Your agents should ask the factory first

Andrew Zigler of LinearB, an engineering-metrics company, proposes that each AI coding agent check how long the team's pull requests wait for review before starting new work.

Andrew Zigler hosts Dev Interrupted, a newsletter and podcast for engineering leaders, and works at LinearB, which makes software that measures and automates engineering work. He argues that AI coding agents, each tuned to finish a task and open a pull request, all add to the same queue for human review. He proposes a short set of instructions that each AI agent runs first, asking an AI model whether to request reviews or start the task.

• LinearB's 2026 benchmarks report found that pull requests from AI agents waited 5.25 times longer for a first review than unassisted ones.
• The same report found that 32.7% of AI-generated pull requests were merged, against 84.4% of those written by people.
• The proposed instructions compare this week's wait for a first review with the median of the previous eight weeks, and list the engineer's own waiting pull requests.
• An AI model then answers one fixed question, whether to request reviews first or start the task, and the engineer approves every review request before it is sent.

This is a vendor's account: the proposal is built on LinearB's own benchmark data, programming interface and review automation. Tracking how long AI agents' pull requests wait for review, separately from people's, is an inference and is not the author's claim.

Derived by Working Surface from the article; more in the Playbook. Source line: That’s how a factory manufactures its own bottleneck, one well-behaved session at a time.

17 October 2026: AI output that looks finished at first glance still needs a named person who checks it against something real, and teams now plan their work around that check.