OpenAI's Navier-Stokes win draws scooping accusations — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI announced Tuesday that an unreleased model solved the Navier-Stokes problem—one of seven Millennium Prize problems carrying a $1 million bounty—in 88 hours by deploying a swarm of roughly 10,000 AI agents powered by an internal model.
- NYU professor Tristan Buckmaster alleged OpenAI rushed the problem after learning he and Levent Alpöge (an Anthropic researcher acting independently) were making progress, and that an OpenAI researcher told him "Why would you ruin your career?" when he threatened to go public.
- OpenAI denied accessing any specific user data to solve the problem, but acknowledged it "cannot rule out" that de-identified data from Buckmaster's Codex usage may have helped train its models—while stressing the two proofs differ significantly.
- OpenAI said it spent millions on the effort but does not intend to claim the $1 million bounty; researcher Sébastien Bubeck told reporters the company began the push after hearing Millennium Prize rumors "on Twitter," only later realizing they concerned Alpöge and Buckmaster.
- Mathematicians at Queen Mary University of London, Carnegie Mellon, the University of South Carolina, and Brown University warned the episode could chill the informal trust norms that allow researchers to share incomplete ideas, with ICARM's Jeremy Avigad calling even the thought that "AI systems might steal ideas from our queries" chilling.
- Oxford professor Andras Juhasz called the announcement a "PR victory" for OpenAI but questioned its sustainability, noting that "suddenly, 10,000 mathematicians jump on your problem"—a scale of competition no individual researcher can match.
Why it matters: OpenAI's own admission that it cannot rule out training on user Codex data means every query a mathematician types could now feed models racing to scoop them, potentially chilling the open exchange of unfinished ideas that mathematics depends on. One frontier lab can now deploy computational firepower equivalent to thousands of researchers against any publicly buzzed-about problem.
Ask SkimNews
