423 high-value problems are only the beginning. The world is full of important problems we have not found yet — hidden inside laboratories, hospitals, companies, engineering teams, and entire industries. And often, the people closest to these problems know better than anyone what truly needs to be solved. If you have a real-world problem that matters, bring it to TRACES.
What we look for, what to send us, and the path from a problem statement to an executable world a solver can enter.
A short description of the problem, why it matters, and who is waiting for the answer. Detailed intake requirements are being written — until they are published, start with an email and we will take it from there.
Please send only information you are free to share. Do not include third-party confidential information, or data you are not licensed to share — including data we would need for step 3.
We will work with you to turn your problem into a challenge AI can pursue. Every accepted problem goes through the same build.
Turn an open-ended ambition into a well-defined question: the objective, the decomposition, and what success actually means.
Recover the path a specialist actually takes: the tractable steps, and how evidence moves from one to the next.
Assemble the domain-specific data the work depends on, and make it legible and accessible to a solver.
Select and wire in the specialized instruments the problem requires, so the solver can compute, query and test.
Stand up the executable world and run it for you, so any solver can enter it and work the problem end to end.
Define how a proposed solution can ultimately be checked: by existing evidence, or by evidence still to come.
Then we put frontier AI to work — evaluating what today’s models can solve, understanding where they fail, and developing better solvers to push further.
The goal is not simply to grow a benchmark. It is to build a community around the problems that matter most — and bring the best of AI to solving them.
“Upgrading exploration from personal gambling to civilizational collaboration; turning the unknown from a forbidden zone into a governable engineering object.”