Linus Torvalds, creator of Linux, recently published a development branch called noref-filemap-read-with-llm in the kernel. In the patch message, he made the use of AI explicit: “The code, the test programs, the review behind the list above and the first draft of this text were done with an LLM coding assistant (Claude Code) under my direction. The measurements were run on my machine.”
The work addresses an old cache-contention bottleneck when many threads perform small reads from the same cached file (a common scenario with shared libraries, databases, and web servers). The approach explores page-cache reads without folio references, using mechanisms such as RCU. Synthetic tests showed significant throughput gains compared with the baseline.
Torvalds also used LLM tools to make patches more readable and, in the process, identified additional long-standing bugs in the assoc_array library code. On the kernel mailing list he criticized rules that require mandatory disclosure or bans on specific tools, calling them “utterly stupid.” His position is that developers should understand what they are doing; the choice of tool does not define competence.
Discussion about the Assisted-by tag
In parallel, kernel maintainers are discussing the possibility of removing or simplifying the Assisted-by tag introduced to mark patches with LLM contribution. Arguments include low adoption, limited practical value of the collected data, and the fact that final responsibility remains with the human who signs the Signed-off-by. A majority of participants at a recent Maintainers Summit supported the change. Torvalds and Greg Kroah-Hartman support treating AI as just another tool, without special prohibition rules.
Why it matters
The episode marks a point at which the most influential maintainer of the world’s most important free-software project openly uses a frontier coding assistant on kernel work and publicly defends normalizing that practice. This influences both Linux contribution policy and the broader debate about attribution, quality, and “slop” in open-source projects. The patches still need broader review and testing before any mainline merge.
Sources
- Phoronix — Linus Torvalds Working On A Hefty Linux Kernel Improvement With Claude Code
- Tom’s Hardware — Linus Torvalds posts his first AI-assisted patch to the Linux kernel
- Phoronix — Linus Torvalds used AI to make code decipherable and uncovered more bugs
Transparency: This content was created, edited, or reviewed with the assistance of artificial intelligence. Information was cross-checked with public posts on X and available internet sources. Consult original sources for the full context.
Por GeekikiBot