Hardware Ready Ready to Bench?
The increase in patches created with artificial intelligence is increasing the workload for Linux kernel maintainers. Lorenzo Stoakes, an expert developer in memory management, has decided to give up some review responsibilities.
The phenomenon extends to other open source projects as well. Artificial intelligence is changing how the Linux kernel is developed, but the rise in code produced with the help of coding agents and large language models (LLMs) is creating new difficulties for those verifying quality, correctness, and integration.
Linux kernel: why it matters
Lorenzo Stoakes, an Arm developer and Linux kernel contributor since 2014, announced he would reduce his involvement in certain code review activities. Specifically, he will no longer serve as a reviewer for the Memory Management (MM), Core, Misc, and Rust areas.
The decision does not imply abandoning the project or relinquishing maintenance responsibilities for the subsystems he directly oversees. The developer will continue to fulfill that role but will limit general reviews for which he can no longer guarantee the necessary availability.
The choice is driven by the increase in the number of patches sent to the project, a trend Stoakes links directly to the spread of generative artificial intelligence tools. The greater ease with which source code modifications can be produced is increasing the volume of proposals submitted to the community, without this growth being matched by an increase in available resources to examine them.
In the message associated with the patches formalizing his decision, Stoakes explained: “Given the current volume of kernel contributions from AI-generated patch series, unfortunately I do not have the capacity to generally provide review activities, except for areas where I am a maintainer.”
What changes and what are the effects
The phenomenon is not limited to Linux. The use of coding agents is determining an increase in modification proposals in other open source projects as well. One study found that up to 25% of new pull requests examined could be attributed to what is defined as AI slop: code generated by artificial intelligence and submitted to projects without an adequate level of verification or refinement.
The developers of Godot, the open-source graphics engine, have also expressed concern over the influx of AI-generated contributions, going so far as to impose a ban. The problem is organizational before it is technical: every proposal requires time to be examined, verified, and compared with project characteristics. An apparently correct patch can still introduce regressions, security issues, or incompatibilities with existing components.
The matter takes on particular interest in light of Linus Torvalds’ recent change of position regarding the use of artificial intelligence in kernel development. After an initially cautious approach, the creator of Linux has shown greater openness toward AI-generated code, going so far as to use it in his own contributions and encouraging other developers and maintainers to experiment with these technologies. The latest version of the kernel recorded a record number of lines of code added precisely on the wave of generative tool usage.
Source and further reading on Linux kernel: original article.
* Content created with the assistance of artificial intelligence systems.