Speeding up code review with small-change

Imprint lets the build decide which small pull requests an AI skill can approve, so engineers spend review time on the larger and riskier changes.

Will Larson leads engineering at Imprint, a financial technology company that runs credit card programmes. Its engineers now merge four times as many pull requests as two years ago, and the typical review takes 77 minutes instead of 33. His team built a rule, called small-change, that lets an AI skill approve a narrow class of changes without a human reviewer.

• A pull request counts as a small-change when it is under 1,000 lines and has no serious findings from the AI review tool. Its author must also have ten or more merged changes in that service.
• Shared libraries are never eligible, and each repository can exclude paths, such as a service that moves money.
• The review skill has two options only, approve or ignore, and never comments, because Larson found that AI-written review comments slowed teams down.
• About 10 percent of changes qualify so far, and any error from one is treated as a failure of the team's checks, not of the author.

An engineering team can start reducing human review with a narrow rule the build checks, then widen it as its automated tests improve. This is an inference and is not the author's claim.

Derived by Working Surface from the article; more in the Playbook. Source line: it’s uncomfortable to reevaluate whether a long-held best practice has become merely dogma.

18 October 2026: as AI agents make code and prototypes cheap, teams are writing down which work needs a human check and which ideas have earned more effort.