Kristian Tanninen

Kristian Tanninen

Product CTO

Chat with AI
Kristian Tanninen

Observation·

The MVP got cheap. Proof didn't.

AI made the MVP cheap. Founders spend the saving on looks, and proof still costs what it always did.

An MVP takes a weekend now. Anyone can build one with an AI tool. So the bar moved: a rough MVP gets called AI slop, and founders answer with polish.

Build three rough versions, not one polished one.

Our data shows where the money goes. Of more than a thousand requests to build an MVP, one in six asks for a polished design. Only one in nine asks for a way to test demand or measure use. And the request that follows is growing faster than MVP requests overall: make the MVP hold up for real users.

Polish does not answer the question an MVP exists for. A buyer looks at a beautiful demo and says it looks great. That is a compliment, not a decision. Weeks of design later, nobody has committed to anything.

The cheap part is the advantage. Use it.

Build three rough versions instead of one polished one. Each takes days, not months.

Show them to real buyers this week, on their own data. A buyer reacts to their own numbers in a way they never react to a demo account.

Ask each buyer to commit: a paid pilot, a signed letter, a start date. Keep the version someone commits to and drop the rest.

The MVP got cheap. Proof didn't. A compliment is not proof. A commitment is.

Reel transcript · 0:51

An MVP takes a weekend now. Anyone can build one with AI. So founders spend the money on looks instead. That's the wrong place to spend it.

Next request: make it hold.

Our data holds over a thousand MVP requests. One in six asks for a polished design. Only one in nine asks how to measure if anyone uses it. And the next request is growing fast: fix the MVP so real users can rely on it.

Three rough versions, not one polished.

So founders polish the cheap part and skip the proof. Turn it around. Use the cheap part to get proof fast.

Build three rough versions, not one polished one. Show them to real buyers this week, on their own data. Then ask each buyer to commit: a paid pilot, a signed letter, a start date.

The takeaway

Keep the version someone commits to. Drop the rest. A compliment is not proof. A commitment is.