Despite the immense popularity of LLMs and AI coding agents in software development, the open-source community seems to be quite divided on the topic at times. On the one hand, you have projects like GIMP banning AI from its codebase entirely, while other projects, like Nobara Linux and Omarchy, are very much all-in on AI coding. There is, however, a middle ground, like the Linux Kernel project, which allows AI-assisted and -generated code in the codebase, but all code requires human review and a proper merge request write-up that demonstrates an understanding of the actual code submitted.

This appears to be where KDE Plasma will land as a project, if the recent proposed KDE LLM guidelines are anything to go by. The draft of KDE's LLM guidelines was submitted for discussion by project lead, Nate Graham, and it outlines rules that effectively summarized by the golden rule: "Don't be lazy."

In the KDE LLM guidelines discussion post, the proposed draft explains that "contributions with obvious and lazy LLM usage may be ignored or closed," and that "nobody should know if you use an LLM," because the output should be indistinguishable from what you could produce yourself without an LLM. It also places rules on AI-generated text, forbidding AI-generated text for merge requests, commit messages, or replying to a question or comment.

The current draft also seeks to prevent additional work being placed on maintainers by limiting vibe-coded changes to that maintainers are expected to clean up and implement—an issue faced by Godot maintainers not long ago. The draft has yet to be finalized, and the discussion around the LLM guidelines is expected to continue in the upcoming days, after which a finalized version will be added to the KDE documentation.