LEARNABLE / MODULE 1 / GUIDED BUILD
Your first dinner conversation loop
Load an example, inspect its fields, and answer a follow-up. Then break something on purpose and repair it. You are learning to explain what the application does.
1. Inspect a sample response
Edit a value to test meaning. Delete a comma to test format. Loading a sample replaces the current practice edits.
2. Read what the app understood
A summary appears only after structural validation succeeds.
Now check meaning: compare every value with the original request. The format check cannot establish that the extraction is faithful.
3. Answer one useful question
Validate the sample first.
For dietary restrictions, enter “none” only when explicitly intended. Budget amounts must be positive numbers; fee answers must be yes or no.
4. Make it yours
- Load the worked case and validate. Enter a fictional delivery area. Confirm the budget stays at 25.
- Change 25 to 250. Validate again. Explain why a format pass is insufficient, then repair the value.
- Load the broken case. Repair the missing comma and validate again.
- Download this starter. Open it in a text editor and search for
nextQuestion. Change the delivery-area question wording. Save and open your copy in a browser. - Return to Product Academy and save your test results and check-in.
The starter is a single HTML file. No installation, credentials, or paid service is required. For help, ask your coding assistant to explain the nextQuestion function before changing it.
Extraction prompt for a later live-model experiment
Task: extract dinner preferences from the supplied request. Rules: preserve explicit requirements and corrections. Do not guess missing details. Use null for unknown values; use [] for dietary restrictions only if the user explicitly says none. Do not invent restaurant availability, prices, currency, or location. Return these JSON fields: cuisine_preferences: list of strings or null spice_preference: string or null budget_amount: positive number or null currency: string or null includes_fees: boolean or null dietary_restrictions: list of strings or null delivery_area: string or null If the request specifies party size or a per-person budget, preserve those separately as party_size and budget_basis. Input: [place the user request here]
Prompting alone is not validation. A live version will need a model connection, server-held credentials, response checks, and measured behavior.