The Idea Grill

Five hostile experts interrogate your idea and score it out of 100 before you build

SkillNo-CodeClaudeChatGPT

The problem

The most expensive mistake in small business is building something nobody asked for, and it usually starts the same way: an idea that only ever gets discussed with people who are nice to you. Your spouse says it sounds great. Your friend says they'd buy it. Then you spend three months and real money finding out strangers disagree. What's missing isn't feedback. It's interrogation, from someone with no stake in your feelings.

The tool

Once installed, this skill turns your AI assistant into a five-person grill panel: the unconvinced customer, the bookkeeper, the growth skeptic, your strongest competitor, and the operator who has to run the thing on a bad week. Each one interrogates your idea one question at a time, with no questionnaires and no letting you off with a vague answer, then scores their dimension. You walk out with a weighted score out of 100, the unknowns named, and the three cheapest tests to run before spending a dollar.

Two design choices matter here. "I don't know" is a legal answer: it caps the score and becomes a named finding, because unknowns are exactly what you want surfaced before you build. And the score only moves on new evidence. Re-run the grill after your tests, and telling the same story more persuasively gets you the same number. It also says the honest thing out loud, every time: the score measures your evidence, not your market. A 90 is still a hypothesis until a stranger pays.

How to use it

1Click "Install skill" on this page and copy the install prompt.
2Paste it into Claude or ChatGPT and answer the five setup questions honestly, including what you'd charge. A guess is fine.
3Let each panelist finish their questions before defending your idea. Answer "I don't know" when you don't; that's data.
4Take the scorecard's three cheapest tests and actually run them this week. Each is designed to cost under $100.
5Come back with results and say "re-grill me": only the affected panelists re-run, and the score moves on evidence only.

Example

A bookkeeper wants to launch a $99/month "quarterly tax-readiness check" for freelancers. The Unconvinced Customer asks when a freelancer last felt tax pain. Answer: April. Follow-up: so why would they pay in August? No good answer. Demand: 4/10. The Bookkeeper panelist finds solid margins (9/10). The Growth Skeptic asks where the first ten customers come from; "Instagram" doesn't survive the follow-up. Distribution: 3/10, with an unknown named: no channel where freelancers feel tax pain year-round. Total: 51/100, real idea with named holes. Cheapest test: email 20 past clients in September offering the check at $79, count who bites. That one test would move Demand and Distribution both.

Pro tip

Run the grill twice: once on your idea, once on the version of it you'd build if you had half the budget. When the smaller version scores higher, and it often does, that gap is telling you what the idea actually is.

#validation #strategy #new-business