Day 5: Why More AI Replicas Do Not Immediately Fix a Spike Diagram

The practical problem

The autoscaler requests capacity while users are still waiting.

Conditions and side paths

This diagram shows the capacity-control path; requests remain in a bounded admission path.
The ready pool processes admitted work, whose measured demand feeds the controller again.
A warm minimum helps predictable bursts but carries idle cost. A capacity ceiling still needs an overload policy.

One lesson

Requested capacity becomes useful only after startup and readiness.
Proposed teaching design; not a measured deployment or benchmark.