Linus Torvalds used Claude Code to rewrite how Linux reads cached files (startupfortune.com)

Linus Torvalds, the prominent figure behind the Linux kernel, has recently embraced AI-assisted coding by posting a patch utilizing Claude Code, an AI coding assistant, on the Linux kernel mailing list. This patch, targeting a specific inefficiency in the way Linux manages cached file reads, dramatically improves performance by reducing contention in multi-threading scenarios. The solution shifts concurrent read operations into an RCU read-side critical section, effectively minimizing overhead and increasing throughput from around 10 million to up to 33 million reads per second on benchmark tests Torvalds conducted on his own hardware. This development is significant for the AI and machine learning community as it marks a turning point in how open-source software projects, particularly Linux, perceive AI contributions. Historically skeptical of AI, Torvalds' willingness to acknowledge and leverage AI-generated code signals a broader acceptance within the tech industry, particularly as tools like Google's Sashiko AI reviewer demonstrate efficacy in catching kernel bugs. By taking ownership of the patch and validating its performance, Torvalds sends a strong message about the potential for AI tools in serious infrastructure projects, potentially encouraging more developers to integrate AI into their workflows and altering the landscape of software development practices.