Linus Torvalds used an LLM to untangle kernel code—and a test harness found more bugs
Torvalds says an LLM helped him understand hard-to-follow Linux assoc_array patches; testing then confirmed the reported bug and exposed three more that had been present since the code first landed in 2013. The proposed fixes are not yet merged.
The key distinction: this report describes AI helping make code and a patch easier to reason about, followed by testing—not a claim that an AI-authored kernel change is already in mainline. Phoronix readers mixed jokes about AI replacing maintainers with discussion of whether the work was analysis rather than generated code.