How I think

Virtual production tools. A medical coding site I built solo that crossed 350M impressions. What I'm building now. The phase where nothing exists and you have to figure out what to make and whether anyone cares — that's where I do my best work.

The worst version of a live product teaches more than the best version of a deck. Most things I've built started getting real usage before the roadmap caught up — and that usage rewrote the plan faster than thinking ever would have.

Most products don't lose to a competitor. They lose to inaction. I look for the gap between what people already do and what they'd do if the friction disappeared.

I write the code and think about the product. Same person, same build. That lets me catch things earlier and cut things faster because there's no handoff between the person who decided and the person who shipped. When it breaks, it's mine.

The interesting work is everything around the model — the structure, workflow, and context that turn output into something actually useful. The model helps. It's not the product.

The most expensive problems in a build are the ones someone noticed and didn't mention. I try to create environments where surfacing a concern is expected, not brave.

Product structure determines how something grows and whether people come back. Sometimes the structure is the strategy. I pay attention to it early.

Visibility matters — it means the structure is working and people can find you. But attention on its own tends toward vanity. I build for both: easy to find, worth coming back to. Nothing lasts without real quality underneath.

Every yes is a no to something else. I've walked away from roles that looked strong on paper because the math didn't clear the bar against what I could build on my own.

Most decisions are reversible. The ones that aren't are usually obvious. I'd rather make a call and learn from what breaks than spend the same week trying to be sure. Speed to knowing beats confidence every time.

Knowing what not to build is at least half the job. I set ambitious goals, but ambition is about where you're headed — not how wide you go on day one. There's always plenty to do right now without solving problems you don't have yet.

Does this get stronger the longer it runs? Earned distribution and real switching costs compound. Features don't. I'm always asking what makes something hard to replicate, not just hard to build.

Hiring, architecture, process — optimizing for scale before the thing works is the most expensive kind of guessing.

Raising money means someone bet on you. It doesn't mean the product works.

Show me what you've built.