Code review, specs, and TDD reborn

What happens after coding is solved? | Fiona Fung (Claude Code & Cowork)

Topic 3 of 20

Published Jun 21, 2026 · Clip 14:41–16:59 (2:17)

Lenny asks how the team keeps code review from bottlenecking 8x throughput. Fiona notes Claude Code Reviews didn't exist a year ago and human reviewers were a major bottleneck; humans still review areas requiring deep subject-matter expertise. The key move is automating a framework for 'what good looks like' - checking specs, scales, and content-design criteria into the repo and keeping them updated - so Claude Code Reviews can validate code against them. Lenny compares this to the evolution of test-driven development, and Fiona recounts her first Claude Code bug fix done test-first, noting Claude removes the 'eat the broccoli first' tax that made TDD painful.