Project · coding agents

Test the instructions, not the prose.

The original Sharper Than AI project is still here. Inspect a public repository, prepare historical tasks, then use the local runner to compare instruction snapshots under controlled conditions.

How the local runner worksView an example
Experiment builder
No signup

Get an instruction inventory, historical PR tasks, and an exportable test plan. Prepares the test; does not run agents.

View example brief

Public repositories only. No signup, cloning, or code execution.