Linux development is really benefiting from AI, but not from its coding

Published Sep 16, 2026, 9:01 AM EDT Simon is a Computer Science BSc graduate who has been writing about technology since 2014, and using Windows machines since 3.1. After working for an indie game studio and acting as the family's go-to technician for all computer issues, he found his passion for writing and decided to use his skill set to write about all things tech. Since beginning his writing career, he has written for many different publications such as WorldStart, Listverse, and MakeTechEasier. However, after finding his home at MakeUseOf in February 2019, he would eventually move on to its sister site, XDA, to bring the latest and greatest in Windows, Linux, and DIY electronics. If you haven't been keeping up with the world of Linux distros, each one has been debating adding AI code. Some platforms have outright banned any output from an LLM — not just the code, but even comments and commit notes, too. For the distros that do accept AI code, they usually adopt the same model; people are allowed to use as little or as much LLM-generated code as they please, but they're on the hook for it. If they submit shoddy code, they can't point at their AI agent and say it's all its fault. But one thing I have noticed is that the true power of LLMs in Linux development may not be AI-generated code. It's not even with making documentation. It's sniffing out the bugs in the code, even if they don't actually fix them. AI's ability to fix errors in code hasn't been ideal... We're seeing people complain about its quality So, to clarify up front, I am by no means a Linux developer. However, I have been keeping an eye on how AI has influenced how people have contributed to both the Linux kernel and the distros it relies on. So far, to my knowledge, two people have discussed how AI has helped them code fixes for their software; or, more accurately, how it hasn't. The first is Linus Torvalds, who recently used AI to fix a bug in the Linux kernel: And this was a debug session from hell, enormously helped by an AI doing much of the grunt-work. [...] But while the AI was ready to give up several times, it did keep adding debug code and analyzing it faithfully when I pushed. So credit where credit is due and I let the AI write the commit message above. The second is Linux developer Lorenzo Stoakes of Arm, who recently used an LLM to find issues in the Linux kernel: An LLM was used to first determine where the bottlenecks were then to figure out how to improve them. It generated a lot of code, much of it hideous. I extensively audited and rewrote a lot of it, and heavily edited commit messages, the cover letter and comments. Both of them allude to the fact that the code the LLMs were making wasn't up to par for the Linux kernel, and either required a lot of to-and-fro to get the code looking just right, or required lots of manual tweaking to get it right. And if we leave it at that, we can say that LLMs are handy when it comes to coding, but by no means perfect. ...but its ability to find the bugs in the first place is tremendous AI is acting as a better sniffer hound than a hunter But let's hold our horses a little here. Yes, both Torvalds and Stoakes said that the code the AI generated wasn't up to snuff and needed work to get going. But they also noted that the AI came in handy with something that neither of them really focused on: finding the bugs in the first place. This is where I think LLMs truly shine when it comes to Linux kernel development. Instead of treating them as fellow developers submitting code to the kernel, I feel they're best used as some kind of digital bloodhound. They go out, they sniff out the bugs in the code, and then bring back their findings to a skilled developer who has in-depth knowledge of the kernel and its guidelines. The reason why I bring this up is that it adds an interesting element to distros that have banned AI submissions, such as Gentoo and Void Linux. Sure, submitting AI-generated content is disallowed, but where do critics stand when it comes to using the agent to find bugs? From what I've seen, people who dislike AI tools usually cite at least one of three reasons: plagiarism, copyright, and the environmental factor. The first two are totally subverted if someone uses AI to find the bugs but not fix them; the LLM doesn't leave its "fingerprints" on the code whatsoever. However, while using an LLM to analyze code uses less power than generating content, it's still making a small impact on the environment and encouraging AI companies to roll out more data centers. Linux maintainers are already feeling the brunt of these new bug-hunters They're getting swamped For further proof of AI working better as a bug finder than a fixer, just look at the Linux maintainers. For a few months now, Linus Torvalds has noted that each release candidate for the Linux kernel has seen far more bug reports and fixes than a year ago. This is due to AI agents finding the bugs and either fixing them themselves or shooting off a report for someone else to handle. The problem is, all those bug reports and commits end up on the laps of people maintaining the kernel. What was once a quieter job has since been turned into a tsunami of bug fixes and reports, pouring in from all angles, mostly from AI agents. Worst of all, there are some people who aren't fully aware of how Linux handles bug reports and are using AI agents to escalate minor bugs through the emergency security channels, which has clogged things up a bit and annoyed Torvalds. AI is very good at finding bugs...a little too good, in fact While people rightfully debate the positives and negatives of using AI-generated code in an open-source project, I believe that the real value is using agents as bug hunters. Whether or not the agent has permission to then create a patch itself is up to the people running it, but either way, it seems they do a fantastic job, if not a little too fantastic.