We sell "validate before you build." The only thing I've opened every day this week is the thing I never validated.

Our whole pitch is that you should check an idea against real data before you build it. I've written that sentence more times than I can count. Two of our launches were built on it.

Last week I built something with zero research. No competitors checked, no reviews read, no market anything. I didn't even ask a second person if they wanted it.

What it is: I work from home and I kept missing the sunset. The good light happens while I'm looking at a screen, and by the time I look up it's the blue part, already over. So I made a thing that shows today's sunset time and slowly turns the screen that color as it gets closer, so I notice before instead of after.

That's it. That's the whole product. One screen.

The numbers, since that's the part nobody publishes:

one sentence in, and the structure came back first 3 paths, 9 screens, 39 components
palette pulled from 253 styles, 6 recommended for this structure, I took one
0 lines of code
time spent: [X] hours, during what should have been work
market research done beforehand: none
people other than me who have asked for this: zero

Here's the part I keep chewing on.

Everything we've carefully validated over the past year the analysis engine, the wireframe output, the structure extractor is sitting there being used politely and then forgotten. Signups don't convert. I've written about that here more than once.

The unvalidated impulse thing is the only one I've opened every single day since I made it.

I know the obvious rebuttal, because it's mine: n=1 isn't a market. A tool with one user is a preference, not a business, and me using it daily proves nothing about anyone else. Fine. But I don't think that fully closes it, and here's why.

Validation made sense when building was the expensive part. You spent two weeks researching because you were about to spend three months building. The research was cheap insurance on an expensive bet.

That ratio has flipped and I don't think our advice has caught up. If the build is an afternoon, then the research costs more than the thing it's protecting. At that point "validate first" isn't caution, it's just a slower way to not build something.

Which is awkward, because we sell the caution.

So the question I actually can't answer: where's the line now? At what build cost does validation start paying for itself again? A week? A month? Or is it not about time at all is it about whether you're the user?

And for anyone who's shipped something with no research whatsoever, purely because you wanted it: did it stay a thing only you used, or did it turn out other people had the same problem and you just couldn't have found them by researching?

I'd rather hear "it stayed n=1" than nothing.