System76, the company behind the Ubuntu-based Pop!_OS and the upcoming COSMIC desktop environment, has updated its Pull Request (PR) template for the COSMIC project. The new template includes a mandatory checklist that explicitly prohibits contributors from submitting code that has been assisted by AI.
COSMIC’s New Stance on AI Code
Effective immediately, all contributors must confirm that their submissions contain no AI-generated code, comments, or descriptions. Failure to comply with this new requirement will prevent contributions from being accepted.
COSMIC is System76’s in-house Linux desktop environment, developed from the ground up using the Rust programming language. Prior to the development of COSMIC, System76 relied on a heavily modified version of the GNOME desktop environment for its Pop!_OS distribution. However, persistent disagreements between System76 and the GNOME upstream developers regarding design philosophies and feature priorities led the company to pursue its own independent desktop environment.
Concerns Over AI-Generated Code Complexity
This move by System76 is not entirely unexpected. The company has previously voiced concerns about the potential complexity of AI-generated code. System76 has argued that AI models often lack the comprehensive context required to understand the intricate interdependencies within large software projects. This can result in code that appears functional but is, in reality, difficult to maintain, significantly increasing the time and effort needed for code reviews.
The only codebase within the System76 ecosystem currently exempt from this ‘No LLM contributions’ rule is cosmic-flatpak. This project is dedicated to managing software components that are deeply integrated with the COSMIC desktop environment and are not easily adaptable for inclusion in the standard Flathub repository. This includes items such as COSMIC panel applets and desktop extensions.
A Growing Trend in Open Source
System76 is not alone in its cautious approach to AI-generated code in open-source projects. The Ladybird browser project, for instance, implemented similar restrictions back in June. The maintainers of Ladybird reported that a significant number of inexperienced developers were submitting AI-generated code. This influx placed an unsustainable burden on the project’s maintenance team, as reviewing and refactoring this complex, AI-generated code consumed far more resources than the project could afford.
For both System76 and Ladybird, the core issue isn’t that AI-generated code is inherently unusable. Instead, the challenge lies in the long-term maintenance costs associated with integrating such code into large, collaborative open-source projects. Human maintainers are ultimately responsible for understanding, reviewing, testing, and supporting this code. By implementing these restrictions, projects aim to mitigate future maintenance overhead and ensure the long-term health and stability of their codebases.
