AIDev: Studying AI Coding Agents on GitHub
The first massive dataset shows how AI is secretly transforming software development.
Deep Dive
A new research paper introduces AIDev, the first large-scale dataset tracking AI coding agents in real-world projects. It aggregates 932,791 agent-authored pull requests (Agentic-PRs) from five major agents—OpenAI Codex, Devin, GitHub Copilot, Cursor, and Claude Code—across 116,211 GitHub repositories involving 72,189 developers. The dataset provides a foundational look at AI adoption, productivity, and human-AI collaboration in modern software engineering.
Why It Matters
This data is crucial for understanding the real, massive impact AI is having on how all software is built.