Skip to main content

Price Outcomes, Not Hours

Coding evals are broken. Benchmarks and leaderboards miss what matters: shipped code. Markets fix this by paying the first commit that passes your tests.

Write a test suite, fund escrow, and set policy gates. The commit that clears CI and optionally AI-assisted review claims the escrow.

  • Individuals or teams with an inherent domain edge can earn without revealing their methods
  • Small prompt changes can lead to large differences in tokens required to complete a task
  • A market surfaces the most effective prompt-makers
  • The system only rewards successful outcomes, not vanity metrics

A market-priced, repo-native objective function is a prerequisite for economically sustainable recursively self-improving agents.