OpenAI claims an AI-generated proof of the Navier–Stokes problem, and it is hard to verify

A month ago, OpenAI announced that it had obtained an AI-generated proof of the Navier–Stokes problem. It is one of the seven "Millennium Prize Problems." The Clay Mathematics Institute offers $1 million for solving each one, and since 2000 only one of them, the Poincaré conjecture, has been closed. Bloomberg Tech has now devoted an episode to the topic with Justin Solomon, associate dean for engineering education at MIT. One important caveat: the piece is a podcast announcement, not an analysis of the proof itself. It contains no confirmation that the proof is correct.

What the problem is and why it is so hard to verify

The Navier–Stokes equations describe the motion of a viscous fluid: water in a pipe, air over a wing, currents in the atmosphere. Engineers and climate scientists solve them numerically every day. But mathematicians have still not proved that in three dimensions these equations always have a smooth solution that does not "blow up" in finite time. The practice works; the rigorous theory does not exist.

According to the episode description, Solomon explains both the problem statement and why OpenAI's text is hard to read even for professionals. For mathematics, this is a crucial point. A proof counts as a proof only after the community has checked it. If a text hundreds of pages long is written by a model and full of unfamiliar moves, verification can take months. A press release cannot speed up that process.

Models that once could not count

The piece includes a detail that conveys the scale of the shift: language models once made mistakes in elementary arithmetic, and now they solve problems that stayed unsolved for decades. Solomon discusses how pure and applied mathematicians react to this. Their interests differ. Applied mathematicians care more about a working tool and a numerical result. Pure mathematicians need understanding, meaning an argument structure that can be passed on to another person.

A separate segment of the episode covers how film portrays the work of mathematicians. Usually it is a genius at a blackboard with a flash of insight in the middle of the night. Real work consists of long verification, dead ends, and correspondence with colleagues.

Key facts from the announcement

• Claim: in September 2026, OpenAI announced an AI-generated proof of the Navier–Stokes problem. • Status: the source mentions no independent verification. • Stakes: the Clay Mathematics Institute prize for solving a Millennium Prize Problem is $1 million. • Expert: Justin Solomon, associate dean for engineering education at MIT. • Problem: even professionals find the text of the proof hard to parse. • Parallel front: LLMs complete students' homework, and instructors cannot adapt fast enough.

What instructors and engineers should do

Education hit this problem before science did. If a model solves standard homework, instructors have to assess the reasoning rather than the answer: oral defenses, problems with explanations, exams without devices. According to the description, Solomon is already rethinking his own teaching approach, but the announcement does not reveal details.

For engineering teams, the takeaway is more down-to-earth. Any model output in mathematics, from optimization to flow simulation, needs independent verification: formal proof systems, numerical tests, or a second specialist. Until the community confirms OpenAI's work, it should be treated as a claim. If it is confirmed, mathematicians will get a new class of tools and a new problem: how to read proofs that humans would not have come up with themselves.