AI wrote patches for Linux kernel: 23x speed boost
Linux creator Linus Torvalds published a series of patches written by a language model. They address the locking issue when multiple cores read files from the page cache simultaneously.
Synthetic test results on a Ryzen Threadripper showed a speed increase from 13.3 million to 304 million reads per second. This represents an approximately 23-fold improvement over previous solutions.
Testing limitations and AI usage
Linus Torvalds noted that this figure does not reflect real-world performance under actual workloads. However, the fact of improved metrics remains a significant achievement.
- The specific AI model used for these patches is unnamed
- Previously, the developer had used Claude for code generation
- Commit descriptions have become more detailed thanks to neural networks
Concurrently, the kernel development community has abandoned the requirement to disclose AI tool usage in commits. This change simplifies the process of contributing code changes.









