Math
Problem sets and formal proof environments, from competition mathematics to Lean-verified theorem proving, where correctness is machine-checkable. Math offers the cleanest reward signal in RL, which is why it was the first domain where reasoning models showed dramatic gains.
6 startups
| Name | Domains | Location | Team | Funding | Raising |
|---|---|---|---|---|---|
| Axiom Math Self-improving AI mathematician with formally verified reasoning | Math | San Francisco | 11-25 | Series A, $200M at $1.6B valuation (2026); $264M total | - |
| Epoch AI Nonprofit research institute behind FrontierMath and AI capability benchmarks | Math, Machine Learning | Remote | 11-25 | Philanthropic grants (nonprofit) | - |
| Harmonic Mathematical superintelligence via formally verified reasoning | Math | Palo Alto | 26-50 | Series B, $100M at ~$900M valuation (Kleiner Perkins, 2025) | - |
| Hillclimb Math environments emphasizing verifiable correctness | Math | San Francisco | 1-10 | - | - |
| Math, Inc. Autoformalization agents for verified mathematics | Math | Palo Alto | 1-10 | Seed, $15M (Torch Capital, Robot Ventures, Chapter One) | - |
| Ulam Math RL environments and RLVR trajectories | Math | Warsaw, London | 1-10 | - | - |