The trap of vibe coding

I wasted a week.

Not in the dramatic, startup way. Just quietly, steadily, in the way where you feel busy the entire time and then sit down on a Sunday and go "wait, why did I build this."

Here is what happened. I had an idea. The models are good enough now that within a day I had something running. Within three days I had something that looked like a product. I kept going because the momentum felt real, and momentum is its own kind of drug when you are building solo. By the end of the week I did what I should have done before I opened my editor: basic first-principles thinking. Does this problem actually exist at the scale I need it to? Who is paying for it today? Why would they switch?

The answers were not good.

And the frustrating part is that I knew this. Not the specific answers, but the process. I have read enough, thought enough, been around enough early-stage stuff to know that you start with the problem and then find the tech, not the other way around. That was just the rule. For a long time it was the rule because it had to be. If building something took six months, you could not afford to be wrong about the premise. So people were careful. They did the uncomfortable thing of talking to people, sitting with a problem, making sure it was real before they committed.

Now building takes a week. And so somehow I convinced myself that the discipline could wait.

It cannot. The discipline is still the whole thing. The speed just makes it easier to skip it without realizing you have.

I think there is something specifically weird about this moment where the capability of the tools has started to feel like a reason to build. Like, I can build X, therefore there must be someone who needs X. That logic has never worked. It did not work in the early days of mobile apps, it did not work in the no-code wave, and it is not working now for me personally in the AI wave. The technology being impressive is not the same as the technology being needed.

Anyway, early days. I am not being hard on myself about it, just honest. The correction is pretty simple in principle: problem first, always, and then use the speed to get to an MVP that actually has a target. That part I am genuinely excited about. When you do know the problem is real, being able to get to something testable in days rather than months is still kind of unbelievable. That has changed something fundamental about what a solo founder can do.

I just need to make sure I earn that speed rather than borrowing against it.