Apenwarr says review layers crush velocity

This is a sharp diagnosis of the current AI dev paradox: code gets generated faster, but delivery still stalls in process queues.

• –The “10x slower per review layer” rule is deliberately provocative, but it captures real wall-clock delays in PR, design, and cross-team handoff workflows.
• –The Deming angle is the core insight: quality improves more sustainably by preventing classes of defects than by adding late-stage inspection passes.
• –Early community reaction is polarized between “shift review left into planning/pairing” and “cutting review too hard invites chaos,” which is exactly the tension this post surfaces.
• –For AI-assisted teams, the practical move is smaller increments, better interfaces, and stronger automated safeguards rather than piling on more human gates.