Platform
One platform for the actual loop of AI-assisted engineering.
Five surfaces, one rubric: defend your decisions. Take a challenge, replay your session, review the official solution, climb the leaderboard, ship in a competition.
Platform · Surfaces
Five surfaces, one rubric.
Every part of the platform feeds the same scoring loop — time-to-correct-fix, 30-day regression rate, and a blind defensibility score.
Challenges
Production-shaped challenges across debugging, security, refactoring, and feature build. Easy / Medium / Hard. Sourced from real codebases.
Replay System
Every session is recorded: prompts, redirects, reasoning. Re-watch your own decisions or step through a panel review.
Official Solutions
Each challenge ships with an annotated reference solution — including the path the panel took, and the paths they rejected.
Leaderboards
Ranked by aggregate defensibility score and 30-day regression rate, not commit count. Reset quarterly.
Competitions
Weekly timed events. Same scoring rubric as the certification panels. Top finishers get pulled into the next cohort.