second, both entries dated inside a single season. That is a pace the field is not built for.
The claim now has to survive the only test that matters in mathematics, which is other mathematicians. Proofs at this level are typically long, technical, and checked line by line over months, sometimes years, by people who take a professional pleasure in finding the hole. Some celebrated arguments have collapsed under that scrutiny. Others have held. Until the checking lands, what exists is a claim and two institutions with their names on it. Anyone repeating "AI solved Navier-Stokes" as settled fact this week is running well ahead of the evidence.
Here is why a pure-math result should register outside the seminar room. Every simulation an aerospace firm runs, every weather model a national agency operates, every combustion study inside an engine maker, rests on numerical solutions to these equations. Practitioners have always proceeded as if smooth solutions stay smooth long enough to be useful. A blowup proof changes the theoretical ground under that practice: it says the equations themselves allow a singularity, and that eternal smoothness is not something you get to assume for free. That does not invalidate the simulations. It redraws the boundary of what they can be claimed to represent.
The second signal is about instrumentation. Two months between a scoop and a counterpunch is not normal for a field where a single lemma can eat a year. If AI assistance genuinely compresses the search for a proof, the binding constraint in elite mathematics shifts from generating ideas to verifying them, and verification is slow, manual, and done by a small number of people worldwide. Labs that have staff working at this level now hold an asset that never shows up in a product launch. Anthropic's name sitting next to a Navier-Stokes result is a recruiting advertisement aimed at precisely the graduate students every AI company wants to hire.
What to watch, in order: independent verification, whether the argument survives contact with people who have spent careers on this problem, and whether the technique generalizes to other open questions. For operators and boards, the practical takeaway is that frontier AI is now producing artifacts in domains where correctness is binary and reputation is the whole balance sheet. That raises questions most companies have not written a policy for. Who owns a result when a model contributes? How is credit assigned when a lab employee and a professor publish together? What happens to the partnership if the claim is later challenged? The mathematics will be settled by mathematicians. The governance will be settled by whoever moves first.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.
Anthropic mathematician says AI proved the Riemann hypothesis. Verification will take time.
A single unverified claim from inside Anthropic now sits against more than 160 years of open mathematics, and the gap between asserting a proof and surviving peer review is the entire story.


