A recent rise in "AI slop" code submissions is "draining and demoralizing" the team behind open-source game engine Godot

Godot developer Rémi Verschelde says a rise in AI-generated code is weighing down on the resources of its dedicated maintenance team. The open-source game engine has been around since 2014, but has gained increasing relevancy in recent years as an alternative to the likes of Unity, with Mega Crit switching to Godot for the development of Slay the Spire 2 as just one example. That growing popularity has been reflected in increasing submissions from budding contributors, but Verschelde cautions that an abundance of "AI slop" code is becoming overwhelming.

Godot has now been used in many notable projects. While indie games ranging from Slay the Spire 2 to Buckshot Roulette make up a lot of this count, it can even be found in the likes of Battlefield 6's Portal map editor. Verschelde, who has worked on Godot for more than a decade and is the co-founder of one of its key backers, W4 Games, says that the engine's team "prides itself in being welcoming to new contributors, letting any engine user have the possibility to make an impact on their engine of choice."

That team is now facing an increasing rise in pull requests that contain AI-generated code. A pull request, or 'PR', tells the submission review team that an in-development branch is ready for consideration to be integrated into the core code base it was originally adapted from. "Honestly, AI slop PRs are becoming increasingly draining and demoralizing for Godot maintainers", Verschelde writes. "We find ourselves having to second-guess every PR from new contributors, multiple times per day."

Verschelde's latest comments come in response to Hidden Folks developer Adriaan de Jongh, who writes that the increasing number of pull requests generated by large language models are "a massive time waster for reviewers - especially if people don't disclose it". Verschelde expands on this, detailing a number of key questions the maintenance team is forced to consider with every such submission.

"The description is verbose LLM output, is the code at least partially written by a human? Does the 'author' understand the code they're sending? Did they test it? Are the test results made up? Is this code wrong because it was written by AI, or is it an honest mistake from an inexperienced human contributor? What do you do when you ask a PR author if they used AI because you're suspicious, and they all reply, 'Yes, I used it to write the PR description because I'm bad with English'?"

Verschelde notes that an AI-assisted declaration is part of Godot's contribution guidelines, but that it is frequently ignored. He explains that Godot's maintenance team spends "a lot of time assisting new contributors to help them get PRs in a mergeable state," but stresses, "I don't know how long we can keep it up." For the time being, the best suggestion that he's able to propose is asking for "more funding so we can pay more maintainers to deal with the slop (on top of everything we do already)." If you'd like to help in that regard, you can do so here.

Among the other alternatives are moving off of Github, or employing trust systems that rate users based on community votes. "I dislike raising the barrier to entry, but we might have no other choice," Verschelde remarks. "I don't expect Github to shield projects against AI slop its parent company actively facilitates. But there's also a lot of pros for Godot being on Github, which aren't easy to replicate elsewhere."

While moving to a less-popular platform could reduce the number of these submissions, it "increases the barrier to entry not just to AI slop, but also to genuine contributors or even Godot users who just want to report issues or participate in discussions." For now, Verschelde says he is "monitoring how the situation evolves and we'll see."