I Scoped This as a Dropdown. It Was Four Services and a Streaming Rewrite.
Adding a second agent backend exposed direct callers, different CLI event shapes, and missing message persistence. The historical actual-phone re-entry check remained open.
Adding a second agent backend exposed direct callers, different CLI event shapes, and missing message persistence. The historical actual-phone re-entry check remained open.
One tester exposed keyboard-focus and feature-discovery problems I had missed. A later correction separates fetch errors from unverified root causes.
Seven successful units reported an average $0.659 API-price estimate. No metered bill was charged. The logs show orchestration volume, not a controlled savings comparison.
My first exception concerned multi-file refactors. GPU contention complicated the local-coding policy later. The chronology matters when explaining why the rule changed.
A cloud model wrote 1,900 lines in one session and burned my five-hour limit — then the next session remembered none of it. Three layers of context files plus hooks, and what each one is actually for.