THE CRUNCH
Linus Torvalds has posted the first release candidate of Linux kernel 7.4 and declared that the project has found a "much healthier model" for handling AI contributions. The merge window is unusually large: almost 20,000 commits against the usual figure of around 15,000.
Torvalds said in his weekly state of the kernel update that he is "much happier" seeing a flood of AI-triggered work arrive during normal development than showing up mainly as extra fixes during the release candidate period, when the kernel is meant to be stabilising. He hedged, though: people expect 7.4 to be the next long-term support release, which independently makes some releases larger. "So it may not *just* be the flood of AI patches," he wrote.
He also remained wary of the submissions themselves, describing 7.4 as "large, and full of fixes for issues found by AI", with more and more patches suggested by AI, though these often get rewritten by humans. AMD GPU register offsets and bit masks account for a third of the changes. His advice to anyone overwhelmed: "Just ignore the noise."
It is a notable shift in tone: in mid-October 2026 Torvalds posted his own first LLM-assisted patch, an attempt to fix an old cache contention problem that he claimed delivered a 23x improvement in a synthetic benchmark, while stressing real-world impact would be far smaller. Phoronix rates the headline inclusions in 7.4 as improved support for processors from AMD, Apple and Google, plus security tweaks for virtual machines, support for the Raspberry Pi 10-inch monitor and further work on Qualcomm's Snapdragon X2 SoC for PCs.


