← the index
vllm-project/vllm ↗
11.6%
of 7,312 commits were committed by an AI agent — a floor, not an estimate
11.6% agent · top contributor 3.3%413 contributors
org-wide

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

850
agent commits
6,462
human commits
413
agent contributors
3.3%
largest contributor

Assisted or autonomous

850assisted
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: 17.4%. Bars are scaled to that peak, not to 100%.

Which agents

Claude CodeAnthropic6058.3%
Gemini CLIGoogle1091.5%
Cursor AgentAnysphere741%
OpenAI CodexOpenAI530.7%
GitHub CopilotGitHub300.4%
DevinCognition20%

Verify it yourself

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

  • ee0a4c46[Bugfix] Account for PCP in multi-node world size validation (#55111)Claude Code
  • 859dd395[Bugfix][Core] Wait for the previous PP tensor sends before the next forwaClaude Code
  • d539de1c[Docs] Add missing return annotations flagged by griffe (#54980)Claude Code
  • 6f74337c[Bugfix][Spec Decode]Preserve user --speculative-config overrides for specGemini CLI
  • 8fea1d30Fix uniform_random routing simulation to sample without replacement (#4368Gemini CLI
  • bf2b45b5[Attention] Integrate FlashAttention 4 SM100 headdim 256 support (#42669)Gemini CLI
  • bb363db9feat: Add support for reasoning_token_count to reasoning parser (#54982)Cursor Agent
  • 5d09eb2c[Bugfix][MoE] Preserve unquantized weight storage on ROCm (#46009)Cursor Agent
  • 8f03625b[CPU][Zen] Route Int8 MoE inference through zentorch on AMD (#44834)Cursor Agent
  • 848ab131[Perf] Accumulate Conformer attention scores with baddbmm (#55062)OpenAI Codex
  • 2a4e3cc3[Bugfix][KV Offload] Ensure tracker progress for oversized offers (#54759)OpenAI Codex
  • f870b929[Qwen3.8-Flash-Next] Fuse Qwen4Exp PLE kernels (#54517)OpenAI Codex
  • a0d3e5c1[XPU] Add fused GemmaRMSNorm path for eager execution (#53678)GitHub Copilot
  • c296cf82[Bugfix] Add forward_xpu to XDRotaryEmbedding for HunyuanOCR on XPU (#5217GitHub Copilot
  • 19d61b12[Bugfix] Initialize DeepGemmQuantScaleFMT oracle lazily; bound QuantFP8 UEGitHub Copilot
  • d3d79ffc[Bugfix][Spec Decode] Capture the widest uniform decode batch by default (Devin
  • 044b0522[Kernel][Perf] Tune fused_moe FP8 config for Qwen3.5 on L40S (+7%) (#53819Devin

What was excluded

Commits in window7,315
Merge commits removed−0
CI / dependency bots removed−3
Considered7,312

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 vllm-project/vllm

Scan another repository

Try