For years, open source operated on a relatively simple cultural premise:
if you send useful code, you’re welcome.
It didn’t matter much:
- where you learned
- whether you were junior or senior
- whether you worked at Big Tech or from your bedroom
- whether your English was perfect or not
The implicit contract was:
- understand the problem
- propose a reasonable solution
- take responsibility for what you send
Generative AI just broke that balance.
And maintainers’ reactions are starting to harden.
The RPCS3 Case Is Not Isolated
Last week, RPCS3 — one of the most technically sophisticated open source projects in the gaming ecosystem — published new rules against AI-generated PRs.
They didn’t ban AI.
That’s what’s important.
What they banned was:
- contributions without real understanding
- auto-generated PRs
- spam submissions
- changes without human ownership
The most quoted phrase from the announcement was brutal:
And honestly, that’s probably going to end up being the most important organizing principle of the entire coding agents era.
Because the easier it becomes to generate code, the more valuable the ability to:
- understand systems
- reason about tradeoffs
- critically review
- keep software alive
AI can generate patches.
It still can’t be held accountable for them.