Student
Undergraduate and master's students, with mentors assigned to every team.
Warsaw · Dates announced soon
Europe's first hackathon built for research-level mathematics.
The problems are revealed when the clock starts. Teams work through two nights with frontier AI, serious compute and Lean, hand in a report when it stops, and then defend what they found before a jury.
Every team works with frontier models and a compute budget. The point is to measure what they can do on real problems, so the event does not use weaker substitutes.
Results can be formalised in Lean, and a Lean track rewards formal work. Workshops before the event bring newcomers up to speed.
After the stop, teams present to jury panels organised by field. The jury reads the report, asks questions and decides.
Every problem starts with a scientist who needs the answer. Pure research problems in a model's scaffold tend to go either instantly or nowhere. Questions from physics, biology or transport have a reason to exist and a person who will use the result.
The scientist who owns the problem works with mathematicians to turn it into a precise statement with definitions, known results and data.
Frontier models are run on it first. A problem stays only if they make progress without solving it outright.
Each brief says what a good result looks like in forty hours, at three levels.
The open square is a problem nobody has solved. A team fills it only with something another person can verify: a proof, a counterexample, a computation that reruns, ideally all of it in Lean.
Undergraduate and master's students, with mentors assigned to every team.
Doctoral researchers in mathematics, computer science or any of the sciences.
Everyone else: postdocs, researchers, engineers, people from industry.
Teams have up to four people. A team competes in the division of its most advanced member, and any team may choose Open.
In Warsaw, in spring 2027. The exact dates go to the waitlist first.
No. Lean helps and is rewarded, but strong mathematics and good use of compute count on their own. Lean workshops run before the event for anyone who wants to start.
A problem with its brief, access to frontier models, a compute budget, a ready Lean environment, mentors and the problem's scientist to talk to.
The teams do. Results the jury accepts are meant to be published openly, with their Lean code. [Licence and authorship rules to be confirmed]
Yes. If you are a scientist with a question mathematics could settle, write to us. We will work with you to turn it into a problem for the event.
Functor Foundation, a research foundation in Warsaw. Warsaw Mathathon is an independent event with no formal ties to other mathathons.