Clay Mathematics Institute says Navier-Stokes is "apparently" settled as AI-linked proof faces review

CMI seems to be starting the clock for scrutiny while treating the problem as presumptively solved. It is sensible that they are staying out of the credit dispute and the Fields Medalists' open letter.

Waiting until the drama cooled down and issuing a neutral statement was smart. The wording is so sterile that it never even says "OpenAI."

The phrase about "new technologies" accelerating mathematical research is doing a lot of work here. I like the combination of vagueness and confidence.

"Apparently" is load-bearing in CMI's line that Navier-Stokes has been settled. They are clearly leaving room for the result to fail review.

CMI's rules require publication in a qualifying outlet and then two full years before it will consider a Millennium Prize solution. A self-published OpenAI proof has not started that clock.

A Lean proof is available to download, and checking the Lean statement is much easier than reading the whole proof. But I missed the qualifying-outlet requirement, so publication status still matters.

The rule is not unusual. The 1906 Wolfskehl prize for Fermat's Last Theorem also required two years after publication.

The qualifying outlet appears to mean a peer-reviewed math journal, so OpenAI may not be eligible until roughly 2029. arXiv and a company website do not automatically qualify under the written rules.

CMI retains sole discretion over what counts as a qualifying outlet and may relax conditions after advice from experts that a published solution is likely correct. That leaves room for CMI to handle an unconventional publication.

OpenAI says it will not claim the prize anyway. I think the real issue is the disputed credit: OpenAI says its work is separate, while critics say both efforts use the same smooth-forcing mechanism.

CMI may decide not to award the prize if the contribution cannot be clearly attributed to individual mathematicians, especially if AI made the main contributions. The rules appear to allow that outcome.

I am uneasy with huge prizes attached to arbitrarily difficult problems because they invite competition and credit fights. AI also changes the older model where one person could spend decades alone on a hard proof.

Neither of the two Millennium-problem solvers has claimed the prize money, so the money plainly is not the main reason people tackle these problems.

I see mathematics becoming more like engineering, and I think that is overdue. Mathematicians and other scientists will have to adapt, much as manual craftspeople did during industrialization.