OpenAI's Navier-Stokes Win Draws Mathematician Backlash — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI claimed it cracked the Navier-Stokes Millennium Prize problem in 88 hours using roughly 10,000 AI agents and "tens of millions of dollars" of compute, after hearing that other researchers were closing in on the same problem.
- Tristan Buckmaster (NYU) and Levent Alpöge (an Anthropic researcher) were both pursuing Navier-Stokes; Buckmaster told The Verge that OpenAI researcher Sébastien Bubeck offered him "unlimited compute" and sole authorship of OpenAI's paper — provided he cut Alpöge out — which he called a "bribe."
- OpenAI spokesperson Laurance Fauconnet "categorically" denied that Buckmaster's Codex prompts could have influenced the system in any way, though Buckmaster said he had been using Codex during his own work and remains "unconvinced."
- Sébastien Bubeck acknowledged making similar authorship-and-compute arrangements with other mathematicians but declined to name them, and confirmed to The New York Times that Alpöge's Anthropic affiliation was "a sticking point."
- Andreas Thom (TU Dresden) found that OpenAI built a major math result heavily on work by him and Gábor Kun, quietly amending its acknowledgment only after press attention; he told The Verge the company "probably don't even know" whether ChatGPT conversations about the research fed back into model training.
- Of the seven Clay Mathematics Institute Millennium Prize problems — each carrying a $1 million bounty and set in 2000 — only one (the Poincaré conjecture) has been solved in the 25 years since, making Navier-Stokes a uniquely high-profile target for any lab seeking a "win."
Why it matters: OpenAI's Navier-Stokes sprint was framed as proof of frontier AI reasoning, but the real story is procedural: mathematicians told The Verge the company pressured competitors, dangled authorship to isolate a collaborator, and quietly built on outside work. If a single well-resourced AI lab can scoop a Millennium Prize problem and sideline the people who developed the underlying ideas, credit norms in mathematics get rewritten by whoever has the biggest compute budget.
Ask SkimNews
