Early Adoption of Agentic Coding Tools by GitHub Projects

2026-05-29 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Expert, extended

Summary

An analysis of 25,264 agentic pull requests (PRs) from 2,361 popular GitHub repositories, using data from May-July 2025, reveals early adoption patterns of agentic coding tools like Copilot, Codex, and Claude Code. The study found that the median repository generates only one to two agentic PRs during this three-month period, indicating that intensive adoption is concentrated in a small subset of projects. Small projects (1-5 contributors) show significantly higher participation ratios and average agentic PR activity compared to medium or large projects. Project-level agentic PR productivity varies substantially, with only 25 out of 2,361 projects (1%) exceeding an industry benchmark of 36 PRs per participant. Furthermore, human-agent collaboration is predominantly a single-human oversight model, where one developer both reviews and modifies the agent's contributions, accounting for 78.9% of agentic PRs. Multi-human collaboration remains comparatively rare.

Key takeaway

For software engineering teams considering agentic coding tools, recognize that adoption is currently uneven and often concentrated in smaller projects. You should anticipate a predominant single-human oversight model for agent-generated contributions, requiring robust individual review processes. Focus on adapting your team's review practices and organizational processes, not just agent capabilities, to ensure sustainable integration and manage varied productivity outcomes.

Key insights

Agentic coding tools see uneven adoption, concentrated in small projects with single-human oversight, and varied productivity.

Principles

• Agent integration needs human/org processes.
• Project size impacts agentic workflow participation.
• Human review remains central to agentic workflows.

Method

Analyzed 25,264 agentic PRs from 2,361 GitHub repos (AIDev-pop, May-July 2025). Classified projects by contributor count and PRs by five human participation patterns to assess adoption and productivity.

Topics

• Agentic Coding Tools
• GitHub Pull Requests
• Human-Agent Collaboration
• Software Development Workflows
• Open-Source Projects
• Code Review Practices

Best for: AI Scientist, Research Scientist, Software Engineer

Related on AIssential

• Early Adoption of Agentic Coding Tools by GitHub Projects — Early adoption of agentic coding tools is uneven, with single-human oversight dominating collaboration patterns.
• Collaborative AI Engineering: One Dev, Two Dozen Agents, Zero Alignment — Maggie Appleton, GitHub — AI agents accelerate code generation, but current tools fail at team alignment, creating a new bottleneck.
• How Do AI Coding Agents Contribute to Software Development? an Empirical Study of Agentic Pull Requests — The study empirically characterizes AI coding agent contributions to software development via pull request analysis.
• A Year Of Vibes — Agentic coding tools are fundamentally reshaping software engineering workflows, challenging traditional practices and human-machine relationships.
• Augmentation with Dilution: A Large-Scale Empirical Study of Human Contributor Ecosystems After AI Coding Agent Adoption — AI coding agents dilute human participation and increase review burden in open-source projects without displacing contributors.
• Code Is Not Cheap: How to Multiply Your AI’s Output With Software Fundamentals — AI-assisted coding, while prevalent, demands human engineering expertise to mitigate quality issues from "jagged" model outputs.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.