The committee managing the GNOME Circle catalog has approved new rules that prohibit the publication of applications generated using AI tools. GNOME Circle provides a platform for hosting applications and libraries created by third-party developers using GNOME technologies, facilitating their entry into the GNOME ecosystem.
Projects showing signs of AI usage for code generation, such as nonsensical code inserts, style inconsistencies, forced API usage, and comments with hints for AI, will no longer be accepted in GNOME Circle. The new rules also state that developers must have a good understanding of the submitted code, be able to justify the methods used, and answer any related questions about the code. However, using AI for learning and solving related development tasks, such as code autocompletion, is not prohibited.
The main reason for prohibiting code generation through AI is the attempt to expedite the mandatory review process for projects submitted for inclusion in the GNOME Circle collection. The queue for review has grown significantly, and some applications have to wait years for a decision regarding their inclusion in the catalog.
The recent observed increase in low-quality software publications generated by AI creates additional pressure on reviewers and further congests the application queue. To clear the backlog, the acceptance of any new applications for review is temporarily suspended. After the suspension is lifted, new projects created via AI will be rejected without review (applications submitted before the introduction of the new rules will be considered).
The second reason for banning AI usage is the tightening of rules in the Flathub catalog, which is used for hosting software packages presented in GNOME Circle. Indirectly, Flathub's rules also apply to the GNOME Circle project, making it pointless to spend time reviewing programs whose packages will not be accepted into Flathub.
A survey conducted among those accompanying GNOME Circle showed that
62% of participants do not use AI, 34% turn to AI assistants for minor questions or to prepare small code snippets, and only 3% generate large portions of code through AI. None of the respondents admitted to completely abandoning hand-coding in favor of AI generation.
Source: opennet.ru
