110K PR Dataset Reveals How Autonomous Coding Agents Actually Contribute to Open Source Projects
arXiv·medium signal
Constructs a novel dataset of approximately 110,000 open-source pull requests including contributions from autonomous coding agents, analyzing activity patterns and code change quality over time. First large-scale empirical study of real-world autonomous agent contributions to production repositories, examining effects on code quality, team dynamics, and software maintainability. Directly relevant as agent-authored PRs become commonplace.