Suspending the Mathathon won’t stop the flood
On the open letter about the Caltech Mathathon
A coalition of current and former Caltech mathematicians has asked for the Mathathon to be suspended. They have quite a few points, but the economics are wrong: the $20,000 of AI credits they say no mathematician could reasonably expect is about what an individual subscription already buys over several months. And the letter assumes that if you don’t welcome AI labs to attempt mathematics, the flood will stop.
Of the recent declarations about AI and mathematics, the Open Letter about the Mathathon is the one aimed at a specific event, which is what makes it interesting (A coalition of current and former Caltech mathematicians 2026). It is a response to this:
On October 30th, 2026, Caltech will begin hosting the Caltech Mathathon, a math hackathon in which participants are encouraged to solve open research problems by prompting LLMs. Two prominent AI companies, Anthropic and OpenAI, will supply participants with 2 million dollars in AI credits.1
And they think
This event is likely to have destructive impacts for the mathematical community.
They have quite a few points, but this statement
There is no reasonable future model of mathematical research in which every research mathematician receives 20000 USD in AI credits to prove a result.
is simply not true. The equivalent amount of tokens is at least an order of magnitude cheaper in an individual subscription. Somebody on OpenAI’s pro tier or Anthropic’s max tier would have a similar token budget over several months: as of September 2026, heavy subscribers are reported to run $1,000–$5,000 a month of API-equivalent usage on a $200 plan, in one case ten billion tokens over eight months — more than $15,000 at list API rates — for $800 in subscription fees.2
The response from the organizers is here (Arulandu et al. 2026).
I think fundamentally, this rests on the assumption that if you don’t welcome AI labs to attempt mathematics, then the flood will stop (cf. Gowers’ view on the flood). I’m not sure it would: the labs will attempt mathematics whether or not they are welcome, and the letter says as much itself:
Major AI companies are engaged in an arms race to declare themselves the first to prove prominent open conjectures. New announcements of AI-generated results appear on social media on a weekly basis.
If the announcements are already weekly, suspending one hackathon would not change that.
One way I see it is that this experiment is much better designed than OpenAI irresponsibly dropping the Navier–Stokes result (see progressive disclosure). So in the view of “AI labs will continue to use math as a benchmark for their models to hill climb”, then if you can’t beat them, join them: engage with them and shape how AI can and should be used in mathematical research.
References
Footnotes
OpenAI withdrew its sponsorship on 10 September 2026, the same day the letter was published, citing “concerns raised by members of the mathematics community”. It had been supplying $10,000 of each team’s $20,000. Anthropic and the other backers remain, and the totals on the event site have been changing since, so the figure quoted above is the one the letter was written against, not the current one.↩︎
These are third-party measurements; neither lab documents what a subscription buys in tokens.↩︎