All the code review is a stage
Andrew Zigler · 12 November 2026
Read the originalEngineering leaders told Andrew Zigler of Dev Interrupted that AI-written pull requests have outgrown careful review, so they now sort changes by risk before deciding who reads them.
Andrew Zigler writes Dev Interrupted, the publication of LinearB, which sells engineering metrics software. He asked engineering leaders where pull request review had become a formality as AI agents wrote more code, and what replaced it. Their answers describe sorting changes by risk, routing risky ones to a qualified reader and spreading the review load.
- LinearB's figures from more than 8.1 million pull requests show AI-generated pull requests merging at 32.7%, against 84.4% for human-written ones, and waiting 5.3 times longer for a reviewer.
- At Postman, an API platform, tests, documentation edits and internal business logic can merge without a person, while authentication, billing, public APIs and core data model changes still need one.
- At Deel, a payroll company, each AI reviewer's configuration, approved instructions and findings can be traced, and a missed issue is fixed in the reviewer's instructions.
- At Harness, a software delivery company, automatic merging is paused for a class of change after an incident until the check that missed it is fixed.
An engineering team could write down which kinds of change may merge on automated checks alone and which need a named reader. This is an inference and is not the author's claim. The piece is published by LinearB and ends by promoting its product.
Sort pull requests by what they could break, let test and documentation changes merge on passing checks, and send authentication, billing or public API changes to a qualified reviewer.
After a production incident, stop pull requests of the same kind from merging on automated checks alone until the team fixes the check that missed the problem.
Derived by Working Surface from the article; more in the Playbook. Source line: A rubber stamp is a governance policy too.
14 November 2026: as AI agents produce more code and tests than people can inspect, teams are writing down which work gets checked closely, by whom or by what, at what cost, and how they will know it worked.