PR Spam Mirrors Early 2000s Email Spam Wave
Original: PR spam today looks like email spam in the early 2000s
Why This Matters
Highlights emerging quality control challenges in open source as AI-generated contributions scale, requiring new reputation and trust infrastructure.
Greptile analyzed pull requests on OpenClaw, which saw submissions surge from 2/week to 3,400/week by February 2025, with merge rates dropping from 48% to 9.3%, revealing AI-generated spam patterns similar to early email spam.
Greptile, an AI pull request review platform, studied OpenClaw, GitHub's fastest-growing repository, and documented a dramatic shift in contribution patterns. In December 2024, OpenClaw received approximately two pull requests per week. By February 2025, that number jumped to 3,400 per week. Simultaneously, the merge rate collapsed from approximately 48% to fewer than 9.3%, indicating quality degradation. Analysis revealed many low-effort submissions generated by AI coding agents. One contributor submitted 106 PRs in a single day with a median submission interval of three seconds. The study identifies three key observations: First, contributor reputation will become essential filtering mechanism, similar to email sender reputation scores that emerged to combat spam. OpenClaw data shows merge rates correlate with contributor history: 8.2% for first-time contributors, 10.3% for those with 2-5 PRs, and 18.6% for those with 5+ PRs. Second, tools like Vouch, created by Ghostty maintainer Mitchell Hashimoto, implement trust management systems for open source projects. Vouch prevents unvouched users from contributing and flags bad actors, functioning as a sender reputation score equivalent for open source. Third, homogenization of AI-generated contributions threatens diverse problem-solving approaches. The analysis found multiple contributors independently submitted identical PRs with the same titles, reducing the diversity of perspectives traditionally valued in open source development.