AI tools let developers write code faster, but shipping speed hasn't improved proportionally. The bottleneck has shifted to code review, not code writing. Applying the Theory of Constraints (and Amdahl's Law), the argument is that speeding up non-bottleneck steps (code generation) yields little gain — you must fix the bottleneck (review). Code review isn't primarily about catching bugs; it's a socio-technical practice that spreads knowledge, enforces standards, and shapes long-term codebase health. The post outlines what good code review looks like, argues against fully automating it (since mentorship and taste-building are personal), and recommends subordinating the whole team's workflow to clearing the review queue rather than generating more code.
