Map your idea,
or find out the hard way.
Describe an idea. PremiseCheck breaks it into the claims it rests on, scores each one 0–100, and flags the ones that would sink it. Then research agents check the shaky claims against the web.
- 0–100 score per claim
- load-bearing flags
- web research agents
- what-if assumptions
- stress-test checklists
- read-only share links
Profitable within 12 months
Enough households pay $75/week
Agent checked 12 sources — comparable box services churn hard after week 6.
Weekly cadence builds the habit
Affluent, health-conscious local market
A real survey: the conclusion on top, the claims holding it up below. One of them is about to sink the whole thing — better to know now.
How it works
Describe the idea
A paragraph is enough. The map assembles in front of you, claim by claim — premises, hidden assumptions, evidence.
Read the structure
Every claim is scored 0–100 and the ones that would sink the conclusion are flagged. The weakest load-bearing claim is called out first.
Send the agents
Each agent researches one claim against the web — for and against, primary sources first — then revises its score. Challenge or redirect it in chat.
Add what you know
Deals signed, numbers you have, pilots you ran. First-hand context re-scores the claim, and the overall verdict tracks every change.
What you get
Success maps
For a business idea: every condition that must hold for it to work, including the ones you haven't thought about.
Steelman mode
For an existing idea: the strongest honest case for it — and exactly where that case is thin.
Adversarial research
Agents try to break claims, not confirm them. Verdicts cite sources you can open.
What-if assumptions
Flip any claim to "holds" or "fails" and watch what it does to the conclusion.
Stress tests
A concrete checklist per claim — an experiment, a dataset, a person to ask. Tick them off as you go.
Share links
Send anyone a read-only map. They see the structure; only you can change it.
FAQ
What kinds of ideas work?
Business ideas, strategies, investment theses, policies, beliefs — anything with a conclusion resting on claims. If it can be argued, it can be surveyed.
Where do the scores come from?
Each claim gets the model's honest probability that it's true, spread deliberately — a map where everything sits at 70 hasn't been examined. Research and your own context revise the scores, and the overall verdict re-scores whenever a claim moves.
What do the research agents actually do?
Each one takes a single claim, searches the web for the strongest evidence on both sides, and files a report: verdict, evidence for and against, sources, revised score. You can question its reading or point it somewhere new — it remembers everything it found.
Can I correct the map?
Yes — reword claims, re-score them, change what supports what, add claims the model missed, delete ones that don't belong. The verdict re-scores after every structural change.
Who can see my surveys?
Only you. Surveys are private to your account. If you want someone to see one, you create a revocable read-only link.
Is it free?
Yes, while in beta. Each account gets daily limits (3 surveys and 8 research runs a day) so the research agents stay affordable to run — they reset every day.
Find the weak points before reality does.
Map your ideaFree to try — creating an account takes a minute.