I build things
and help people use them.

Pen and paper first, always. I think through the pieces myself, and I dictate a lot. Then I take it to a model and tell it to push back. The goal is to have my assumptions challenged. After that it is a mockup, then the UX and UI, and only then the pipes. I work in the terminal.

Small, working, shipped. WedVis went from idea to a live product in about a week, then about six months of tuning it against what real couples and wedding pros actually did inside it. I watch usage and change the prompting behind the product from that.

I do not lock into one platform. I have moved between Gemini, ChatGPT and Claude on purpose, across different interfaces, to learn what each one is good at and what works for me. That comes from playing quarterback under a lot of different coaches. No single coach is completely right. You take the best from each and assemble your own version.

Where AI was the wrong tool

There was a stretch where I looked at every problem and AI was the hammer, whether or not the problem needed a hammer.

Outreach was the clearest case. I tried hard to automate it. The right answer was hand to hand: personal LinkedIn messages, personal Instagram messages, email. Only after doing it by hand and understanding the motion did scaling any of it make sense.

Checklists were the other one. I over-engineered a Notion-style AI checklist. Users wanted a plain Excel-style table. Better for a handful of people, wrong for almost everyone. I shipped the table.

What I got wrong

My worst mistake was building features nobody needed instead of doing outreach. I optimized the signup form too late, while I was paying for traffic. I polished onboarding before I had solved getting anyone to sign up at all. The lesson under all of it: building with AI gives you a dopamine hit that makes you feel more efficient than you are, and it is very easy to keep building without stepping back to ask why.

One habit I brought over from marketing: I build with scale in mind, so the pieces can connect to each other later instead of shipping closed doors. Image generation in WedVis was built so images could flow into other features instead of sitting in a gallery. I will admit that sometimes tips into over-engineering.

My whole life I could not make anything with my hands, and I envied people who could. My father was a military mechanic who could fix anything that did not involve electricity. My grandfather, Poppy, was a machinist who built torque converters by hand. I could not build anything tangible, could not fix anything, could not play an instrument, could not dance. Then AI coding arrived and for the first time I felt the thing I had been missing: producing something real that could affect someone else's life. That, to me, is everything. It is why I love this.

I played quarterback in college. I host AI literacy sessions for young professionals around New York. I write at Dispatches, teach a free course for people who have never opened a terminal, and build things nobody asked for because it keeps the work fun. One of them reminds people to call their mothers.