The economics of open source were already barely working. Now they're approaching a triage crisis.

Maintainers of Godot, the popular open-source game engine used by thousands of indie developers worldwide, are publicly warning that a flood of low-quality AI-generated code submissions is overwhelming their volunteer review capacity. "I don't know how long we can keep it up," one maintainer told PC Gamer.

Let me explain why this matters beyond one game engine.

The open source software that runs the internet - the libraries, frameworks, databases, and tools that every tech company depends on - is maintained by a relatively small number of volunteers who do this work for free, in their spare time, out of genuine commitment to the ecosystem. The economics are strained under normal conditions. Burnout among major open source maintainers is a documented crisis. The Log4Shell vulnerability of 2021 was a wake-up call: critical infrastructure depended on code maintained by a handful of people, underfunded and overworked.

Now add AI.

The barrier to submitting a pull request has dropped to near-zero. Anyone with a GitHub account and access to an AI coding tool can generate plausible-looking code and submit it to any open source project with no real effort. The code might not work. It might introduce subtle bugs. It might technically run but violate the project's architectural principles. But it looks like code, it has comments, it might even pass basic linting - and now a volunteer maintainer has to read it carefully enough to catch the problems.

For experienced maintainers, reviewing AI-generated submissions is more work than reviewing human-written submissions, not less. Human contributors usually have a specific problem they're trying to solve. Their code reflects actual understanding - even if imperfect - of the system they're modifying. AI-generated submissions are often technically plausible in isolation but architecturally naive: the code doesn't understand what it's interacting with, because the person submitting it doesn't either.

The scale problem is the critical one. If a project receives 100 genuine pull requests per month, that's a manageable review load. If AI tools enable 1,000 low-quality submissions per month, the maintainers face a choice: spend four times as much time reviewing, or lower their review quality and let problems through. Neither option is sustainable.