A founder spent a year building a tool that tells you whether your idea is good. It launched to silence. That silence is the most useful data point of the week.
Why do founders skip validation even when it is free?
Because a verdict is a dead end and building is a direction.
A founder wrote about this recently. He built something that took your idea, checked it against real market data instead of vibes, and told you honestly whether it was worth building. Nobody used it. His conclusion was that people do not want to be told their idea might be bad.
He is half right. People do not want to be told. What they want is to move.
Every product that ends in a score has the same shape. You put in your idea, you get back a number and a paragraph, and then you are exactly where you started except slightly more anxious. There is no next step inside that experience. So founders close the tab and go build, which at least feels like progress.
Is an AI score the same as validation?
No. An AI score is a well-organised summary of things other people have already written about markets that already exist.
That has value. It is not nothing. But the model has never had to decide whether to hand you money, and it never will. It has no budget, no boss, no competing priority, no vendor it already pays for something similar. It cannot tell you that your pricing is fine but your onboarding is the reason it churned, because it has never onboarded.
A smarter model does not fix this. More intelligence applied to secondhand text produces a more articulate guess.
What is the difference between a verdict and a signal?
A verdict is a judgement about your idea. A signal is a fact about a person.
Verdicts sound like: this scores 7 out of 10, the market is crowded, similar products exist.
Signals sound like: this specific operations manager is currently paying a consultant four hundred pounds a month to do this by hand. This designer downloaded it, used it for nine days, then stopped, and when asked said the export was useless. Three people asked about pricing before you had pricing.
You can act on a signal tomorrow. You cannot act on a verdict at all, except by ignoring it.
Why does stated interest keep fooling smart people?
Because agreeing is free and paying is not.
Someone telling you the problem is real costs them nothing. It makes them feel helpful. It makes the conversation pleasant. It is also completely compatible with them never giving you a penny, because you never asked them for one.
This is the trap in most early research. Founders come back from ten conversations with ten confirmations and read that as demand. What they collected was ten people who did not want to be rude.
The population that confirms a problem and the population that pays to solve it overlap far less than anyone expects. Sometimes they barely touch.
What does real demand actually look like?
It looks like inconvenience. Somebody already pays for a worse version. Somebody has built a spreadsheet to work around it. Somebody asks what it costs before you bring up cost. Somebody comes back a second time without being reminded.
One founder posted numbers that make the point better than any argument. His app converted 74 percent of the people who hit the paywall. Extraordinary number. Then 70 percent cancelled inside a month, 60 percent barely used the product, and 77 percent never created a single task. He wrote that he had not been selling a habit tool, he had been selling the moment of deciding to change.
A high conversion rate behind a hard wall measures the wall. Retention measures the want.
What should you do with an idea you are excited about?
Do not put it through a machine that will tell you to stop. That is a fight you will win against yourself, every time, and you will learn nothing.
Put it in front of people who fit the profile, who have no relationship with you, and who have nothing to gain from being encouraging. Then watch what they do rather than what they say. The number of people who say yes is a vanity metric. The number who change their behaviour is the business.
The founders who survive are not the ones who validated hardest. They are the ones who found out early who was lying to them, including themselves.
Key takeaways
- Validation gets skipped because verdicts have no next step, not because founders are reckless.
- An AI score summarises what has been written about existing markets. It has never had to pay you.
- A verdict is a judgement about your idea. A signal is a fact about a specific person.
- Confirmation of a problem is free. That is exactly why it is unreliable.
- High conversion behind a hard paywall measures the paywall. Retention measures the want.
- Watch behaviour, not agreement. What people do is the only honest answer.
FAQ
Is it possible to validate an idea without building anything?
Yes, but only if you are measuring behaviour rather than opinion. The question is not whether a product exists. It is whether a specific person with a real budget changes what they do because of it.
Why do AI idea validators keep launching if nobody uses them?
Because they are cheap to build and the promise is appealing. The product that scores your idea is easy. The product that gets a stranger with the right profile to react honestly is hard, which is why so few people build it.
How many conversations are enough?
The count matters less than who was in them. Sixty-four conversations with people who like you is worth less than five with people who have no reason to be kind and a real budget for the problem.
My friends and family love the idea. Does that count for anything?
It counts as affection. Your friends are answering a different question from the one you asked. They are answering whether they support you, and the answer is yes, and it tells you nothing about the market.
What is the single fastest honest signal?
Someone asking what it costs before you raise price. It means they have already decided the problem is worth money, and are now deciding whether it is worth yours.
Building has never been cheaper. Being wrong costs exactly what it always did.