Post #1090582
2026-04-01 21:39 UTC
An @lwn article on Sashiko captures a real tension - one side wants mandatory LLM patch review for linux-mm, the other says it's too noisy.
I pulled 466 findings from the public API to look at what's actually in those reviews. Duplication across reviews? Zero. Pre-existing bugs landing in your inbox? 1-in-5 chance, averaging 28% of the review text when it happens.
When a tool tells you about a bug you didn't write, who owns it now?
https://www.bexelbie.com/2026/04/01/whats-in-a-sashiko-review
#Linux #LinuxKernel
Replies (1)
-
@cgwalters@fosstodon.org 2026-04-03 13:40
@bexelbie "The sharpest question the data raises isn’t statistical. It’s social." I'm so sure that's not your voice, it's an LLMs. https://rfd.shared.oxide.computer/rfd/0576#_llms_as_writers And yes, when I read it it gave me exactly that feeling about wondering how much of the rest of the article you actually looked at at all.