Playbooks · Coding tasks
180M repos: most agent work is unattributed
Written November 8, 2026
A 180M-repo census found 850,157 Claude Code commits — bot lookups found 28,154. Why GitHub counts miss most AI coding, and what that means for reviewers.
What the census did
A paper by Arsham Khosravani and Audris Mockus, submitted in June 2026, ran a multi-method census across World of Code, more than 180 million Git repositories. It combined four detection methods: configuration-file scanning, commit-message analysis, author-identity matching, and bot-signature lookup, classifying agent traces into four behavioural types. Every detection pattern was hand-validated, with 495 labels checked and per-cell precision reported.
The headline comparison is between two ways of counting the same thing. Multi-method detection found 850,157 Claude Code commits in one snapshot. Bot-account lookup, the signal most adoption studies rely on, recovered only 28,154 of them: 3.3 percent, a 30-fold relative-recall gap. The paper's conclusion is that single-signal prevalence estimates are biased low by at least that factor. Across snapshots from December 2024 to April 2026, commit-attributed agents generated over 320,000 commits per month. Claude Code leads, with 886,122 commits across 17,295 projects, and dominates silent, configuration-file-only adoption, with 21,078 projects showing agent configuration and no other trace.
Why bot counts miss most of it
No single method captures more than a fraction of agent activity, and the methods capture different populations. Compared against an independent pull-request census, the two channels capture nearly disjoint agent populations: a pull-request census misses 79 percent of commit-detected Claude Code adopters and essentially all Codex adopters. The work profile follows deployment and detection mode rather than the tool itself. Pull-request-deployed cloud agents, such as Codex and Cursor, surface as feature work, while commit-deployed in-editor agents, such as Claude Code, OpenHands, and Aider, surface as maintenance. No single channel is representative, because each channel sees the work its deployment mode produces.
That is the mechanism behind the gap. A bot account exists only where the agent acts as a bot: a cloud agent opening a pull request, or a commit authored under a bot identity. An agent running inside an editor, committing as the developer, leaves no bot account behind. It may leave a configuration file, a commit message pattern, or nothing at all. Counting bot accounts therefore counts one deployment mode, and mistakes it for the whole population.
What a reviewer should do instead
- Do not quote a bot-count share as an adoption rate. A number built on bot accounts is a count of one channel, biased low by at least a factor of 30 for Claude Code in this census. Treat it as a floor for one deployment mode, not as a share of all agent work.
- Attribute from the commit content, not the account name. Configuration files, commit message patterns, and author identity each catch populations the others miss. A review that looks only at who authored a commit will miss the in-editor work entirely. The fingerprinting guide covers what to look for in a pull request.
- Ask which channel a number came from before repeating it. A pull-request census and a commit census measure different populations doing different kinds of work. A figure without its channel attached is not comparable to a figure from another channel.
- Do not use the census to attribute any single commit. The census measures populations across 180 million repositories. It proves nothing about who wrote any given commit, and this page does not present it as doing so.
Limits
- The paper is on arXiv, submitted in June 2026. It is a research preprint, and its figures are the paper's, from its detection framework and its validation. This page reports them as the paper reports them.
- Detection is probabilistic. Patterns were hand-validated on 495 labels, with per-cell precision and confidence intervals, which means the counts carry measurement error the paper itself quantifies. Neither number on this page is an exact count of agent-written commits.
- The census covers open-source repositories in World of Code. It says nothing about private repositories, where in-editor agent use may be higher still, and this page does not extrapolate it there.
- No census was reproduced for this page. The figures are read from the paper, not measured again.
Related playbooks
- Dependabot pauses alerts on quiet repos
- Triage a backlog of issues with an agent
- Find a DOI by title
- Document a codebase so a stranger can work in it
Sources
Back to: Coding playbooks hub · Also: Fingerprinting AI coding agents from 33,580 PRs · METR: Anthropic's AI R&D sped up about 1.9 times · BrokenML: a benchmark of broken machine-learning jobs