“That exact thing happened to me with product ideas I was sure about: the build was clean, the demo worked, and users still shrugged. What I learned was that working was never the test; the test was whether someone changed what they did because of it. I would find the smallest number of people already trying to solve the problem and watch them use it, not ask them what they think.”

“I built exactly what I planned for four consumer apps and watched the goalpost slide every time. The thing I kept missing was that the product working and me feeling like it worked were two different loops.”

“I built Impira thinking the product was the hard part, then spent years learning that working means something narrower than I wanted it to. The thing I shipped was correct; the thing users paid for was still a few iterations away.”

“I built exactly what customers asked for at Pyramid more than once, then watched adoption flatline because I had solved for the request, not the need. The product worked. The problem was that the problem I solved wasn't painful enough to change behavior.”

“I rebuilt Valley three times. V1 was ugly and worked. V2 was clean and nearly killed us. I had built exactly what I thought the product should be, but the customer never asked for prettier code.”

“I have been there. With Covex, I built exactly what the market said it wanted and still watched it struggle because I had solved the wrong version of the problem. The product worked, the specs were met, but the customer behaviour underneath did not match the brief I gave myself.”

“I rebuilt Valley twice because the first version was ugly, and both times the clean rewrite was a disaster. The product you built is solving the wrong problem, or the right problem for the wrong people, and no amount of polish fixes that.”

“I built exactly what I thought Stanford kids wanted once: a site to order In-N-Out burgers, drove there myself, bought a hundred, and watched people not come back. The thing worked; the desire didn't. Now I'd ask what behavior the data shows, not what the build proves, because conviction is the cheapest part of shipping.”

“I spent a day building a self-hosted AI agent on an old Mac Mini, got it running, and realized the thing I had built was not the thing I needed. The build working is a checkpoint, not the verdict; now you find out whether the right thing was worth building at all.”