Linus Torvalds is using AI to develop Linux kernel code now
Original reporting by ZDNet
The latest set of Linux kernel patches, released by Linus Torvalds, are notable for being largely co-authored with Anthropic’s Claude Code AI. This marks a pivotal moment, as the patches tackle a persistent, unsolved memory cache slowdown issue prevalent when multiple cores and processors contend for shared data. Specifically, the collaboration targeted inefficiencies in Linux’s page cache during repeated small reads, a problem that could severely impact system performance. The proposed solution involves deferring memory reuse, thereby circumventing the bottleneck created by atomic writes to a shared cache line. Initial experimental benchmarks on Torvalds' laptop show substantial performance improvements, with one specific eight-thread workload demonstrating a dramatic three-fold increase in read throughput.
A Fundamental Shift
This practical application of AI reflects a profound evolution in Linus Torvalds’ long-held perspective on the technology. Moving from initial skepticism to openly embracing AI as a valuable programming tool, this development represents his most significant endorsement yet, integrating AI directly into the heart of Linux kernel development. While the patches remain experimental, carrying inherent engineering trade-offs like delayed memory reclamation and requiring further refinement, their very existence is a testament to AI's growing utility. Torvalds, the architect of the world's most critical open-source operating system, has now leveraged AI to push the boundaries of system optimization, establishing a new paradigm for how large language models can contribute to fundamental software engineering.
Torvalds' decision to leverage Anthropic’s Claude Code for a critical kernel patch is more than just an interesting anecdote; it represents a definitive turning point in the adoption of AI within foundational software development. While the experimental patches for the page cache bottleneck are still under review, highlighting both significant performance gains and unresolved technical trade-offs, their mere existence signals a profound shift. The creator of Linux, long a careful observer, has now moved from skepticism to hands-on embrace, applying AI to one of the world's most vital codebases. This action legitimizes AI as a potent, practical tool capable of generating production-level code, even if human oversight remains paramount.
A New Era for Development This milestone extends far beyond the Linux kernel. Torvalds' endorsement acts as a powerful validation for AI's role in intricate, high-stakes coding environments, likely accelerating its widespread adoption across critical infrastructure and enterprise software. It reframes the conversation around AI from merely assisting with bug detection or patch review to actively participating in core code generation. We can anticipate a future where developers, even those building the bedrock of our digital world, increasingly integrate AI assistants into their daily workflows, shifting their focus towards strategic design, rigorous validation, and the nuanced refinement of AI-generated solutions. This integration will inevitably foster new discussions on code ownership, accountability, and the evolving skill sets required for human developers, pushing the boundaries of what collaborative software engineering entails in an AI-augmented landscape.
Frequently asked questions
- Has Linus Torvalds used AI to write code for the Linux kernel development?
- Linus Torvalds recently used Anthropic's Claude Code, an AI coding assistant, to generate patches for the Linux kernel. This marks a significant shift in his approach to development. The AI helped craft the code, test programs, and even initial drafts of the accompanying documentation for a specific memory cache issue. This demonstrates a growing acceptance of AI as a valuable tool in open-source projects.
- What specific technical problem did Linus Torvalds address with AI-generated Linux kernel patches?
- The AI-assisted patches target a memory cache slowdown problem in the Linux page cache, particularly when multiple processor cores access shared file data. This contention, due to repeated updates of reference counts on cached data, can significantly slow down specific workloads. The proposed fix defers memory reuse to reduce competition for shared cache lines, improving read performance in certain scenarios.
- How has Linus Torvalds' perspective on using AI in software development changed over time?
- Linus Torvalds initially expressed skepticism about AI for coding but has progressively embraced it as a powerful tool. His views evolved from seeing AI as useful for bug identification and code maintenance to actively employing it for generating new kernel code. He now considers AI a valuable assistant that enhances the programming experience, though he cautions careful use for critical projects like Linux.