RPCS3 Maintainers Ask Contributors to Stop Submitting AI-Generated Code

The RPCS3 maintainers have asked contributors to stop submitting AI-generated pull requests to the project. The request is direct: low-quality, LLM-produced diffs are consuming reviewer time without advancing the codebase.

This is not an isolated complaint. Maintainers across high-complexity open-source projects are running into the same pattern. AI coding tools lower the barrier to submitting code, but they do not lower the bar for what a project like RPCS3 can accept. Emulator development operates at a layer where correctness is non-negotiable and subtle hardware-behavior bugs can go undetected for a long time. A diff that looks plausible to a language model is not the same as a diff that is correct.

The practical problem is throughput. Every PR that arrives requires a human to evaluate it. When a significant portion of incoming contributions are generated by tools that have no understanding of the PS3's Cell architecture, RSX renderer, or the existing codebase conventions, the review queue fills with noise. Maintainers either spend time rejecting submissions or risk merging regressions.

For engineers using AI tooling on their own projects, this is a useful calibration point. Code generation is fast; code review is not. Pushing generated output upstream without validating it against the target system's actual behavior shifts the cost of your tooling onto someone else.

For open-source project maintainers, RPCS3's response is a reasonable template. Setting explicit contribution policies that address AI-generated code is now a practical necessity, not a philosophical stance. Projects that rely on deep domain expertise are especially exposed.

The broader implication: as AI code generation becomes default behavior for many developers, the social contract around open-source contribution needs explicit renegotiation. RPCS3 is doing that work now.

Source