Linus Torvalds uses Claude Code for Linux kernel improvements

Linux creator Linus Torvalds has confirmed he is using an AI coding assistant to help develop new features for the Linux kernel. In a recent patch message, Torvalds disclosed that he used Claude Code to create code, test programs, and draft documentation for a new kernel branch. This disclosure comes amid ongoing debates within the open-source community regarding the use of AI-generated content.

Kernel Development with Claude Code

Torvalds recently created a new branch in his main repository named noref-filemap-read-with-llm. The branch aims to improve memory management by allowing the system to read from the page cache without folio references. This change addresses a bottleneck where small reads of a single file from many threads contend on folio reference counting.

In the patch message, Torvalds explicitly stated his use of AI for this work: "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." According to Phoronix, the impact of these patches on addressing the read bottleneck is significant.

Torvalds' Stance on AI Tools

Speaking at the Open Source Summit Europe, Torvalds discussed his evolving view of AI. He told Dirk Hohndel, head of Ericsson Software Technology, "I used to think it was just not very good — but now I’m at the point where I actually really like using AI." However, he emphasized the need for caution in critical projects: "I think that you need to be very careful using AI when you’re doing something real and important."

Torvalds described AI as a "wonderful tool if you use it correctly and treat it as a tool," noting that it makes programming more enjoyable. He highlighted its utility for beginners and for tasks outside his primary expertise. For example, while working on a guitar pedal project, he used AI to generate a Java user interface because he does not typically write Java code. He noted the AI-generated result was "horrendous" but served as a functional comparison to his own C implementation.

Impact on Kernel Maintenance

The integration of AI tools into the Linux kernel workflow has introduced new challenges for maintainers. Torvalds acknowledged that while AI can identify genuine security problems and issues in old drivers, it also creates stress for the human maintainers responsible for reviewing code. He stated, "It is, I think, improving our code base… Although it is also stressing maintainers out to the point that it’s a problem."

Some subsystem maintainers now expect patches to have received an AI review, such as from the Sashiko system, before acceptance. Torvalds noted that "three-quarters of the preceding day’s Linux Kernel Maintainer Summit discussion concerned making AI generation and review less stressful and more useful." He pointed out that the current bottleneck is not generating patches, but processing them through review without exhausting the people responsible.

Why it matters

Torvalds' direct use of Claude Code for kernel development signals a shift in how major open-source projects integrate AI. While he maintains that AI is best used for specific tasks and hobby projects, its role in generating code for the Linux kernel highlights the tension between efficiency and the human review capacity required to maintain stability. The community continues to debate rules for AI-generated content, with some projects banning it and others, like the kernel, adapting to its presence.

---

Sources

• phoronix.com: Linus Torvalds Working On A Hefty Linux Kernel Improvement With Claude Code - Phoronix
• zdnet.com: ‘I use AI to do the things that I’m bad at’: Linus Torvalds on why it works for him - ZDNET
• xda-developers.com: As the Linux community debates LLM-generated content rules, Linus Torvalds says he "really likes using AI"

This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.