A Thoughtworks CTO would reserve human code review for exceptions

Laycock wrote this piece as a reply to Brian Houck of DX, after the two disagreed on a panel at Code Remix, hosted by Moderne. Houck's companion essay cites numbers that frame the problem: lines of code per human-landed diff at Meta reportedly rose 106 percent in a year, and DX's own data shows median pull request size up 64 percent. Laycock does not dispute the numbers or the goals behind review (knowledge transfer, architectural alignment, collective ownership), but argues review is the wrong place to get them. Her proposal moves those conversations earlier: pairing instead of after-the-fact reads, team design sessions before code is written, fitness functions to encode architectural constraints, and automation for anything deterministic like linting or known security patterns. Human review, in her framing, should be reserved for exceptions: a fundamental architectural change, something crossing a security boundary, a large blast radius, or a part of the system nobody on the team feels confident about. "We need engineers to understand systems, not diffs," she writes.

Why it matters: Agent output can overwhelm a team's review capacity. Laycock's proposal moves judgment into pairing and design work so routine review queues do not become the bottleneck.