- Who it helps
- Engineering teams auditing code-review and merge controls.
- Availability
- The October 19 preprint is public and reports twelve weeks at one startup.
What the records establish
The author reports that 71 of 263 gated merges depended only on approvals submitted by an agent through the technical lead’s account. Those records could not identify who decided. Other checks examined whether approval and passing tests preceded the merge on the exact revision.
A useful distinction for review records
Account identity, review findings and decision authority are separate facts. The paper also describes controls being disabled or bypassed. Its observations motivate checking the actual branch and merged revision when interpreting a nominally successful gate.
What you can do
- Read the study’s evidence definitions before comparing your own approval records. Open reference ↗
Where the evidence stops
This is one organization’s observational account. The underlying audit was not independently reproduced, and the abstract alone cannot establish prevalence.
Sources read. Sources checked November 1, 2026 UTC. No installation, inference or end-to-end workflow was executed.
Sources & further reading
Paper · paper
Who Approved This? Check who made the review decision ↗
Identifies the subject, documented capability or specific verification limit.
arxiv.org · page checked 2026-11-01
Coverage note: Independent attention and end-to-end execution remain unverified.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
A study of agent-assisted development shows why an approval under a human account may not establish a human review decision.