Built on the side · with AI

I built a real app in the margins of a full life.

This was never a full-time project. I run a school, have one wife and two kids, play pickleball and hit the gym most days, produce podcasts, read, take meetings and drinks, and pretend to have a full social life. The app got built in whatever hours were left over.

104
calendar days
329
features
1
person, part-time

It's a training platform for youth soccer. Kids get a phone app with drill videos and log their practice. Coaches get a website to hand out workouts and see who's doing them. Parents get a report on their child.

A phone app, a full website, a secure database, and an automated testing system behind all of it is normally the work of a small product team. This was one person, part-time, in about three months.

why the boring part matters

Think of putting up a building. Anyone can frame the walls quickly. What makes it a place people can actually live in is the wiring behind them, the plumbing, a lock on every door, and an inspector who walks every room and signs off before anyone moves in. Skip that and you get something that photographs well and fails the first cold night.

Software works the same way. The demo everyone claps at is the easy part. The real work is making sure it doesn't quietly break when a family is depending on it. Here is what that work looked like.

7,300+
automated checks. Little programs that inspect the app every time I change something and flag what broke before anyone else sees it. Nearly half of everything written was code that checks the other half.
11,663
times the whole thing re-checked itself. Every single change ran through the full test system before it counted. That's closer to how a professional software team ships than how a weekend project does.

And about a quarter of the total work wasn't new features at all. It was going back and rebuilding things that already worked, because good enough wasn't. That's the first corner people cut when time is tight. I didn't cut it.

who this is really for

Why should any of this matter to someone who will never look at the code? Because the app holds real families' information. A practice that doesn't save, a report showing the wrong kid, a login that leaks — to the parent on the other end those aren't glitches, they're broken trust. All that checking is what makes it safe to hand to real people. That was the whole point of it.

The AI wrote most of the code. I won't pretend otherwise. My job was the rest of it: deciding what to build, reviewing what shipped, tightening the security, and pushing it through tests until it behaved like real software. That part doesn't automate, and it's where the hours went.

One person, a few good tools, and hours I probably owed the social life.

329 features shipped 7,300+ automated checks 11,663 CI runs ¼ of the work was rework part-time · 104 days

Every number here is pulled straight from the project's own build history, not estimated. A real, shipped product; the private details stay private.