THE CRUNCH
Linus Torvalds has no objection to AI-assisted Linux kernel development, provided the person wielding the tool knows what they are doing. In comments reported by The Register, he called arguments against AI in commit messages "entirely misguided", saying "we want competent people to use AI as a tool". The flip side: he wants the kernel community to discourage "drive-by patches" from people who send a contribution once and never intend to learn more.
The comments follow a decision announced by kernel developer Sasha Levin at the Maintainers Summit in Prague: the Assisted-by tag, which flagged whether AI was used in a submission, is being deprecated. That does not open the door to careless code. Submissions still have to pass review, and Torvalds stressed that clueless AI use should be discouraged even as newcomers who genuinely want to learn are welcomed.
There is evidence the balance is already shifting. Long-time maintainer Greg Kroah-Hartman said the flood of obviously wrong "AI slop" security reports the project received months ago has given way to real, useful AI-generated bug reports, though maintainers are now buried in good ones. He suggested AI bug finders may eventually exhaust the readily discoverable problems, easing that pressure.
AI is also being put to more concrete work. Red Hat's Christian F.K. Schaller built hw-harness, an open source tool that automates the slow business of working out how a piece of hardware behaves so a Linux driver can be written for it, making it easier to build drivers from Windows drivers. That could help with one of the desktop's oldest irritations: missing hardware support.