Pulse
2026-09-18-v3.0
Dataset
2026-10-27-catalog-ai-v2
Ruleset
b2e8889c66f7
Scanned
2026-10-27
Window
February–September 15, 2026
Catalog tracking observation. This repository is measured with the same fixed window and ruleset, but it is not part of the published benchmark or its aggregate claims.

Multimodal RL training framework for diffusion & omni models

01Result

verl-project/verl-omni ↗

verl-project/verl-omni AI coding statistics: 19.9%

of 377 commits in February–September 15, 2026 were committed by an AI agent — a floor, not an estimate
19.9% agent · top contributor 24%16 contributors
org-wide

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

75
agent commits
302
other considered commits
16
agent contributors
24%
largest contributor

02Drivers

What shapes this result

Mode, timing, and detected signatures show whether the result reflects broad adoption or one concentrated workflow.

Assisted or autonomous

75assisted
0autonomous

Assisted — a developer’s identity authored the commit and the agent added itself as co-author. Autonomous — a bot identity authored the commit without a human identity attached.

Agent share by month

Agent-attributed commit share by month
MonthAgent shareConsidered commits
2026-0415.1%53
2026-0524.4%45
2026-0626.2%65
2026-0717.5%57
2026-0817.2%93
2026-0920.3%64

Peak month: 26.2%. Bars are scaled to that peak, not to 100%.

Which agents

AgentVendorCommitsShare
Gemini CLIGoogle318.2%
Cursor AgentAnysphere236.1%
GitHub CopilotGitHub195%
Claude CodeAnthropic71.9%
OpenAI CodexOpenAI10.3%

03Evidence

Inspect the matches

Public commit samples let you inspect the signatures behind the aggregate result.

Matched commit samples

  • ab5dfbaf on GitHub[trainer] feat: add Qwen3-Omni multimodal DPO training (#269)Gemini CLI
  • 887a6b3e on GitHub[1/N][data, omni, tests] feat: add Omni-Preference offline MLLM DPO dataseGemini CLI
  • 18a99343 on GitHub[rollout, diffusion] refactor: integrate Qwen-Image step execution into FlGemini CLI
  • 298eb0cc on GitHub[4/N][recipe, doc, trainer, rollout] feat: add Qwen-Image FlowGRPO v1 syncCursor Agent
  • 62f056d6 on GitHub[doc] chore: add slides badge next to Slack in README (#579)Cursor Agent
  • 63668387 on GitHub[recipe, doc] chore: default Wan2.2 DanceGRPO docs to v1 and deprecate v0 Cursor Agent
  • 4e16d84d on GitHub[ci] chore: route bug reports to maintainers (#577)GitHub Copilot
  • b92534c8 on GitHub[2/N][reward] feat: add pointwise audio HTTP reward support (#528)GitHub Copilot
  • 33a2317b on GitHub[tests] feat: add Bagel PickScore LoRA e2e smoke (#379)GitHub Copilot
  • dbf4ad2f on GitHub[trainer, tests, doc] feat: async teacher scheduling for diffusion OPD on Claude Code
  • 75f19f6d on GitHub[trainer, tests] fix: union colocated reward output into the v1 diffusion Claude Code
  • dc2c4b0b on GitHub[diffusion] feat: add Boogu-Image FlowGRPO support (T2I + Edit) (#332)Claude Code
  • 67d4e1c3 on GitHub[trainer] fix: fix the rollout seed (#136)OpenAI Codex

04Method

What entered the denominator

The result begins with commit history, removes merges and routine automation, and reports literal agent signatures as a lower bound.

Exclusion ledger

Commits in window379
Merge commits removed−2
CI / dependency bots removed−0
Considered377

Bots are excluded from the denominator entirely. Agent attribution requires a literal supported signature; no code-style inference is used.

Continue exploring

Claude Code commitsBrowse repositories with detected Claude Code signatures →GitHub Copilot commitsBrowse repositories with detected GitHub Copilot signatures →Cursor Agent commitsBrowse repositories with detected Cursor Agent signatures →OpenAI Codex commitsBrowse repositories with detected OpenAI Codex signatures →Gemini CLI commitsBrowse repositories with detected Gemini CLI signatures →Python AI coding statisticsCompare evidence across Python repositories →GitHub AI commit analyzerInspect another repository with the same evidence model →
Reproduce this observation

Run the same detector locally.

Deterministic signature matching; no code-style inference. Squash merges can remove trailers, so this result remains a measured floor.

npx qmmit-cli verl-project/verl-omni

Scan another repository

Try