Things I've made for the fun of it.

Every time I wanted to find somewhere to study or co-work with a friend in NYC, I'd end up with three browser tabs open — Yelp, Google Maps, the café's own website — and still not know if it'd actually work for me. One tells you if the coffee is good. One tells you how far away it is. Nobody tells you whether you can set up for three hours, find an outlet, and hear yourself think.

I had a product idea and not enough technical expertise. So I learned by doing — leaning on what I know about how projects run, and using Claude to execute. I started with a product brief, built the visual direction and mockups in Claude Design, then handed off a full design package to Claude Code to build a working prototype. Even as a team of one, I set up a Linear project, wrote tickets, and ran a real feedback loop: backlog grooming, prioritization, dependency mapping. It kept Claude and me aligned on the roadmap and gave the whole thing the structure it needed to move fast without getting messy or losing momentum.

Coding wasn't the hardest part. The hard part was the strategic thinking — being clear and precise enough about what to build that AI could actually build it well. In the end, it wasn't so different from running an end-to-end project with a full team of designers and developers, just a much smaller standup.

The tools were the easy part to pick up — Vercel, Supabase, SQL, OpenStreetMap, a handful of integrations. What took more work was developing an instinct for when to go deep and when to stay at the product level. There were moments where I had to resist the urge to just ask Claude to build something, and instead slow down to think through the implications: how does this data model affect what we can show users? What happens to this feature when the database has 10,000 entries instead of 300? That kind of thinking — anticipating downstream consequences before writing a line of code — is something I'd always relied on designers and engineers to do. This project forced me to learn how to do it myself, building confidence to work more fluidly with future project teams.

A small beta with real users in NYC, using their feedback to sharpen the data and validate whether Outpost is as useful in practice as it is in theory!