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.

A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.

01Result

NVIDIA/TransformerEngine ↗

NVIDIA/TransformerEngine AI coding statistics: 6.6%

of 564 commits in February–September 15, 2026 were committed by an AI agent — a floor, not an estimate
6.6% agent · top contributor 35.1%17 contributors
org-wide

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

37
agent commits
527
other considered commits
17
agent contributors
35.1%
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

37assisted
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
2026-022.2%45
2026-035.8%69
2026-042.3%88
2026-0510.6%66
2026-0616.9%71
2026-073.9%76
2026-082.2%92
2026-0910.5%57

Peak month: 16.9%. Bars are scaled to that peak, not to 100%.

Which agents

AgentVendorCommitsShare
Claude CodeAnthropic335.9%
OpenAI CodexOpenAI20.4%
Cursor AgentAnysphere20.4%

03Evidence

Inspect the matches

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

Matched commit samples

  • f72bc1ad on GitHubThread tanh logit softcapping through FlashAttention (FA2 and FA3) (#3391)Claude Code
  • 39a54018 on GitHub[PyTorch] Fix NaN expert weight gradients at num_groups == 1 with SReLU (#Claude Code
  • b97f5ab9 on GitHub[PyTorch] Use torch's register_custom_class API for opaque quantizers whenClaude Code
  • f95573f1 on GitHub[PyTorch] Refactor grouped linear and grouped MLP tests (#3122)OpenAI Codex
  • 91bb9cfe on GitHub[PyTorch] Refactor grouped MLP into joint forward-backward fused op (#3117OpenAI Codex
  • be37e9b7 on GitHub[common] Grouped gemm update - nvfp4 for blackwell and fp8 blockwise hoppeCursor Agent
  • 282b4fb2 on GitHub[torch.compile][PyTorch] Prepare linear for torch compile (#2967)Cursor Agent

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 window564
Merge commits removed−0
CI / dependency bots removed−0
Considered564

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 →Cursor Agent commitsBrowse repositories with detected Cursor Agent signatures →OpenAI Codex commitsBrowse repositories with detected OpenAI Codex 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 NVIDIA/TransformerEngine

Scan another repository

Try