Choose a Prototype Tool
Asha's biggest worry about Pantry Pal is not the design. It is whether AI-suggested dinners will feel useful to a tired person at 8 p.m. A polished, clickable app would take her days to build. A question like hers might need only an hour.
The rule
Pick the cheapest prototype that answers your riskiest question.
A prototype is a question in disguise. Write the question first, then choose the lowest-fidelity tool that can answer it. The options form a ladder:
| Rung | What it is | Good for answering |
|---|---|---|
| Sketch | Screens drawn on paper or slides | "Do people understand what this does?" |
| Clickable mock-up | Linked screens with fake data, often generated from a prompt | "Can people find their way around?" |
| Hand-run AI | You run a chat AI yourself and pass its answers on | "Is the AI's output actually useful?" |
| App builder | Describe the app in chat and get a working one with stored data | "Does the whole flow work with real use?" |
| Real project | An AI coding assistant working in a codebase you own | "Can this grow into the product?" |
Higher rungs feel more real, but they cost more time, invite polishing, and make you reluctant to throw the thing away. And every prototype should be easy to throw away: it exists to learn, not to launch.
When you compare tools on the same rung, three questions age well: can you export what you made, how does the price grow as you use it more, and can you see and change the code and data it creates?
The shortcut
Question → rung. Name the riskiest question, then climb only as high as that question needs.
"Are the AI's dinner ideas actually useful?"
The riskiest question goes first. Asha pastes five friends' real fridge lists into a chat AI, sends each friend their ideas, and asks which one they would cook. Hand-run AI, and no app at all.
The trap is climbing too early: a beautiful working app built to answer a question a sketch could have answered on Monday.
Drills
Answer fast — trust the rule.
What should you decide before picking a prototype tool?
Asha wants to know whether people understand what Pantry Pal does from its first screen. What is the cheapest fitting prototype?
What does a hand-run AI prototype test best?
Why should prototypes stay easy to throw away?
Remember
- Pick the cheapest prototype that answers your riskiest question.
- Question → rung: sketch, mock-up, hand-run AI, app builder, real project.
- Trap: climbing to a polished working app before the cheap questions are answered.