Everyone who has used AI coding tools interactively or as agents knows they're unpredictably hit or miss. The old, non-agent Copilot has a dashboard that shows org-wide rejection rates for for paying customers. I'm curious to learn what the equivalent rejection-rate for the agent is for the people who make the thing.