ai-tagged commits — tracked daily since 2025
THE AI CODE INDEX
How many public GitHub commits carry an AI coding tool’s own tag, every day. Daily volume, share of self-identified AI commits, and doubling time across 7 tools shown (11 tracked). Every number is a lower bound: tools that leave no tag are invisible, and so are private repos. how we count →
this series is available via the api — developers →
§ 1 — total daily commit volume, all tools
fig. 01
172M commits · last 6 months · ▲ 319% vs prior span
latest complete day — nov 10 · 1.9M
§ 2 — share of self-identified ai commits
fig. 02
share of commits that carry a tool’s tag — not share of developers, seats or enterprise use. each tool is counted differently; the table below says how.
how we count — one row per tool
method v1
Every series counts only commits that carry the tool’s own tag on public GitHub. It is a lower bound on self-identified AI commits, not market share: private and enterprise repos are invisible, and a vendor changing its default tag setting moves its line overnight.
▸ how each of the 11 tools is counted▾ hide the per-tool detail
| tool | counted when a commit… | tag default | since | shown |
|---|---|---|---|---|
| claude code | carries the trailer Co-Authored-By: ClaudeClaude Code adds the trailer by default; users can turn it off. Counts commits where it was left on. | tag on by default | 2025-10-08 | yes |
| cursor | carries the trailer Co-authored-by: CursorTagged on commits made by Cursor’s cloud and background agents; in-editor edits leave no tag. The series has detector regime breaks between Jan and May 2026, so compare within a regime. | tagged in some modes | 2025-10-08 | yes |
| github copilot | is authored by author:copilot-swe-agent[bot]Counts only commits authored by the Copilot coding agent. Autocomplete and chat use, most of Copilot’s volume, leave no trace in commits. VS Code’s optional "Co-authored-by: Copilot" trailer is tracked as its own series below, not here. | tagged in some modes | 2025-10-08 | yes |
| github copilot (co-author tag) | carries the trailer Co-authored-by: CopilotVS Code’s git.addAICoAuthor trailer. Opt-in, except default-on from 22 Apr 2026 (VS Code 1.117) until the revert to off in 1.119 (May 2026). Kept separate from the coding-agent series so that window is visible as a regime, not hidden inside a trend. Do not add the two Copilot series together: agent commits can carry the trailer too. | tagged in some modes | 2026-02-01 | collected only |
| aider | carries the trailer Co-authored-by: aiderAider’s default attribution edits the git author name rather than adding a trailer; this series counts the trailer only. We have not verified the current default. | default not verified | 2025-10-08 | yes |
| gemini cli | contains the text gemini-code-assistLoose text match. May also count commits from the Gemini Code Assist GitHub app, not only the CLI. | default not verified | 2025-10-08 | yes |
| devin | carries the trailer Co-authored-by: DevinDevin often commits under its own bot account rather than a trailer; those commits are not counted here. | default not verified | 2025-10-08 | yes |
| codex | carries the trailer Co-authored-by: CodexSources disagree on whether and since when Codex tags commits. Treat the level as a floor and the trend as indicative only. | default not verified | 2025-10-08 | yes |
| junie | carries the trailer Co-authored-by: JunieCollected since 5 Mar 2026. Not yet shown on the public chart. | default not verified | 2026-03-05 | collected only |
| amazon q | contains the text amazon-q-developerLoose text match. Collected since 5 Mar 2026. Not yet shown on the public chart. | default not verified | 2026-03-05 | collected only |
| cody | carries the trailer Co-authored-by: CodyCollected since 5 Mar 2026. Not yet shown on the public chart. | default not verified | 2026-03-05 | collected only |
§ 3 — the other tools, same scale
fig. 03
cursor
37.8k/day · ▲ 1%
codex
7.3k/day · ▼ 20%
devin
6.4k/day · ▼ 27%
github copilot
1.6k/day · ▼ 51%
aider
165/day · ▼ 57%
gemini cli
125/day · ▼ 2%
claude code is left out so the smaller tools are readable — one shared scale for everyone else.
§ 4 — momentum, ranked
fig. 04
| no. | tool | momentum 30d | commits/day | doubles every | share |
|---|---|---|---|---|---|
| 01 | claude codefastest | ▲ 9% | 1.9M | 242 days | 97.2% |
| 02 | cursor | ▲ 1% | 37.8k | 2238 days | 2.0% |
| 03 | gemini cli | ▼ 2% | 125 | — | 0.0% |
| 04 | codex | ▼ 20% | 7.3k | — | 0.4% |
| 05 | devin | ▼ 27% | 6.4k | — | 0.3% |
| 06 | github copilot | ▼ 51% | 1.6k | — | 0.1% |
| 07 | aider | ▼ 57% | 165 | — | 0.0% |
01claude codefastest
▲ 9% · 1.9M/day · 242d to double · 97.2%
02cursor
▲ 1% · 37.8k/day · 2238d to double · 2.0%
03gemini cli
▼ 2% · 125/day · — · 0.0%
04codex
▼ 20% · 7.3k/day · — · 0.4%
05devin
▼ 27% · 6.4k/day · — · 0.3%
06github copilot
▼ 51% · 1.6k/day · — · 0.1%
07aider
▼ 57% · 165/day · — · 0.0%
§ 5 — the pr arena
fig. 05
fig. 05 — autonomous agent prs, all time
github copilot and devin don’t sign their commits, so they score zero in the charts above. zero there isn’t absence — it’s invisibility. this is their arena.
§ 6 — the open model layer
fig. 06
models — release snapshots (big stars, low velocity)
- mistralai/mistral-vibeMistral Vibe is a free, open-source AI assistant that lets developers have a conversation with their codebase directly from their computer's command line, using plain English to explore, edit, and manage code projects. It's built by Mistral, a leading European AI company, and runs on their own AI models.15/100Active5.4k★ · 796⑂
- meta-llama/llamaThis is Meta's official release of Llama 2, a family of powerful AI language models that anyone — from individual developers to large companies — can download and run on their own systems. It provides the core model files and basic code needed to load and use these AI models, which range in size and capability, for tasks like answering questions, writing, and conversation.12/100Active59.6k★ · 9.8k⑂
- deepseek-ai/DeepSeek-V3DeepSeek-V3 is a powerful open-source AI language model — similar to the technology behind ChatGPT — developed by the Chinese AI lab DeepSeek. It can understand and generate text, answer questions, write code, and handle complex reasoning tasks, and is freely available for anyone to use or build products on top of.6/100Active105k★ · 16.7k⑂
- deepseek-ai/DeepSeek-R1DeepSeek-R1 is a family of open-source AI models that are exceptionally good at reasoning through complex problems in math, coding, and logic — trained using a technique called reinforcement learning, where the AI essentially teaches itself by trial and error rather than being hand-guided. DeepSeek claims these models perform on par with OpenAI's most advanced reasoning AI (o1), and they're freely available for anyone to download and use.3/100Active92.0k★ · 11.7k⑂
- zai-org/GLM-4.5—4.4k★ · 486⑂
runtime — where open-model energy lives
- ollama/ollamaOllama lets you run powerful AI models — like DeepSeek, Gemma, and Qwen — directly on your own computer or server, without sending data to any third-party cloud service. It works on Mac, Windows, and Linux, and connects to popular AI-powered coding tools and chat apps you may already use.52/100Hot187k★ · 19.1k⑂
- unslothai/unslothUnsloth is a desktop app that lets you run and train AI models — including chatbots, image generators, and text-to-speech systems — directly on your own computer, without sending data to the cloud. It supports all the major AI models making headlines right now, like DeepSeek, Gemma, and Qwen, and gives you a visual interface to interact with and customize them.45/100Hot79.4k★ · 7.4k⑂
- ggml-org/llama.cppllama.cpp lets developers run AI language models directly on their own computers or servers — no internet connection, no cloud service required — with support for virtually any modern hardware, from laptops to high-end machines. It strips away the complexity typically involved in running these models, making local AI accessible without expensive infrastructure.41/100Hot135k★ · 25.5k⑂
- MoonshotAI/kimi-codeKimi Code is an AI-powered assistant that lives in your computer's command line and acts like a tireless junior developer — it can read and modify your codebase, run programs, search the web, and make decisions on its own to complete tasks. It's built by Moonshot AI and connects to their Kimi AI models, though it can also work with other AI providers.15/100Active8.2k★ · 1.4k⑂
the tools above run on these. open models don’t tag their commits, so this layer is repo traction — not usage share. anything else would be made up.
§ 7 — ecosystem adoption
fig. 07
config files in repos
sdk / package adoption
note: agents.md (an open format) beats every vendor-specific config file.
§ 8 — community pulse
fig. 08
fig. 08 — community pulse, 7 days
| tool | mentions | hn | github | |
|---|---|---|---|---|
| claude code | 1.2k | 86 | — | 1.1k |
| github copilot | 339 | 1 | — | 338 |
| codex | 182 | 0 | — | 182 |
| gemini cli | 50 | 1 | — | 49 |
| cursor | 38 | 0 | — | 38 |
| aider | 10 | 0 | — | 10 |
| devin | 3 | 0 | — | 3 |
§ 9 — intelligence brief
fig. 09
$ tail -f /var/log/gitfind/index.log
09:41:02 index sync completed — 1.9M commits counted across 7 of 7 tools
09:41:03 claude code carries 97.2% of the tool-tagged commits we count
09:41:04 claude code added 149k/day this month — biggest absolute gain
09:41:05 claude code is the fastest doubler — every 242 days at current pace
09:41:06 devin leads the pr arena with 1.5k agent prs — without signing a single commit
09:41:07 aider down 57% this month — watching for a floor
09:41:08 █
form 27-b — subscription
THE TUESDAY BRIEFING
The repos that moved this week, why they matter, and what to watch next. One email. No noise.
building on this data? the catalog and the indexes are also an api — developers →