Coffee With Kusari, episode 10: LLM reviews in the Linux kernel
Appeared on Coffee With Kusari on 21 April 2026
This conversation with Ben Cotton of Kusari is drawn from my posts What’s Actually in a Sashiko Review? and A Few More Thoughts on Sashiko and the Kernel.
We talked about the two hypotheses behind the analysis, why repeated findings do not seem to be the real problem, how pre-existing bugs show up unevenly in review output, and why the harder issue is the social burden of making humans respond to machine-generated review comments.
The conversation also touches on the broader question of what open source projects and AI tool creators should think about when considering this kind of tooling.