AI code in the Linux kernel: a record 1,634 changes in a week
XORit Editorial · Sources, AI assistance and corrections

An observer at The Lunduke Journal reported another record week for AI in Linux development: by their count, 1,634 changes written by neural networks landed in the kernel last week.
By the author's own tally, since the start of September 2026 AI-generated code has accounted for 17.25% of all Linux kernel patches. In other words, roughly every sixth patch this month was, by that estimate, produced with the help of generative models.
An important caveat: this is one observer's estimate, not official project statistics. The Linux kernel is developed by a large community, and the project publishes no reporting on the share of AI code, so the figure should be taken as a rough indicator rather than a measured fact.
If the estimate is close to reality, it shows how deeply generative tools have worked their way into systems software development — a field where reliability and security requirements are traditionally high and the cost of a mistake in code is especially steep.
Opinions in the comments under the post were divided. Some readers are convinced the real share is higher: developers don't always disclose that code was written with AI. Others describe a similar practice at their own jobs, where models are used to hunt for bugs in existing code, and wonder what will happen once the easy bugs run out.
For now these are only estimates and projections. There is still no official data on what share of the code in the Linux kernel is actually written by neural networks.
Translation editor: Салтанов А.
Source: x.com ↗