How it works
Where every claim comes from
There are plenty of tools that will tell you an idea is promising. Almost none of them will show you why. This one is built the other way round.
1. It searches, it does not remember
When you submit an idea, nine sources are queried live:
- Hacker News
- GitHub
- GitHub Issues
- Community forums
- Buying signals
- Stack Exchange
- npm · PyPI · Docker
- Web search
- Dev.to
GitHub Issues and the buying-signal search carry the actual voice of users: complaints and “I’d pay for” statements, quoted rather than summarised. Hacker News gives discussion and, asked once per year, a real trend line. GitHub answers “has someone open-sourced this already”. Stack Exchange shows people stuck on the underlying problem. npm, PyPI and Docker Hub turn “there are alternatives” into a monthly download count, which is usage rather than approval. Eight Discourse community forums add people describing the problem itself rather than reviewing products. Web search fills in products and pricing pages, Dev.to catches write-ups the others miss. Typically 80–120 records come back per idea.
If a source is down or finds nothing relevant, the report says so on the page. Stack Exchange in particular returns nothing for most non-technical businesses, and that is reported rather than padded with loosely-matching questions.
Two extras are computed the same way: mentions per year on Hacker News, and a live registry check on whether the obvious domain names are still free.
2. The model cannot invent a link
Collected records are numbered and handed to the model, which cites them by
index — [3], [7]. The server turns those indices
back into URLs. An index that does not exist is dropped.
That makes “every claim links to its source” a property of the system rather than a promise in a prompt. It is also why a block cannot claim to be evidence-backed if nothing in it cites anything — the server downgrades it automatically.
3. Evidence and estimate are marked differently
Every claim in the block cites a collected record, with a link and a date. Verify it in one click.
No source supports this. Market size, risks and next steps live here. Useful as a starting point, not as a fact.
Roughly half of a typical report is estimate. Showing that honestly is the point — a tool that presented all six sections with equal confidence would be easier to sell and worse to rely on.
4. What it is bad at
The sources are strongest for software and technical businesses. Ask it about a local service business — a flower shop, a gym in a particular city — and it will find very little and tell you so. That is a real limit, not a temporary one.
It also cannot tell you whether you should build something. It measures the outside world, not your fit for it.
5. Limits
Five new analyses per day per visitor. Reading anything already checked is unlimited and always will be. There is no account, no paywall and nothing to buy — the site exists to find out whether it is useful at all.