Each exercise is a PR: title, description, diff. Most hide the mistakes LLMs actually make — off-by-one pagination, verify_exp=False in a JWT refresh, runaway agent tool loops. Flag the buggy lines, or approve with LGTM: judging a clean diff correctly pays full XP. Miss a bug and it "ships" — incidents, 3am pages, and postmortems land on your dashboard. 244 exercises across Python, React/TS, Go, SQL, and LLM agents. No signup, free, progress lives in your browser.
