Open source maintainers are running out of patience. The team behind RPCS3, the long-running PlayStation 3 emulator, publicly asked contributors to stop dumping AI-generated pull requests on its GitHub page, according to Hacker News. Their post on X promised bans for anyone submitting AI code without disclosing it, and their replies to defenders of “vibe coding” were even sharper than the original warning.
This isn’t a one-off rant. Hacker News reports that Godot Engine project manager Rémi Verschelde flagged the same problem back in February, saying the Godot repo had become so swamped with AI-generated PRs he was considering hiring extra maintainers just to triage them. RPCS3 has been shipping since 2011 and now runs 70% of the PS3 library. Godot powers thousands of indie games. These aren’t fringe projects, and the maintainers aren’t anti-AI cranks. They’re drowning in code that doesn’t compile, doesn’t solve real bugs, and wastes hours of human review time.
Why this matters now
The economics of open source assume good faith on both sides. Maintainers review code for free. Contributors submit code they understand and can defend. Generative coding tools broke that contract. Anyone can now produce a 200-line PR in 30 seconds without reading a single function in the codebase. The cost of generating slop dropped to zero. The cost of reviewing it didn’t.
RPCS3 put it bluntly when asked how they distinguish AI code from human code: “You can’t possibly handwrite the type of shit AI slop we have been seeing.” Pattern recognition is doing the work. Wrong abstractions, hallucinated APIs, fake commit messages, confident explanations of code that doesn’t run.
The broader signal
This is the first visible crack in the “AI makes everyone a 10x developer” narrative. A few patterns worth tracking:
- Maintainer burnout is becoming a public story. Until now it was a private Slack complaint. Now it’s on X with thousands of replies.
- Disclosure rules are coming. Expect more projects to require contributors to flag AI-assisted PRs, the way academic journals now require disclosure of LLM use.
- Reputation systems will tighten. First-time contributors with AI-shaped PRs may get auto-flagged. Trusted contributors get faster review. The barrier to entry goes up.
- Tooling will adapt. GitHub itself will likely ship AI-detection signals for PR queues, similar to spam scoring on email.
Practical takeaways
If you contribute to open source:
- Don’t paste AI output you can’t explain line by line. Maintainers can tell, and the cost of getting flagged is your reputation across every project you ever want to contribute to.
- If you use AI to draft a fix, disclose it and walk through what you verified. That’s a different conversation than dumping a PR and hoping.
- Read the contributing guide. Several projects already ban undisclosed AI submissions.
If you run an open source project:
- Add an AI disclosure clause to your CONTRIBUTING.md now. RPCS3’s policy is a clean template.
- Set up a triage label for suspected AI PRs so maintainers can batch-review without burning a full code-review session on each one.
- Be public about enforcement. The deterrent effect compounds.
If you’re a CTO or eng manager watching this from the outside: the same dynamic is hitting your internal code review queues. The volume of low-quality PRs is going up. Reviewer time isn’t. Plan accordingly.
The RPCS3 incident is small on its own. The trend it represents is not. More details at the original Hacker News discussion.