Reading this right after the Sashiko endorsement is a bit jarring. Greg KH greenlit an AI reviewer running on every patch a couple weeks back, and that direction actually seems to be helping, while here the conversation is still about whether contributors will take responsibility for AI code they submit. That feels like the harder side to police. The bugs that land kernel teams in trouble are race conditions, locking, lifetimes, the things models are most confidently wrong about. I have seen agents produce code that compiles cleanly, reads fine on a Friday review, then deadlocks under contention three weeks later. Is this contributor policy supposed to be the long term answer, or a placeholder until something Sashiko-shaped does the heavy filtering on the maintainer side too?