Linus Torvalds's first LLM-assisted patch series

Commenters reacted to Linus Torvalds using an LLM (Claude) to produce a kernel patch series and noted that he acknowledged the assistant and ran the measurements himself. Several people, including niemandhier and manbash, said the diffs looked moderate and reviewable, and pointed out that human review uncovered a race condition. Others, like tombert and DiabloD3, expressed mild surprise that Torvalds - long seen as pragmatic and hands-on in reviews - used AI at all, while manbash and dorianniemiec raised questions about the broader significance of AI assistance in core infrastructure.

Opinion divides over tool choice, quality, and policy. manbash, ACS_Solver and tonis2 argued pragmatism: pick the best tool even if proprietary, while eikenberry and mcbuilder criticized using closed models instead of open weights. ASalazarMX, steveklabnik and fasterik urged nuance: AI as a useful tool when combined with careful human review, whereas more alarmist voices (jmclnx, customguy, ciupicri) floated moves to “AI-free” alternatives. Legal and licensing points sparked contention: DiabloD3 claimed machine-written code can't be copyrighted, while throwaway7356 and bbor disputed that. Many commenters emphasized prompting skill and reviewer expertise as decisive for safe AI use.