Linux creator Linus Torvalds has spoken about how he uses artificial intelligence tools in programming, including work connected to the Linux kernel.
He discussed the topic at Open Source Summit Europe in Prague, describing AI as a helpful tool while cautioning developers about its risks in critical software.
Torvalds said AI has become useful for tasks outside his strongest areas and has also played a role in Linux kernel reviews.
He said maintainers must weigh potential improvements against the extra workload created by AI-generated submissions.
"I actually really like using AI," Torvalds said, according to the source material.
He told Dirk Hohndel of Ericsson Software Technology that AI can make programming more enjoyable when treated as a tool.
Torvalds described his role in Linux development as a maintainer rather than a primary kernel programmer.
He said AI is most valuable when developers understand its limits and remain responsible for the final work.
"I think that you need to be very careful using AI when you're doing something real and important," he said.
AI as a path for new programmers
Torvalds also pointed to AI's role for new programmers who want to experiment with software.
He said modern programming environments can feel intimidating because professional software has become much more complex.
"A lot of people make fun of vibe coding, but I think it's a wonderful way to find joy in programming," Torvalds said.
He said it makes people feel relevant and allows new programmers to do things that would otherwise be very hard.
Torvalds recalled beginning programming in 1981, when computers and software were simpler, and contrasted that with today's larger engineering systems.
He said AI can help bridge gaps for people building projects outside their expertise.
"So, I use AI to do the things that I'm bad at," he said.
He described using AI on a guitar pedal project to improve a user interface he had created in C.
In Linux kernel development, AI tools are being used for code review and bug detection.
According to ZDNet, some subsystem maintainers now expect certain patches to receive AI-assisted reviews before acceptance, though requirements vary across the project.
Torvalds said AI reviews have identified genuine security issues and problems in older drivers, but they have also increased pressure on maintainers.
He said the challenge is processing useful work without overwhelming reviewers.
"It is, I think, improving our code base… Although it is also stressing maintainers out to the point that it's a problem," he said.
Torvalds also discussed Linux's development history, including the shift from manually distributed patches to more advanced collaboration systems as the project grew.
He said early Linux development was centered on a small personal workflow before the community expanded.
"It was really just me, myself, and my computer," he said.
Recent reporting from Startup Fortune said Torvalds also disclosed an experimental Linux kernel patch series created with an AI coding assistant under his direction.
The work targeted page-cache performance and included measurements run on his own machine, but the patch had not been merged at publication.
