feat: run evals on the human
Every AI tool scores the model. Nothing scores the person typing.
So I built one. It reads my own Claude and Codex chats, on my Mac, and gives me a number.
Current build: 73/100
Failing tests: Hand over, Check
Turns out "looks good, ship it" is not a test suite.
korture.ai
KORTURE.aiKORTURE.ai SenseAs agents do more of the work, know whether you're still good, and keep getting better at the part only you can do.
commit 602ddbd