跪拜 Guibai
← All articles
Frontend · AI Programming · Claude

Godot Bans AI-Generated PRs Outright, and That's a Good Thing

By kyriewen ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Open-source maintainers are drowning in AI slop, and Godot's ban is the canary in the coal mine. The same dynamic is hitting internal teams: when engineers submit code they can't explain, the cost of review and debugging shifts downstream, eroding trust and velocity.

Summary

A flood of low-quality, AI-generated pull requests pushed the Godot Foundation to update its contribution policy with a permanent ban for anyone using an AI agent to submit code directly. The rule is the most severe penalty an open-source project has enacted against Vibe Coding, and it arrives alongside a new term circulating in developer circles: Vibe Slop—code that appears functional but that no one understands or takes responsibility for.

The core problem isn't AI assistance; it's AI abdication. Maintainers can no longer judge a submitter's intent when reviewing machine-generated patches, and submitters often cannot explain why the code was written a certain way. The result is a rising review burden that makes PR volume a liability rather than a strength.

For individual developers, the distinction that matters is whether AI output gets a line-by-line review before submission. The policy forces a hard reset from 'the AI wrote it, so it's not my problem' back to 'I stand behind every line I commit,' a boundary that applies just as much inside a company codebase as it does in open source.

Takeaways
Godot permanently bans any contributor who uses an AI agent to submit code directly to the repository.
Contributors with three or fewer merged PRs are classified as new and face additional restrictions.
Every PR must be reviewed and approved by a human before merging; the policy strengthens a rule that already existed.
The term 'Vibe Slop' now describes AI-generated code that runs but that no one understands or owns.
Maintainers report that reviewing AI-generated PRs is harder because the submitter's intent is opaque.
AI assistance is fine when the developer reviews every line; AI abdication happens when code is submitted without understanding.
A four-question self-check can reveal whether a developer is sliding from assistance into abdication.
Conclusions

The ban targets a specific failure mode—submitting unreviewed AI output—not AI tooling itself, which makes it harder to dismiss as Luddism.

Calling the penalty 'permanent' rather than temporary signals that the trust cost of an unreviewed AI submission is treated as irreparable.

Vibe Slop as a term captures a real shift: code quality is no longer just about correctness but about whether the author can explain and stand behind it.

The new-contributor threshold suggests that AI-generated PRs are disproportionately coming from drive-by contributors with no ongoing stake in the project.

Concepts & terms
Vibe Coding
A style of programming where a developer prompts an AI to generate code and accepts the output with little or no review, relying on the 'vibe' that it looks correct rather than on verified understanding.
Vibe Slop
A derogatory term for AI-generated code that appears to work but that no one—including the submitter—truly understands or is willing to take responsibility for.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗