The hidden cost on your engineering roadmap this quarter isn't the work in flight. It's the pull requests sitting idle, waiting on a reviewer, piling up faster than humans can read them. AI assistants are drafting more code than ever, and most of it lands in the same queue that was already understaffed a year ago. The queue is the bottleneck now, and almost nobody is scheduling against it, measuring it, or staffing it the way they staff feature work.
The assumptions that got teams here are worth pulling apart one at a time. For a longer argument about why the writing inside review matters more than ever, see Colombo Telegraph on code review in the age of AI-written code.
Myth: Review Is Overhead, Not Engineering Work
Logging review time as "the cost of doing business" was easier to defend when a senior engineer burned a few hours a week on it. That isn't the shape of the work anymore. When an agent drafts the first pass, the human reading the diff is doing the judgment work the author used to do: weighing trade-offs, catching the subtle wrong-but-plausible choices, deciding whether the change belongs in the codebase at all. Google's own engineering practices treat review as a first-class part of writing software rather than a tax on it, and that framing is worth borrowing.
