← Back to articles

Caltech Mathathon – first hackathon ever devoted to research level mathematics

2026-09-07 | original

read original get Rhodia Dot Grid Notepad → more articles

Why This Matters

Caltech is hosting what it bills as the first hackathon aimed at research-level mathematics, giving roughly a hundred teams 40 hours and over $2M in AI credits to attack open conjectures and defend their results before working mathematicians. It's a public test of how far frontier models have come in pure math and how fast AI-assisted work can move from idea to verifiable, peer-reviewed result. The event's two-stage prize structure, with a second round after community verification, acknowledges the core problem of trusting machine-generated proofs.

Key Takeaways

Caltech's Mathathon (Oct 30–Nov 1) is framed as the first hackathon devoted to research-level mathematics, with frontier models and $2M+ in AI credits provided to ~100 teams.
Teams must defend results before leading mathematicians, testing human understanding, not just output — and prizes are split between the event and a later verification round.
The organizers cite recent claimed AI-assisted breakthroughs on decades-old problems (one still unverified) to ask what a mathematician's role becomes when AI can solve conjectures faster.

Worth a Look

Rhodia Dot Grid Notepad — Even when frontier AI models are drafting proofs, someone still has to sketch the diagram and check the argument by hand. Rhodia's dot grid pads are a favorite for mathematicians because the light dots guide equations and figures without boxing you in, and the sheets tear off cleanly to pass around a team table. Perfect for a 40-hour push at an open conjecture.

See Rhodia Dot Grid Notepad on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.

Update: we have changed the event date to October 30th.

40 hours, $2M+ AI credits,

solve an open problem.

AI has made groundbreaking progress in pure math in the past few months:

May 20, 2026: Erdos's planar unit-distance conjecture was disproved (open 80 years) .

. August 1, 2026: The first explicit non-sofic group was constructed (open 27 years) .

. August 23, 2026: The six-sphere was shown to admit a complex structure (open 78 years; unverified).

This AI advancement raises the following questions: (a) How much can AI speed up the process from ideation to peer-reviewed publication? (b) What is the role of a mathematician when AI can solve conjectures faster?

We are assembling the top math talent worldwide to answer these questions. On October 30th, a hundred teams will be given frontier models to solve open conjectures and build new mathematical theories. Then, they will defend their results before leading mathematicians, who will assess their understanding of the results.

We will award prizes to the most promising and well-explained results at the event. Then, after the math community has had time to verify these results, we will provide a second round of prizes.

... continue reading

Explore topics: caltech mathathon hackathon open problems frontier models ai in mathematics