THE CRUNCH
Lorenzo Stoakes, a lead reviewer for memory management in the Linux kernel, has announced he is stepping down from several of his review duties, citing the flood of low-effort AI-generated patches, often called AI slop. Writing on the Linux kernel mailing list, he said he is leaving the core, reclaim, Rust, THP (huge pages) and miscellaneous sections of the memory management area, while keeping remaining duties such as KVM. The departure appears amicable: he plans to finish his current patch series and will still pitch in occasionally, but says mounting pressure on his inbox forced the step back.
The underlying cause is hard to miss. "Of course the AI wave has been horrible, and is getting significantly worse," Stoakes wrote, describing how slop went from a rare sight to at least weekly, then every few days, and now "pretty much half of what I see, if not more." He had flagged the problem before: in July he warned that "the amount of time wasted on AI slop is already too much and we're only at the start of this," and a month ago noted that "what kills the project is having 1,000 slop submissions that 2 or 3 people have to do something with."
The open nature of kernel development means anyone can contribute, which is a strength when submissions are good and a burden when they are not. Under the current stance discussed on the mailing list, AI-assisted commits do not need to be marked as such, so reviewers are left filtering. That tension sits against a pro-AI drift from Linus Torvalds himself: in October 2026 he posted his first LLM-assisted kernel patch, a fix for cache contention when many CPUs read the same file, claiming a 23x improvement in a synthetic benchmark while cautioning real-world gains would be far smaller.
Stoakes is not alone in feeling the squeeze. One recent study found as much as 25% of new pull requests were reported to be AI slop, and maintainers of other projects such as Godot have complained about similar floods. Open-source projects are arguably hit hardest, since they rarely have the manpower or funds to sift the output of countless automated contributors.