08 - Going further

Time: 5 minutes. Goal: Explain your result and transfer the method without assuming framework parity.

Give a one-minute handoff

Use your worksheet to say:

On GET /api/insights/eager-orders, run <before> observed <sample count> requests
with <SQL count> SELECTs per request. Action <id>, plan <version>, changed only <method>.
The acceptance test passed and the live changed method executed in run <after>.
The response stayed equal and <sample count> requests now showed <SQL count> SELECTs.
Comparison coverage was <coverage>; we cannot conclude <unmeasured claim>.

Have a partner identify one unsupported claim. A good answer can include “not measured.”

Take the workflow to your application

Start with local activation, masked values, one real route, and its owning code. Read availability before adding optional dependencies. Build a reproducible workload and a real regression test, then use a bounded coding-agent assessment.

Do not copy intentionally unsafe fixtures or expose the developer console to production. Keep diagnostics separate from the application's own authentication/CSRF policy.

MVC supplies the full reference exercise. WebFlux has reactive/SQL timing limits; Quarkus uses native integration points and has no transaction capture. See Appendix C before translating an exercise. The outcome is the same evidence discipline, not identical panel counts.

Pick a take-home lab

Appendix D covers durable history, JFR, optional sensors, services, and application AI. Choose one question to investigate, record the separate action approval, and keep its bounds explicit.

Clean up by stopping the local application and coding-agent session. Keep your worksheet and the exercise branch if useful. Remove only your own generated report/capture files; do not run a destructive repository reset.

Previous: 07 - Agent change loop. Reference: Workshop overview and participant worksheet.