Linus Torvalds accepts AI-generated patch for Linux kernel cache contention
Linus Torvalds has incorporated an AI-generated patch into the Linux kernel to improve cache performance while updating the project policy to no longer require disclosure of AI tool usage.
Linus Torvalds released a development branch patch for the Linux kernel created using a large language model. The patch addresses a cache contention issue occurring when numerous CPUs read an identical file. This problem affects web servers, databases, and shared libraries.
While a synthetic benchmark shows a 23x performance improvement, Torvalds cautions that real workloads will not approach that figure, and gains will likely matter more for servers than desktops. Torvalds previously posted his own draft in early 2024 and ran it on his laptop before turning to an LLM.
The kernel project treats AI tools like compilers or editors. Torvalds supports LLM tooling within the review-heavy process, noting last July that an automated code review tool used by the team outperforms human reviewers by catching basic programming mistakes. Additionally, the project is changing its policy so patch authors no longer need to disclose AI tool usage.
Meanwhile, AI tooling increases work for maintainers. AI agents excel at spotting security flaw patterns, making the Linux security mailing list "unmanageable" earlier this year. The kernel fixes about 2,000 CVEs per release. This reporting volume is forcing the team to drop older hardware drivers that have small install bases yet attract many reports.