World's first math hackathon at Caltech: 100 teams, 40 hours, $2M in AI credits
FinanceYF5 · x · 2026-09-07
A team led by Sathvik Redrouthu, with backing from Anthropic and OpenAI, is hosting what it calls the world's first research-level math hackathon (Caltech Mathathon), October 30–November 1, 2026.
Format and highlights:
- 100 top teams recruited globally; candidates include holders of 10 IMO/IOI medals and researchers with 15+ Annals papers.
- Each team gets $1M in Fable API credits and $1M in Astra API credits to solve an open problem of their choice in 40 hours, then defend the result orally before expert judges.
- The main track is judged by independent scholars without sponsor involvement; sponsors like Cognition, Cloudflare, and PrimeIntellect run side tracks ranging from number-field sieve algorithms to superalignment.
- Responsible-AI rules: all AI conversations logged and open-sourced, oral defense required, results published to MathDB for community review; prizes split in two rounds, the second after verification in early 2027.
- Backers include Anthropic, OpenAI, DARPA expMath, Cloudflare, Modal, a16z, YC, 8VC, Baseten, and more.
Context: the site cites recent AI-driven pure-math breakthroughs — a disproof of Erdős's 80-year-old planar unit-distance conjecture and the first explicit non-sofic group — framing the event around how fast AI can compress ideation to peer review.
Related event: Caltech to Host World's First Research-Grade Math Hackathon(4 posts)→
More from Companies & People
- AI futures org Mechanize reportedly acqui-hired by Google — himanshustwts · 2026-09-07
- Indie AI Educator Trains ~100 Internet Bank Engineers, 20+ Already Knew His Work — AlchainHust · 2026-09-07
- Google didn't cut staff, it compressed a two-year roadmap into three months, insider says — victor_explore · 2026-09-07
- Ex-NeurIPS SAC: decline one invite and you're never asked to serve again — andrewgwils · 2026-09-07
- Caltech Mathathon sets fairness rules: all AI conversations open-sourced for math community review — FinanceYF5 · 2026-09-07
- Researcher calls undisclosed AI safety incident 'very bad,' disclosure excuse absurd — eliebakouch · 2026-09-07