Omarchy

The AI Patch Flood Is Already Costing Kernel Reviewers

Leave a comment

Lorenzo Stoakes has been in the Linux kernel since 2014, and he is the author of The Linux Memory Manager. This week he stepped back from some reviewer roles, Phoronix reported, because of the AI-generated patch flood. He is dropping reviewer duties for memory management, core, misc, and Rust. He is keeping the maintainer work in the areas he oversees. In the series that removes him from those reviewer slots, he wrote that, given the current throughput from AI-generated series, he “unfortunately lack[s] the bandwidth to generally be able to offer review except in those areas I maintain.” He wants a more manageable inbox. That is not a rant about tools. It is a scarce-attention problem.

FreeBSD is trying to get ahead of the same issue before it arrives as a pile in the review queue. The project’s Q3 2026 status report said 88% of developers who answered want an AI policy. 35% want AI-assisted code disclosed. 22% would treat it like any other contribution. Only 16% want AI-generated code rejected outright. 65% are fine with AI-generated documentation if a human reviews it. A working group is drafting the policy, with final reports expected in Q4. I already liked that they started by asking the people who maintain the tree. It is a different instinct than hoping the volume will just sort itself out.

My take

If AI help buries the humans who keep a project honest, it is not help. The kernel runs on review, and review runs on people who still have time to read. FreeBSD writing a policy now is the grown-up version of that lesson. Disclose it, review it, and don’t pretend a fluent patch series is the same thing as a trusted one.

God Bless, and Use the Source!

References & Further Reading

Subscribe (RSS) Follow News via RSS

Leave a note

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.