A recent analysis of 25,264 agentic pull requests across 2,361 popular GitHub repositories reveals that intensive adoption of agentic coding tools remains limited to a small subset of open-source projects. The study indicates that the median repository generates only one to two such PRs over a three-month period.
- Small projects with 1-5 contributors exhibit higher participation ratios and average levels of agentic PR activity compared to medium-sized and large projects.
- Productivity varies substantially, with most projects falling below an industry estimate of 36 PRs per participant during the observation window.
- Human-agent collaboration is predominantly characterized by a single-human oversight model, where one developer reviews or modifies agent contributions.
The findings suggest that successful integration of agent-generated contributions depends on human and organizational processes governing their use, not just agent capabilities.