- The cost of producing plausible code has collapsed, but broad claims about dramatic speedups, obsolete practices, or halving headcount are unproven.
- Management practices should be audited based on their underlying assumptions: those resting on code cost need review; those on human coordination, trust, or verification remain valid.
- Evidence of speedups is smaller than noise; honest instrumentation is needed to measure actual acceleration, not felt speed.
- Old metrics like velocity and PR counts become misleading because volume is cheap; focus on business outcomes and treat code volume as a cost.
- Plumbing and mechanical verification (machine-checkable correctness) are fast, but semantic verification requiring human judgment and ownership remains slow and does not compress.
- The junior engineer pipeline is broken because AI absorbs tasks that built judgment; no proven solution exists yet.
- Old rules like 'director should not code' need recalibration; 'shield the team' still valid but requires deliberate context selection; 'consensus before commit' applies only to irreversible decisions.
- Headcount requests should be scrutinized because output per person changed, but coordination costs remain.
- The job is sorting which practices rest on typing costs vs. human nature, not blindly keeping or discarding the old playbook.
- Management's information-routing function is at risk from AI, but judgment, ownership, and signing off on decisions remain human; orgs will shrink to those who own accountability.