IMPLICATOR
OpenAI withdrew its sponsorship of the student-led Caltech Mathathon on Thursday, September 10, after members of the mathematics community objected to the event. An open letter opposing the contest listed 771 signatories when it was published that day. The decision removes one provider of computing credits from a competition intended to put frontier AI models in mathematicians’ hands.
Dan Roberts, an OpenAI research lead, wrote that the company had notified the organizers and acted after “concerns raised by members of the mathematics community.” He added, “We recognize that the rapid progress of AI in mathematics is disruptive.”
As of Saturday, September 12, OpenAI was absent from the Mathathon’s sponsor list. Anthropic, DARPA’s expMath program, a16z, Y Combinator, Cognition, Cloudflare, Prime Intellect and roughly two dozen other backers were still listed.
What Changed
- OpenAI withdrew its Caltech Mathathon sponsorship on Thursday, September 10, after an open letter that listed 771 signatories at publication asked organizers to suspend the event.
- As of September 12 OpenAI is absent from the event's sponsor list, while Anthropic, DARPA's expMath program, a16z, Y Combinator, Cognition, Cloudflare and Prime Intellect are still listed.
- The letter accused AI companies of a campaign of scientific misinformation and said, 'To put it bluntly, AI companies are engaging in research misconduct.'
- Organizers say the hackathon proceeds. OpenAI funded $10,000 of each team's $20,000 in AI credits, and whether other firms replace that share is unresolved.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
The letter’s objections
A coalition of current and former Caltech mathematicians published the letter on September 10 after it was received on September 8. Signatories must belong to the Caltech community or be research mathematicians at the PhD level or above. The total of 771 is the letter’s count at publication, and the document remains open for signatures, making that figure a snapshot.
FREE · ABOUT FIVE MINUTES
Follow the fight over AI in research math.
Implicator cuts the flood to the developments worth your attention, with concise reporting on what changed and why it matters. Delivered every weekday at 4:45 a.m. Pacific, 7:45 a.m. Eastern.
From San Francisco. No spam. Unsubscribe anytime.
“AI companies see research mathematics as an advertising opportunity,” the letter said. It accused the companies of conducting a “campaign of scientific misinformation” and declared, “To put it bluntly, AI companies are engaging in research misconduct.”
The signatories asked organizers to suspend the Mathathon, warning that a rush of AI-generated work would create “slop mathematics” and push verification labor onto researchers without compensation or credit. They also wrote that giving every research mathematician $20,000 in computing credits could not serve as a reasonable future model for the field.
By September 11, a separate letter had 25 signatories, all Fields Medal winners. It said rushed AI announcements can arrive before a proper writeup or citations to prior work, creating attribution and plagiarism problems.
A dispute in the background
Two days before the withdrawal, OpenAI said an internal model running roughly 10,000 agents had produced a Navier-Stokes blowup proof in 88 hours. New York University professor Tristan Buckmaster said the company began work after learning of research he was conducting with Anthropic researcher Levent Alpöge. OpenAI denied seeing their work. The dispute followed the Leiden Declaration on AI and mathematics, endorsed by the International Mathematical Union in June 2026.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
The event goes ahead
The Mathathon is organized by Caltech undergraduates and billed as the first hackathon for research-level mathematics. Organizers said the event is student-led and sponsors “don’t have a say in our decisions.”
Its first round is scheduled for October 30 through November 1 on the Caltech campus. The plan calls for 100 teams to work for 40 hours with at least $20,000 in AI credits per team. The September 10 letter said OpenAI and Anthropic had together promised $2 million in credits.
In a September 11 statement, a Mathathon spokesperson said, “OpenAI specifically sponsored 10k of the 20k,” and that other firms were discussing a similar contribution per team. Whether those talks replace OpenAI’s share is unresolved. The spokesperson said, “We do not anticipate that this will affect the event in any substantial way.”
Winners of the first round are due to receive additional credits for six months, followed by a six-month judging and verification period. Roberts wrote that OpenAI would seek more discussion with mathematicians about integrating the technology and communicating its effects.
Frequently Asked Questions
Why did OpenAI withdraw from the Caltech Mathathon?
OpenAI research lead Dan Roberts said the company acted after "concerns raised by members of the mathematics community" and had notified the organizers. He added that OpenAI recognizes the rapid progress of AI in mathematics is disruptive.
What is the Caltech Mathathon?
A hackathon organized by Caltech undergraduates and billed as the first devoted to research-level mathematics. Its first round runs October 30 through November 1 on the Caltech campus, with 100 teams working 40 hours using at least $20,000 in AI credits per team.
What did the open letter say?
A coalition of current and former Caltech mathematicians asked organizers to suspend the event, warned it would produce "slop mathematics" and push uncompensated verification labor onto researchers, and said AI companies are engaging in research misconduct. It listed 771 signatories at publication and remains open for signatures.
Is the event still going ahead?
Yes. Organizers say the event is student-led and that sponsors "don't have a say in our decisions." A spokesperson said OpenAI sponsored $10,000 of each team's $20,000 and that other firms are discussing a similar contribution, adding, "We do not anticipate that this will affect the event in any substantial way."
How does this connect to OpenAI's Navier-Stokes claim?
Two days before the withdrawal, OpenAI said an internal model running roughly 10,000 agents produced a Navier-Stokes blowup proof in 88 hours. NYU professor Tristan Buckmaster said the company began that work after learning of his research with Anthropic researcher Levent Alpoge, which OpenAI denies seeing.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
