Stay Away from My Trash
tldraw has adopted a new policy to automatically close external pull requests, citing an overwhelming influx of low-quality, AI-generated contributions. This move highlights a growing challenge for open-source maintainers navigating the age of AI-assisted coding. The author provocatively questions the overall value of external code when AI can generate formally correct but contextually flawed solutions with ease.
The Lowdown
tldraw, an infinite canvas SDK, has implemented a new policy to automatically close pull requests from external contributors, citing a significant increase in low-quality, AI-generated submissions. This decision, initially made with trepidation, has received surprisingly positive feedback from other maintainers facing similar issues.
- The problem stems from AI-generated pull requests that, while "formally correct" and passing checks, lack crucial context, ignore project patterns, and are often abandoned.
- The author notes that they and their team already use AI tools for coding, emphasizing that the issue isn't AI use itself, but the lack of contextual understanding that external AI contributions bring.
- This phenomenon, dubbed "Eternal Sloptember," sees repositories receiving a surge of superficially good but ultimately useless PRs, often based on vague issue descriptions.
- A key insight came from the author's own AI-generated issue creation tool, which, while useful for personal workflow, produced "well-formed noise" that external AI contributors would then attempt to "solve" with equally uncontextualized code.
- The author concludes that while their own "shitty AI issue" provided value by capturing ideas, the resulting external AI-generated solutions provided none, as they bypassed the critical human step of evaluating an issue's true intent and feasibility.
This situation has led tldraw to restrict external code contributions, suggesting a fundamental shift in open-source engagement where human judgment and contextual understanding remain paramount over easily generated code. The value of external contribution now lies more in reporting, discussion, and perspective rather than code implementation.