Join Nostr
2026-08-05 17:43:16 UTC

[object Object] on Nostr: today in the Linux kernel turning into slop: this is what LLM code reviews do in ...

today in the Linux kernel turning into slop: this is what LLM code reviews do in every project that uses them

you get spammed with exhausting, mostly irrelevant text from a machine that gaslights you if you disagree with it

you deal with human reviewers who think you’re incompetent if you ignore the LLM review, especially if it catches a bug by pure chance but it got lost in the output or stated in a way you couldn’t decipher

if you disagree with the LLM in matters of style or taste, both the machine and the humans will yell at you. all code bases must look the same. all code must be structured like garbage that sucks to read.

the most expedient and least painful way to deal with the LLM reviewer is to use an LLM or just go LGTM to the bot, and according to all research not paid for by the AI companies that will result in your codebase turning into unmaintainable broken slop.

and yet linus claims that this is just a tool, and the unavoidable future https://social.treehouse.systems/@pac85/117042964033491417