Several large AI era open source projects are moving away from accepting external pull requests and using AI agents to triage issues and write fixes instead, according to Latent Space.
Vercel's AI SDK, which has more than 20 million weekly npm downloads, uses specialized agents to manage over 1,000 open issues and roughly 800 pull requests, with agents authoring 25% to 35% of merged PRs and closing 70% to 80% of issues, Latent Space reported.
Astro, which has 62,000 GitHub stars, built an auto triage system that gets users to verify a fix works before a maintainer reviews it. "Issues came in faster than we could handle them," creator Fred Schott said, according to Latent Space. The project tldraw, with 50,000 stars, and the project Flue now close external pull requests automatically, converting them into issues or discussions, citing what Flue calls "drive-by AI slop PRs" and, for tldraw, security concerns, founder Steve Ruiz said.
"It just makes less sense to have people contributing code if... the code can be written by agents," Ruiz said, according to Latent Space. Developer Mitchell Hashimoto suggested large open source projects will eventually close contributions completely.
Open source has run on the pull request model for two decades. If the projects with the most traffic are quietly replacing outside contributors with in-house agents, the on-ramp that turned users into maintainers on projects like these is closing just as more people are learning to code with AI help.