Linus Torvalds has released an eight-patch Linux kernel series that he says was largely written with Anthropic's Claude Code under his direction, according to ZDNET. The patches target a known page-cache bottleneck that slows small reads when multiple threads repeatedly access the same cached file on multi-core systems.
In tests on his four-core, eight-thread Intel Core i7-8565U laptop, eight-thread throughput rose from roughly 9.9–10.4 million reads per second to 30–33 million, and kernel cycles per read fell from about 2,950 to 725. Single-thread reads improved less dramatically, from about 452–454 cycles to 408–409 cycles per read. ZDNET notes these are experimental results from a specific workload, not a general speedup for Linux.
The approach defers reuse of memory removed from the page cache until an RCU grace period elapses, letting readers copy data without repeatedly updating a shared reference counter. Torvalds called reclamation "the part of this series that is least solved," and the branch remains outside mainline Linux 7.3-rc6 as of Oct. 13. A regression also appears when a read targets a user page that is not yet mapped.
The release marks a notable step in Torvalds' evolving attitude toward AI. He said he now "really like[s] using AI" and finds it makes programming more enjoyable, while cautioning that it should be treated as a tool. ZDNET frames the patch series as the culmination of his shift from skepticism to hands-on use.