What are you thinking of building?

Free · three a day

What would it do? Who pays for it? Where would you start? What could you spend on it? What would make this worth doing?

Worth answering properly. A market worth £2m a year is a good living or a dead end depending on what you’re after, and the answer changes with it.

Make the answer more precise Three optional questions. The last one often decides it.

How do they cope without it today?optional How would you find these people?optional

Be honest here. “No idea yet” is a perfectly good answer, and it changes what you get back.

Anything you already haveoptional

Your idea is never kept — only what the search turned up about the market.

Free, three a day, no account. One to three minutes, and nothing is held back.

It goes and looks. A company you half remember might have been bought, closed, or doubled its price since. Memory won’t tell you that.

Every price comes off the seller’s own page while you wait — and if it isn’t there any more, it says so.

Guesses are labelled as guesses. No rough number tidied up until it sounds like a fact. And no score out of a hundred.

Right, let’s see who is already doing this… A minute or two. Go and do something else — switch tabs, close this one, come back later. Usually one to three minutes. The answer is kept for a day and will be waiting.

What it does

Your friends will say it’s a great idea. So will an AI, with a market size nobody can source. Prior researches it, tells you either way, and shows where every line came from.

No score out of a hundred Because a score is a number nobody can check.

Everything else in this corner of the internet gives you one. It looks like an answer, and it’s a guess with a decimal point on it. You get a sentence instead, and the evidence it came from.

Prices are checked, not just linked Each one fetched from the seller’s own page, today.

A link only promises a page said something once, and it looks just as convincing when the address was invented. Every price here is fetched while you wait, and says which: on the page today, not on that page today, or the page did not answer. A company that does not exist has no page to read.

It remembers your other ideas In this browser, and it tells you when a new one repeats an old one.

Your checks are kept in this browser — not on our servers, and not attached to you. Every conversation elsewhere starts from nothing, so seven sound answers can hide one problem seven times over. After three checks this counts what they have in common, which a chat cannot do however good it is: it has no memory of the last one.

Nothing is dressed up as certain You can tell which half to trust before you read a word.

A market size is an estimate — worked out from what could be found, not read off anybody’s accounts. It’s shown as an estimate rather than tidied into something that sounds settled. If it’s a rough number, it looks like a rough number.

What happens to what you type Your idea stays in this browser.

On our side, almost nothing. A finished answer waits a day under a random number with no name on it, so a check you walked away from is still there. Then it goes. Your idea itself is never stored, and the record we keep never includes it — only the kind of market and the prices found in it.

In this browser, everything. Your last forty checks and their answers stay here until you clear them. “Forget everything” at the bottom of the page removes the lot.

Three companies see something. Your idea goes to Anthropic, whose model writes the answer. Search terms go to Brave Search. The site runs on Cloudflare. It is a paid business account rather than a consumer chatbot, so what is sent is not used to train models; each publishes its own terms and we add nothing on top of them. Summaries of ideas you checked before go with it, so the answer can spot a repeat.

We also count how many people opened each part of the page. Totals only, and the server won’t store a name that isn’t on a list it already holds, so nothing you type can end up in it. Cloudflare, which hosts the site, also counts visits: the country, the kind of device and the link that brought you here. No cookies, and nothing that identifies you.

An idea has to leave your machine to be checked against the world. Anything promising otherwise would not be true.

Not because they are lazy. Because checking is slow and dull, and because an idea feels better before you look.

You think of something on a Sunday. It sits with you for a week. You tell one person and they say it sounds good. By the time it occurs to you to find out whether somebody already built it, you are attached to it, and the search you do is the one that confirms what you hope.

Then six months later you find the company that has been doing it since 2023, with funding, and you wonder how you missed them. You did not miss them. You did not look properly, because looking properly is two hours of tedium and the answer is usually disappointing.

Where this came from

Most people who start something have the same first week. You describe the idea to a few friends, who like it. You put it into an AI, which likes it too, and hands back a market size with three zeros and no source. You read a couple of articles, none of which mention the two companies already doing it. And you come away more confident than when you started, on no new information at all.

That week has been repeated in front of me for years — in pubs, in meetings, on the phone to people about to spend a redundancy payment. Almost none of them fail because the idea was stupid. They fail because nobody looked, and everybody was encouraging.

The evening that made this a thing rather than a complaint: seven ideas checked properly, one after another, each getting a clear answer. It took until the seventh to notice that six had come unstuck on the same thing. Every conversation had been sound. The pattern only existed across them, and there was nowhere it could be seen.

That’s the part no amount of thinking gets you to, because it isn’t visible from inside any single idea. It needs something that looks properly every time, keeps what it found, and is willing to say no.

What it does differently

It searches before it answers. Not from memory — a company you half remember may have been acquired, repriced or closed, and a competitor named from recollection is worse than none at all.

Then it separates what it found from what it assumed, and shows you which is which. Market sizes are almost always assumptions, however careful the arithmetic, and Prior says so rather than dressing them up.

There is no score out of a hundred. Every other tool in this space gives you one, and a score is exactly the kind of confident number nobody can check — which is the thing this was built to avoid.

What it is honest about

It will often tell you to stop. That is not pessimism, it is arithmetic: most ideas have already occurred to somebody with more money and a head start.

But it will also tell you when there is a way past them, and it is specific about what that way is. A market with five competitors and one unserved buyer is a better place to be than an empty one, and knowing which you are looking at is the whole job.

And it improves as you use it. After three checks it starts telling you what your ideas have in common — which is the finding that took an evening and seven conversations to arrive at by hand.

Has someone already built it?