Three problems AI didn't solve
You’re building with AI every day. You’ve shipped more in the last year than the previous five. And you’ve probably noticed: the bottleneck moved. It’s not “can I build this?” anymore. It’s “does anyone want this?”, “can I sustain this?”, and “do I actually know what I’m doing?” These three essays map the process underneath those questions.
They share a single engine I call Genius — a four-phase loop (Current, Desired, Actions, Results) that I’ve tested against twenty years of building, including the last eighteen months of building with AI. Each essay takes one pass through a different problem. They stand alone, but they make more sense together.
Why you can use AI for everything and still not know what you're doing.
You're asking AI to analyze, summarize, and draft. The answers come back fast and plausible. But you don't actually know more than before you asked. This essay maps the kind of knowing no model produces.
ReadYou shipped ten things this year. How many do people actually use?
You can ship a SaaS in a weekend. But 70% of solopreneurs still make under $1K/month. This essay is about the gap between shipping fast and making something people actually use.
ReadAfter the reckoning: what's actually worth building now.
The vibe coding arc ran its course. If you survived it and you're still here, the question isn't 'can I build?' anymore. It's 'can I build something that lasts?' This essay maps what that requires.
ReadI publish slowly and selectively — one essay when I have something worth saying. Subscribers get each piece as it lands, plus occasional notes on what I’m working through between essays.