OpenAI spent part of this week suggesting it had done something extraordinary in mathematics: a formalization related to the Navier–Stokes existence and smoothness problem, one of the Clay Institute’s million-dollar questions. The artifact, as described by people who have looked, is a single commit. The metadata says the work is self-assessed.
That is not how million-dollar problems usually end. They end with papers, referees, and years of arguing about a lemma on page forty. A machine-checkable proof could, in principle, skip some of that. It cannot skip the part where independent mathematicians read it.
Twenty-five Fields Medalists have already signed an open letter about how OpenAI has been using the field’s prestige. The letter and the commit landed in the same news cycle, which is either a coincidence or a communications strategy. Either way, the claim is now public, and the only way out is through the proof.
If the formalization holds, it is one of the strongest arguments yet that these systems are doing something that looks like science. If it does not, it will join a growing stack of demos that were scored by the people who built them. The industry is very good at that kind of scoring. Mathematics is not.
