The coding agent market is already one tool wide. claude code is 80% of the 166.7m agent commits…
the coding agent market is already one tool wide. claude code is 80% of the 166.7m agent commits measured on public github, the top three tools own 93%, and the other 16 tracked agents split the last 1%.
everyone argues about model concentration, but the real single point of failure is the harness.
Context
The GitHub Coding Agent Monitor counts new public commits that carry a coding agent's signature, using the GitHub Search API on a daily schedule. It tracks ten agents, including Claude Code, Cursor, GitHub Copilot, Devin, Aider, OpenAI Codex, OpenCode, Google Jules and Amazon Q. When read, its chart showed a 10 day rolling average of the top three as a share of all public commits: Claude Code 2.66%, Cursor 0.25%, GitHub Copilot 0.01%.
The tracker lists its own limits: only public repositories, only commits where an agent left a signature, and only default branches. It says some agents may be more common in private repos or may not show up because they leave no signature, and asks readers to be wary of conclusions.
The arXiv census (2026) combines configuration-file scanning, commit-message analysis, author patterns and bot lookup across more than 180 million repositories. It reports that bot-account lookup recovers only 3.3% of the Claude Code commits found by the full method, a roughly 30 times gap, and that single-signal estimates are biased low. By its April 2026 snapshot it counts Claude Code at 886,122 commits across 17,295 projects, followed by Jules at 215,804.
The figures in the note are 166.7 million agent commits, Claude Code at 80%, the top three at 93% and 16 other agents sharing the last 1%. None of those numbers appears in the tracker page, the census paper or the other sources read. They are unsupported here, not refuted. The tracker tracks ten agents, not nineteen, and measures share of all public commits, not share of agent commits, so its percentages are not the same quantity.
Both sources say agent share depends on how agents are detected. A ranking built from signatures favours tools that leave a signature, and the census says no single method captures more than a fraction of activity. Whether concentration holds across all agents is therefore a question about detection as much as about the market.
'The real single point of failure is the harness' is the author's opinion. The sources read measure how many commits are attributed to each tool, not how failures or outages spread.
Related work
- GitHub Coding Agent Monitor (powerset-co/github-coding-agent-tracker, GitHub) ↗Source for the signature method, the ten tracked agents and the stated limits.
- Detecting AI Coding Agents in Open Source: A Validated Multi-Method Census of 180 Million Repositories (arXiv) ↗Source for the census of 180 million repositories and the detection gap.
- Agentic Much? Adoption of Coding Agents on GitHub (arXiv) ↗Earlier study of coding agent adoption on GitHub, based on agent-authored pull requests.
Watch next
- Find where the 166.7 million commit total and the 80 and 93 percent shares come from. Check whether any dataset ranks all nineteen agents the note counts.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 8 October 2026 at 22:19 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →