11.1%

of 386 commits were committed by an AI agent — a floor, not an estimate

11.1% agent · top contributor 41.9%14 contributors

org-wide

Adoption is broad. Many contributors are committing agent-authored work, which is what a real organisational rollout looks like.

agent commits

343

human commits

agent contributors

41.9%

largest contributor

Assisted or autonomous

43assisted

0autonomous

Assisted — a developer’s identity authored the commit and the agent added itself as co-author. The agent ran git commit during a working session, with a person there. Autonomous — a bot identity is the author, no human in the loop. Across the whole index, 93.7% of agent commits are assisted.

Agent share by month

Feb26

Mar26

Apr26

May26

Jun26

Jul26

Aug26

Sep26

Peak month: 21.6%. Bars are scaled to that peak, not to 100%.

Which agents

Claude Code	Anthropic	27	7%
GitHub Copilot	GitHub	11	2.8%
Cursor Agent	Anysphere	4	1%
OpenAI Codex	OpenAI	2	0.5%

Verify it yourself

Sample commits that matched. Every one links to GitHub — check the trailers.

e3364bdcoptional tail silence to flush the decoder (#16237)Claude Code
c40f18ddspeechlm2: add deterministic Granary-v2 preference sampling for restorableClaude Code
617a5fd5fix-tutorial-install-line (#15879)Claude Code
e3364bdcoptional tail silence to flush the decoder (#16237)GitHub Copilot
6ffd048bDistributed OOMptimizer for speechlm2 training (#15807)GitHub Copilot
52ba8338Add hi-IN , Ko-KR and pt-BR IPA tokenizer support (#15567)GitHub Copilot
5ff60930fix(asr): make DynamicLengthTensor indexing Python 3.10-compatible (#15797Cursor Agent
c66a3796Add vLLM support for NeMo SpeechLM (#15520)Cursor Agent
86264e7cdocs: fix broken tutorial links pointing to non-existent stable branch (#1Cursor Agent
059d9039feat(speechlm2): add native MTP speculative decoding in vLLM (#16196)OpenAI Codex
e8acc244feat(speechlm2): export portable checkpoints and serve them with vLLM (#16OpenAI Codex

What was excluded

Commits in window	395
Merge commits removed	−0
CI / dependency bots removed	−9
Considered	386

Bots are excluded from the denominator entirely. Counting dependabot as “AI-written code” is the fastest way to produce a meaningless number.

How this was measured

Measured over February–August 2026, cloned directly. Detection is deterministic string matching against signatures agents write about themselves — no heuristics, no style analysis. Merge commits and CI bots are excluded from the denominator. The figure is a floor: squash merges strip trailers, so real usage is higher. Contributor counts are aggregate — no individual is named.

Run it yourself: npx qmmit-cli NVIDIA-NeMo/Speech

Scan another repository

Try