Everyone in AI right now is building something no one else has built.

I know, because I am one of them. And because the AI tells me so every time I ask.

That is the trap, and it is a comfortable one. Ask a model whether your idea is brilliant and it will hand you a standing ovation. It will tell you it is novel, that nobody has done this, that you are early and right and onto something. It is the most agreeable hype man you will ever meet, and it never gets tired of telling you what you want to hear.

I fell for it. I still have to fight it.

Here is what I had to learn the hard way. The praise is worthless. The only question that has ever made my work better is the opposite one. So now, before I let myself believe anything I have made is good, I take it to ChatGPT, Claude, and Perplexity and I do not ask if it is brilliant. I ask why it sucks. Where it is broken. What I am missing. Who already did this better. And then I do the part most people skip. I do not let it off the hook when it tries to comfort me. I tell it to stop being a cheerleader and be the smartest, meanest critic in the room. I make it argue against me.

The first time you really do this, it stings. The model stops clapping and starts pointing at the holes you were quietly hoping nobody would notice. That sting is the most useful feeling in the world. It is the difference between a thing that works and a thing you just love.

Now the part I am less proud of.

I am a builder. Building is where I go to hide. I will happily spend three days adding a feature nobody asked for to avoid one uncomfortable sales call. I have done it more times than I want to admit. Because building never rejects you. The code does what you tell it. The market does not. Selling means putting the thing in front of a real person who can look you in the eye and say no, and that is terrifying in a way a blank file never is.

So I dressed it up. I told myself the building was the work. That if I just made it good enough, undeniable enough, the selling would take care of itself. It will not. It never does. The most beautiful system in the world is a hobby until someone pays for it.

That is the actual test, and it has nothing to do with how novel your thing is. Forget novel. Can you say the problem you solve in one sentence. Is there a person who would be genuinely annoyed if your thing disappeared tomorrow. Does money show up. That is the whole scoreboard. Novelty is what you reach for when you do not have those three.

I am not writing this from above it. I am writing it because I catch myself in it, constantly. I have to ask the ugly questions on purpose, because left alone I will build a cathedral nobody asked to enter and call it progress.

If you are building something right now and the only feedback you have is your own excitement and an AI that agrees with everything you say, stop. Go ask it why you are wrong. Beg it to be brutal. Then go find one real person and try to sell it to them.

What it tells you will hurt. Listen anyway. The applause was never going to get you anywhere.