AI makes building cheap. It does not make building valuable. Those are two different problems, and most builders only notice the first one.
Give a modern AI tool a clear description and it will produce the feature — the research, the code, the tests, the build. Hand it ten rejections from an app store review and it will diagnose and fix each one. That part of the old job genuinely got faster.
What did not get faster is the earlier, quieter question: what should you build at all?
That is not a prompting problem. It is a domain problem. The AI does not know which signal someone can actually act on, because it has never lived their day. It can generate a hundred charts overnight and never once ask which of them answers the one decision the person is actually stuck on — the yes-or-no that makes their morning easier.
The value was never in the output. The value was in the translation: taking a technical signal and folding it into the single choice a person actually makes.
That translation has a name — domain knowledge. It is what separates an AI that produces from a product that matters.
GainCut’s recovery feature is a clean case study. The underlying signal — heart rate variability — is a solved problem; the papers have been published for years. Rebuilding it with AI was never the win. The win was deciding what the tired person actually needed: not three overlapping waveforms and a chart, but a single go/no-go — green, lift; amber, easy day; red, rest. The AI did the research, the SwiftUI, the tests, the builds, the ten app-store rejections and their fixes. None of that was the value. The value was the translation, and the translation came from having lived the problem — not from the model. (GainCut is real — Calorie Counter on the App Store.)
The point is not “don’t use AI.” It is don’t use AI blindly. The people who get real value out of it are not the ones who delegate the most. They are the ones who walk in already knowing what is worth building, and use the AI to build it faster. The AI multiplies the reach of that judgment. It does not supply it.
For a solo builder, that is the whole strategy in one line: spend your scarce hours getting close enough to a real problem — and the person who has it — that you can name the decision they are stuck on. That part nobody can automate away. Everything after it, an AI can carry.
AI does the work. Domain knowledge decides what work matters.
This post lives at the intersection of a few of the site’s neighbours: What AI Agents Still Can’t Do names the capability ceiling, The Supervisor Class is how you stop writing code and start directing it, and “Done” Is a Claim, Not a Fact is the ritual for trusting the output. This one is the earlier question — whether the thing being built is worth building, and where that judgment comes from.