Linus Torvalds’ Surprising Take on AI in Software Development
▼ Summary
– Linus Torvalds supports Rust’s integration into the Linux kernel despite some developer resistance, believing it encourages maintainers to be open to new ideas.
– Torvalds views AI positively for making Nvidia a significant Linux kernel contributor, contrasting with its past reputation as difficult to work with.
– AI presents mixed effects on Linux, aiding tasks like patch management while causing infrastructure strain from crawlers and generating unreliable bug reports.
– Torvalds sees AI-driven “vibe coding” as beneficial for inspiring new programmers by lowering barriers to entry, similar to his early computing experiences.
– He advises developers to pursue hobbies where failure is acceptable for stress relief, drawing from his own experience with guitar pedal projects.
During a recent keynote at the Open Source Summit Korea 2025, Linus Torvalds shared his nuanced perspective on artificial intelligence in software development, touching on its potential to inspire new programmers and its practical effects on large-scale projects like the Linux kernel. Speaking with Dirk Hohndel of Verizon, Torvalds balanced enthusiasm for AI’s educational role with caution about its disruptive influence on development workflows.
Torvalds began by reflecting on his long-term involvement with Linux, noting that for nearly two decades his focus has shifted from programming to technical leadership and maintenance. He pointed out that each new kernel release involves over a thousand contributors, illustrating the project’s ongoing vitality. “Even after thirty-five years,” he remarked, “we’re still refining core code, improving maintainability, and adapting to hardware changes, there’s always more to do.”
The discussion moved to Rust’s gradual adoption into the Linux kernel, a change Torvalds supports despite some internal resistance. He acknowledged that the transition has not been entirely smooth, with some maintainers stepping down, but he defended the move as necessary for progress. “Encouraging maintainers to stay open to new ideas is essential,” he said, “even when it means navigating some disruption.”
When the topic turned to AI, Torvalds highlighted one clear benefit: Nvidia’s transformation into a major contributor to the Linux kernel. He recalled a time when Nvidia was famously difficult to work with, but today the company actively supports Linux development. “They’ve become deeply invested in our community,” he observed, “and that’s a positive shift.”
Still, Torvalds described AI’s overall impact as a mixed bag. On one hand, experimental tools are helping maintainers manage patches and stable releases. On the other, AI-driven web crawlers have placed a heavy burden on kernel infrastructure, consuming resources and creating headaches. Hohndel echoed this, citing cURL developer Daniel Stenberg’s experience with AI-generated security reports that resemble denial-of-service attacks. Torvalds confirmed that while his team hasn’t faced the same severity, they do see fabricated bug reports from people misusing AI, which wastes valuable time.
Torvalds admitted he doesn’t personally experiment with AI coding tools, but he recognizes their appeal, especially for newcomers. He compared today’s “vibe coding”, using AI for quick, informal projects, to his own early experiences typing programs from computer magazines. “Modern computing can feel intimidating,” he said. “If AI helps someone get excited about coding who otherwise wouldn’t, that’s a good thing.” He cautioned, however, that relying on AI for serious, maintainable code remains problematic.
Looking ahead, Torvalds expressed hope that AI will eventually become an unremarkable part of the toolkit, rather than a constant topic of hype. He also addressed concerns that AI-driven productivity could lead to layoffs, comparing AI to compilers: both change how developers work, but neither replaces the need for skilled maintainers.
Toward the end of the conversation, Torvalds lightened the mood by sharing his hobby of building guitar pedals, despite having no particular talent for music or mechanics. He encouraged others in high-stress tech roles to find pastimes where failure is not just acceptable, but enjoyable. “Working on something simple and hands-on,” he said, “can be a great way to unwind from the complexity of kernel development.”
(Source: ZDNET)
