When to use it
When you have a product or feature idea and the urge to "just start building" — but you should test it cheaply first. The role is product strategist. Result: a breakdown into risky assumptions, a way to test each one WITHOUT writing code, and the single biggest risk to tackle this week.
The prompt (copy and paste)
You are a product strategist who validates first and builds second. Reason step by step, be honestly critical.
IDEA: "<PASTE THE HYPOTHESIS>". WHO IT'S FOR: <segment>. WHAT PAIN IT SOLVES: <pain>.
Break it down:
1. State the hypothesis as "We believe that <who> will <do what>, because <why>".
2. The riskiest assumptions (what has to be true, or the whole thing falls apart) — in descending order of risk.
3. For each assumption, the cheapest and fastest way to test it WITHOUT building anything (interviews, a landing page, pre-orders, a manual concierge test).
4. The "keep going" and "stop" signals.
At the end, the main risk to test first and exactly how to do it in a week.
Filled-in example
Idea: "an app that suggests recipes from a photo of your fridge". Segment: busy city dwellers. Pain: "I don't know what to cook from what I've got".
Expected AI response: the hypothesis in the template; risky assumptions — (1) people are willing to photograph their fridge, (2) recognition is accurate enough for the recipes to be relevant, (3) they'll pay for it; no-code tests — a concierge test (they send a photo to a chat, recipes are written by hand), a landing page with pre-orders, 10 interviews; stop/go signals; the main risk of the week — "will they photograph it at all", tested concierge-style on 20 people.
Variations
- Pre-mortem. "Imagine the project failed a year from now — name 5 likely reasons."
- Experiment map. "Lay it out as experiments: hypothesis → metric → success threshold."
- Devil's advocate. "Tear this idea apart — find why it will NOT take off."
Pro tips
- Force the AI to be critical ("honestly critical", "devil's advocate") — by default the model agrees with you and praises the idea.
- The most valuable part is "test it WITHOUT building": the point of the exercise isn't to build, it's to make being wrong cheap. Don't skip it.
- The AI generates hypotheses and ways to test them, but real people do the testing: the conclusion isn't "the AI said it'll work", it's "here's how I'll check it within a week".