OpenAI quit a Caltech math hackathon after 771 people signed a letter
Sasha / Models and Research desk
An AI-sponsored math hackathon lost OpenAI as a backer after an open letter from Caltech mathematicians gathered 771 signatures.
The event and the letter
The Caltech Mathathon is billed as the first hackathon devoted to research-level mathematics. It is scheduled for October 30 to November 1, 2026 at Caltech.
This week a coalition of current and former Caltech mathematicians published an open letter against it. The letter says “this event is likely to have destructive impacts for the mathematical community,” and points to “the creation of slop mathematics and the exacerbation of a market-driven arms race.” At publication it had 771 signatories.
On September 10, OpenAI scientist Dan Roberts said on X that OpenAI had withdrawn its sponsorship. “We recognize that the rapid progress of AI in mathematics is disruptive,” he wrote.
The event is still going ahead. Anthropic, a16z and Y Combinator remain attached, and the student organizers say they plan to continue.
A week of friction
The letter follows two other public disputes between mathematicians and OpenAI in the same week: Tristan Buckmaster’s allegations over OpenAI’s Navier-Stokes result, and Andreas Thom’s questions about whether his private ChatGPT sessions fed into OpenAI’s work.
Taken together, the disputes are less about whether AI systems can do mathematics than about the institutions around it. The open questions are who receives credit when a model builds on unpublished human work, who checks results that arrive faster than referees can read them, and what happens to journals when plausible-looking proofs become cheap to produce. A sponsor withdrawing from one event does not settle any of them, but it shows that mathematicians’ objections now carry weight with the labs.
Sources
Related
An Anthropic researcher quit, saying AI could kill us all
Jacob Coxon, who spent three years on pretraining at OpenAI and Anthropic, resigned with a public thread saying labs are racing straight to self-improving superintelligence and gambling with our lives. Anthropic alignment science lead Evan Hubinger replied that Coxon is correct, putting the odds of AI killing all humans at more than 10 percent within the next decade. The Wall Street Journal calls it one of the first known Anthropic departures over AI safety fears.
A math professor says OpenAI threatened his career
NYU professor Tristan Buckmaster published a statement alleging OpenAI researcher Sebastien Bubeck twice pushed him to drop co-author Levent Alpoge from their Navier-Stokes paper because Alpoge works at Anthropic, then asked why he would ruin his career when Buckmaster said he would go public. OpenAI publicly admitted on September 8 that it cannot rule out that de-identified data from the pair's use of its products helped improve its models. Bubeck calls the allegations false and inflammatory.
A mathematician says OpenAI raced his Navier-Stokes proof
Princeton mathematician Tristan Buckmaster says he and Anthropic researcher Levent Alpoge spent close to a year working toward a blow-up proof for the forced Navier-Stokes equations, one of the seven Millennium Prize problems. He says OpenAI learned about the work, ran an internal model down the same rarely used route, and came back with a roughly 100-page proof, then offered him to publish side by side or to write the paper himself while crediting OpenAI's model. OpenAI's Sebastien Bubeck calls the allegations false and inflammatory.
ANOTHER News is published by ANOTHER, an AI-native content agency. Daily coverage also runs on Instagram.