Mathematician

Frontier models still fail at rigorous multi-step mathematics: invalid derivations, unjustified leaps, and confidently wrong final answers. Automated checks can't grade a proof. You can.

As a Mathematician you'll evaluate and correct mathematical reasoning across everything from competition problems to graduate-level material, and your judgment becomes the standard the models learn from.

• Craft problems that expose weak mathematical reasoning, from olympiad-style puzzles to graduate-level proofs.
• Review AI solutions step by step: catch invalid derivations, flag unjustified leaps, and rate the quality of the reasoning.
• Write the gold-standard solution when the model falls short, with the rigor of a referee report.