A coalition of Caltech mathematicians announces Mathathon's new theme, Old Problems, New Proofs, asking participants to spend 40 hours learning and two months rewriting proofs while disclosing LLM use

The statement shifts Mathathon's theme from open problems to alternative proofs and expositions of already-solved problems, under the banner Old Problems, New Proofs. Previously Mathathon centered on problem solving; the statement says organizers, coauthors of the Open Letter, and community members reached a consensus that problem solving is only one component of mathematics, and that mathematical understanding and exposition are similarly meaningful yet often overlooked by grantmakers and hiring committees. This is a position and format statement from the organizers, grounded in the internal discussion and consensus described in the text rather than in experiments or statistics; the four color theorem, the ABC conjecture, and the Navier-Stokes problem appear only as examples of proofs that many mathematicians find difficult to understand or unsatisfying.

The format is specified as 40 hours of learning plus peer presentation, followed by two months developing an alternative proof or exposition, then submission of an explainer and a GitHub repository. The repository must store whatever was used to produce the explainer, including LLM chat histories and code used to produce visualizations, so the mathematics community can examine the ideas behind the finished product; the explainer can take any form, such as a paper, blog post, video, or interactive game. This is a description of event rules taken from the statement itself; the text gives no participant counts, completion rates, or measures of output quality from past editions.

On tools and funding, Mathathon no longer receives sponsorships from developers of proprietary AI models and instead gives participants a cash grant to purchase any tool they need, whether LLM credits, HPC access, or pen and paper. The statement says proprietary models are still allowed but open-source tools are encouraged, with the goal of empowering each participant to decide for themselves which tools to use at Mathathon and in their own work—a change from accepting sponsorship from proprietary model developers. This is an organizational policy statement; the text also notes that SAIR provides compute, hosts workshops, offers prizes to teams using open-source models, and administers travel grants, with XTX Markets as lead donor.

The statement positions Mathathon as a first experiment rather than a conclusive guide to AI in mathematics. It states explicitly that organizers hold different opinions about AI-assisted problem solving and about open-source versus proprietary models, while all believing mathematicians can use AI to advance human understanding; the event requires participants to fully disclose and justify LLM usage and to consider whether smaller-scale open-source tools would suffice. This is a stance and self-positioning statement; the text reports no experimental results, benchmark data, or quantitative comparison of AI mathematical capabilities.

This statement addresses students and mathematicians who may take part in Mathathon, along with community members interested in AI's role in mathematics, explaining that the event takes place November 13–15, is co-organized with SAIR, is led by donor XTX Markets, and setting out the participation process and tool-funding arrangements. It is meant for understanding the event's positioning and rules and for reference when designing similar activities; the claim that mathematics is undergoing its biggest change in decades is the organizers' judgment rather than a data-based conclusion of this text.

The text was read incompletely and is an event announcement, so there are no verifiable results, participation scale, or evaluation metrics; the event's actual effects, the quality of participants' outputs, and the real mix of open-source versus proprietary model use can only be observed after the November 13–15 event. In addition, the statement is written by the organizers and says it was reviewed by Terence Tao and Andrew Wu, but the text does not describe the scope or conclusions of that review.