OpenAI offered 'unlimited compute' to drop a co-author — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI claims to have cracked the Navier-Stokes Millennium Prize problem using roughly 10,000 agents, tens of millions of dollars in compute, and 88 hours of compute time.
- Tristan Buckmaster, an NYU professor, says OpenAI researcher Sébastien Bubeck offered him 'unlimited compute' and sole authorship of the breakthrough paper if he cut collaborator Levent Alpöge — an Anthropic employee — from the project; Buckmaster rejected the offer, calling it a 'bribe,' and turned his office into a 'war room' with colleagues, outreach coordinators, and lawyers.
- OpenAI spokesperson Laurance Fauconnet 'categorically' denied that Buckmaster's Codex prompts over the prior two months influenced the system, though earlier OpenAI statements had acknowledged it could not rule out that data from Buckmaster's product use improved the model.
- Andreas Thom of TU Dresden, whose work with Gábor Kun underpins a separate OpenAI mathematical result the company later quietly credited, told The Verge the unresolved question is whether his ChatGPT conversations about the research informed the models that ultimately tackled his own problem.
- Bubeck acknowledged to The New York Times that OpenAI made similar arrangements with other mathematicians and said the core issue was having an Anthropic employee on what he called an 'internal OpenAI project,' while Alpöge described his work with Buckmaster as a 'personal collaboration' independent of Anthropic.
- Buckmaster, citing prior experience collaborating with Google DeepMind, says tech companies are 'obsessed with scooping' and treat prestige and being first as 'the only currency' — a worldview he says clashes with mathematicians' focus on the lineage of ideas and beauty of discovery.
Why it matters: For mathematicians, the episode crystallizes a deeper problem: when AI labs treat famous open problems as trophies rather than building blocks, the people whose cumulative work underpins breakthroughs get erased from the record. OpenAI has not yet published a peer-reviewed proof of its Navier-Stokes claim, meaning its biggest mathematics win may also become its most disputed.
Ask SkimNews
