Measure prompt behavior
First complete the teaching lesson. Use the PocketCards codelab and this source revision for the practical work.
Your task
Use a five-fact public passage for three generations. Add one unsupported question to your review notes, not to the source, and see whether any output invents its answer.
Evidence to produce
Record source facts, generated questions, supported answers, omissions and variation. Test an embedded instruction as a separate case. Keep a small learning log in your own branch. Include the command or action you used, the expected result, the observed result and the reason for any difference. Do not include configuration values, tokens or private source notes.
Before checking the answer
Predict the outcome before you run the app. If the result differs, locate the responsible layer: input validation, ViewModel state, provider adapter, storage or rendering. Fixing the wrong layer can hide a symptom without meeting the requirement.
Review guidance
Different wording is not automatically a failure. Unsupported content is a failure even if the JSON parses. Prompt instructions do not replace app validation.
Explain your decision
Would an exact-string assertion be useful for a live model here? Explain why or why not. Answer with a concrete code reference and one observation from your run. If you can only repeat the terminology, revisit the lesson and demonstrate the behavior before moving on.
Check your work after attempting the exercise
Different wording across three runs can be acceptable when answers remain supported. A unique, well-formed question is still unacceptable if its answer relies on facts absent from the notes. Record coverage and support independently.
If your result differs: There is no single expected generated deck. If your review checks exact wording only, rewrite it around source facts. Treat embedded instructions as a separate case so you can attribute a change in behavior.
Ready to continue: Your learning log contains a prediction, an observed result and a source reference. You can explain the result without copying this guidance. Carry the evidence into the final assessment.