Steer an AI agent with your judgment.

Watch an agent take on a real task, then write the rubric that scores it — the human data that measures and improves AI.