Building in the open

On Ramblings of a Designer I talked about side hustles, why people are scared of AI, drawing it on paper first, and what I would do with a hundred dollars if I were starting today.

More of us are building after work with an AI writing the code. It will tell you it is finished. Here is the ten minute check that tells you whether that is true.

I set my price by looking at the wrong shelf. The number felt right, and it was aimed at a customer I do not actually have.

Everyone assumed one thing was true, including me, including the AI. I was a sentence away from promising it to customers. Then I read the actual documentation.

A green checkmark is not a working feature. The habit that separates people who ship reliable things from people who ship surprises.

Twelve people said yes. Five actually installed. The real story of a launch is the quiet funnel between "I would love to try that" and "I opened it."

Localization and white-labeling force you to stop designing a screen and start designing the conditions under which one gets assembled. That is a rules engine, not a catalog of parts.

The bottleneck was never the code. It was the judgment. Start before you feel ready, direct the tools instead of deferring to them, and tell the true story rather than the impressive one.

Handing your work to a reviewer you will never speak to. Two rejections taught me that some review flags are not bugs you patch, they are decisions about what the product actually is.

A passing build proves nothing. The thing writing the code cannot be the thing approving it, and everything the AI got confidently wrong looked exactly like progress.

None of these rules are hidden. They just do not announce themselves while you are designing, and by the time they matter you have already built the thing they forbid.

Two days before launch a rule killed the product I had built. The pivot took an afternoon instead of a month, and only because I had written down why I made each decision.