For years I have watched founders, intrapreneurs, and senior leaders pitch ideas. The patterns are familiar.

Uber for X.

eBay for X.

Computer vision to X.

AI to X.

I have given up trying to guess which ideas are good. I cannot do it reliably. Nobody can. What I have learned to listen for instead is one specific tell: does the person pitching want to know if their idea is right, or do they want me to confirm it?

The test

I ask one question: “What’s the simplest way you can imagine to prove this idea works, before you build it?”

The answer reveals more than the idea itself. If they cannot or will not imagine a simple test, they are not in love with the problem. They are in love with their solution. The solution has a particular shape in their head, and the shape is the part they want to protect. The problem is a pretext.

This matters because love of the solution will defend the solution past the point where the evidence kills it. Love of the problem will let evidence kill ten solutions and go looking for the eleventh.

The Mechanical Turk

In the late 1700s, the world marveled at a machine that could beat most chess players. It was called The Turk. It toured for nearly a century before the secret leaked. Inside the magnificent box was a person with above-average chess skill. The “magic” was a human in a cabinet.

This is the metaphor I keep returning to.

The best innovators I know use Mechanical Turks. They build the smallest possible imitation of their idea, often with people doing the work the eventual technology would do, and they use it to ask one question: do the users actually want this, in this shape, enough to come back for it?

A Mechanical Turk is not a fake-it-until-you-make-it move. It is a fail-fast-and-pivot move. It is a way of converting a belief into a test before spending the budget to convert the belief into a product.

Most ideas do not die because the technology is hard. They die because nobody wanted exactly the thing the founder imagined. The Mechanical Turk asks the cheaper, faster question first.

And now the test costs even less

The Mechanical Turk worked because the person in the cabinet was cheaper than building real chess intelligence. The cheaper the imitation, the lower the cost of asking the real question first.

That economics has shifted again. You can now put a large language model in the cabinet. The “intelligence” is generic, sometimes wrong, occasionally expensive in tokens, but it can be on a screen in front of a real user in an afternoon. What used to require a team and a backlog now requires a prompt, an API key, and a Stripe account.

There is no excuse left for not running the test. If you cannot describe how to put your idea in front of a real user this week, with an AI doing the work that the eventual product would do, you are not failing for lack of options. You are failing for lack of willingness.

What it looks like when someone refuses

There are projects, almost always, where the founder cannot describe a Mechanical Turk for their idea. Sometimes that is because the idea is genuinely technical R&D, where the R is a real unknown. Most of the time it is because the founder doesn’t want to know yet.

That is the tell. The willingness to design a cheap test is proof that you love solving the problem more than you love your solution. The unwillingness is proof of the opposite.

I have watched well-funded teams spend two years building “the technology” that nobody asked for in exactly that shape. Sometimes they ship. Sometimes they don’t. Either way the money is gone and the lesson was available, cheap, eighteen months earlier, in a Mechanical Turk that nobody wanted to build.

Why this is a Change Energy question

The Mechanical Turk is the first link in the conversion chain: belief into test. If a team won’t do the conversion at step one, the rest of the chain doesn’t matter, because they are not really running a test. They are running a defense.

A roadmap is a list of bets, not proof. The Mechanical Turk is the cheapest evidence you can produce that turns a bet into something you can defend or kill. Use it, or be honest that you have no intention of letting evidence change anything.