kodwai

Kodwai addresses the gap in traditional assessments by focusing on how engineers collaborate with AI agents in practical scenarios. Users select from ticket-style problems across categories such as backend and data, then run a command-line tool to set up the environment on their own machine. The platform tracks the entire process including prompts, iterations, and fixes to evaluate judgment and recovery from errors. The workflow involves picking a challenge, initializing it via the CLI with a chosen agent, working through the task independently, and submitting the session for analysis. Scoring breaks down into direction, outcome, and lift, highlighting effective steering over simple test passage. This approach reveals abilities that matter in modern development where agents handle routine work. By examining full sessions rather than final outputs, Kodwai rewards thoughtful interaction and produces transparent feedback. Developers gain insight into their vibe coding habits while competing on a leaderboard that emphasizes real engineering practices.