Pulse
2026-09-18-v3.0
Dataset
2026-09-18-v3.0
Ruleset
b2e8889c66f7
Scanned
2026-09-16
Window
February–September 15, 2026
Local to this browser

01Result

Dao-AILab/flash-attention ↗

Dao-AILab/flash-attention AI coding statistics: 3.9%

of 335 commits in February–September 15, 2026 were committed by an AI agent — a floor, not an estimate
3.9% agent · top contributor 23.1%10 contributors
org-wide

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

13
agent commits
322
other considered commits
10
agent contributors
23.1%
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

13assisted
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. Across the whole index, 96% of agent commits are assisted.

Agent share by month

Agent-attributed commit share by month
MonthAgent shareConsidered commits
2026-021.4%71
2026-032.4%123
2026-045.1%39
2026-057.1%28
2026-063.6%28
2026-076.7%15
2026-086.7%15
2026-0912.5%16

Peak month: 12.5%. Bars are scaled to that peak, not to 100%.

Which agents

AgentVendorCommitsShare
Claude CodeAnthropic92.7%
GitHub CopilotGitHub30.9%
Cursor AgentAnysphere10.3%

03Evidence

Inspect the matches

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

Matched commit samples

  • 3c2c6b31 on GitHub[CuTe, SM90] Fix swapped backward MMA layout and use 2-stage hdim256 confiClaude Code
  • d8dbfad5 on GitHubFix `seqlen_k_loaded` treating a window bound of 0 as unbounded (sliding wClaude Code
  • c68c592f on GitHub[CuTe, SM100] Sparse MLA bwd: don't scatter dK/dV at -1 sentinel indices (Claude Code
  • 59f01d6e on GitHubBump AITER submodule to commit 3b2e6f4 (#2540)GitHub Copilot
  • 3da76cdb on GitHub Build Fix: Update abi3 tag to cp310 and minimum python version to 3.10 (#GitHub Copilot
  • b46c5875 on GitHub[AMD ROCm] Fix NaN in FMHA BWD when seq_q=0 (#2421)GitHub Copilot
  • abaa8787 on GitHub[CUTE]Bump to Cutedsl (#2216)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 window335
Merge commits removed−0
CI / dependency bots removed−0
Considered335

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

05Provenance

Pinned to a source state

This observation names the repository state, release, and detector that produced it. Use these identifiers when reproducing or citing the result.

Dataset release
2026-09-18-v3.0
Scan date
2026-09-16
Ruleset
sha256:b2e8889c66f72c18a61839f0bf1a9f77b5481ba2def044dba33c797b2f2bdcae
Canonical repository
Dao-AILab/flash-attention
Default branch
main
Resolved HEAD
0dc2cb48e484894c48a6cc5c6503fa5cded350c3 ↗
Resolved
2026-09-16
Measurement window
February–September 15, 2026
Download observation JSONDeterministic export · same values as this page

06Peer context

Position among comparable repositories

The cohort uses the same release and measurement window, then narrows to the repository’s ecosystem and removes low-activity observations.

CohortPython · 247 eligible repositories · at least 100 considered commits2026-09-18-v3.0 · February–September 15, 2026 · 145 of 392 ecosystem repositories excluded
Agent-commit share position39th percentile

Midrank percentile; equal shares split the tied positions.

Repository
3.9%
Cohort median
6.3%
Share rank
149 / 247
0th50th100th

Percentile = (repositories below + half of 4 tied repositories) ÷ 247.

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 →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 Dao-AILab/flash-attention

Scan another repository

Try