Godot maintainers struggle with 'draining and demoralizing' AI slop submissions

Rémi Verschelde, a maintainer of the open source Godot game engine, is the latest to complain about the impact of "AI slop PRs [pull requests]", which he said "are becoming increasingly draining and demoralizing for Godot maintainers."

His post was prompted by a comment from Adriaan de Jongh, game designer and director of small gaming company Hidden Folks, who said that LLM-generated PRs for Godot are a "massive time waster for reviewers ... changes often make no sense, descriptions are extremely verbose, users don’t understand their own changes ... it’s a total shitshow."

A comment noted that the Blender 3D design project is facing the same issue and has recently proposed an AI contributions policy, following others including the Linux Foundation, Fedora, Firefox, Ghostty, Servo and LLVM.

Verschelde appealed for "more funding so we can pay more maintainers to deal with the slop" and also spoke of the conflict between being welcoming to new contributors to let "any engine user have the possibility to make an impact" while also dealing with the onslaught of useless PRs. "I don’t know how long we can keep it up," he said.

GitHub itself is to blame, according to some comments, since the company is a big AI advocate. "This platform incentivizes this kind of behavior," said one; and another that it "is just exhausting to watch all this play out and GitHub promoting this, not fighting it."

Linux distro Gentoo is in process of migrating from GitHub to Codeberg thanks to "continuous efforts to force Copilot usage for our repositories."

One project, the self-hosting toolkit Coolify, has created an Anti Slop GitHub Action which its developer claims "could have closed 98 percent of slop PRs." The developer is not opposed to AI itself, and stated that "AI is one of the best things to ever be released and when used with experience and properly according to project guidelines it will pass all checks."

GitHub director of open source programs Ashley Wolf acknowledged the problem of "what happens when low-quality contributions arrive at scale" last week, though choosing her words carefully so as not to blame AI itself. According to Wolf, "maintainers have always dealt with noisy inbound." Nevertheless, GitHub is introducing features to make AI slop easier to deal with, including PR deletion from the GitHub UI (user interface) which she said is coming soon.

Wolf also mentioned relevant features that have already shipped, including the ability to limit PRs to collaborators or disable them entirely. Maintainers can also enforce temporary interaction limits for specific users.

Further refinements are under consideration. Wolf mentioned criteria-based gating, such as requiring a PR to be linked to an existing issue, or defining other rules that contributions must meet. There is also the inevitable suggestion that AI can be used to fix the problem it created, via automated triage.

Wolf’s post follows the creation of an official GitHub discussion on the subject earlier this month, as we reported. A GitHub product manager contacted us to state that "we don’t think counting AI-generated PRs is the right metric," showing again the tension between the company’s strong promotion of AI and the evidence of the damage it is doing to open source.