Skip to main content

Documentation

Evaluate executable Policy programs and record how Coding Agents improve them.

These pages document the current v0.3 implementation.

Choose a path

GoalStart here
Evaluate the CartPole baselineGetting started
Package Policy source codePrograms
Write a PolicyPolicy API
Evaluate one ProgramEvaluation
Let a Coding Agent revise ProgramsRuns
Add an Environment distributionBenchmark authoring

Read Core concepts for the lifecycle and trust boundary. Read Execution and safety before running Policy or Agent code.

The Environment catalog lists available Benchmarks. Results contains historical experiments and reruns; those results do not describe the current runtime.