← Back to digest
KOL Takes

Linus Torvalds on LLMs in kernel development

What's this about?

Linus Torvalds, creator of Linux, weighed in on a kernel mailing-list question about using LLMs for kernel development. Notably, the take is calm and pragmatic — treating LLM-assisted patches as something to be judged by the same standards as any patch — rather than one of his famous outbursts.

What exactly is "the kernel" and why is this sensitive?

The Linux kernel is the core software under most of the world's servers, phones (Android), and cloud infrastructure — mistakes there are security holes for billions of devices. Kernel patches go through famously exacting human review. Whether AI-generated code belongs in that pipeline — and whether submitters even disclose it — is one of open source's touchiest questions, and Torvalds' voice carries more weight than anyone's.

What's the community saying?

HN 39 points — modest points, notable content
  1. "Cogent, relaxed and nice" — surprise at the measured tone; several note that even his outbursts were usually right.
  2. Meta-complaint about attention — "sane take gets zero traction on HN, meanwhile every doomer post rockets to the top."
  3. A dissent — one commenter finds it *too* relaxed: the issue isn't the tools but the scale of impact, which broad statements don't address.
← Pre-clearance for open-weights models?"The LLM Critics Are Right. I Use LLMs Anyway" →