Harry Potter Uno

When I tell somebody I have a home lab server, they look at me like I’ve got a biohazard situation in my basement. I don’t know what they picture. Some blinking wall of machines, cables everywhere, a room you need a keycard to get into.

In reality, it’s two mini PCs, each about the size of a router, sitting on top of a hard drive bay. One is all you’d need. Plug it in, connect it to the internet, and you have a server. I live in Hancock County, Kentucky. Population about 9,000 across the whole county, and my family’s been here since around 1800. This little box in my house is the entire infrastructure behind everything I’m about to describe.

---

For as long as software has been an industry, building it required moving to a city. You learned to code, got hired at a company in San Francisco or Austin or New York, and spent your career building products for millions of people you’d never meet, solving problems you’d never personally had. That’s not a knock on anyone. It was structural. Software was expensive to build and expensive to run, so it needed scale to justify itself. Scale meant cities. Cities meant leaving.

The whole industry was a sorting machine. People with the ability to build got pulled toward metros. People with the deepest context, the ones actually embedded in their communities, who knew their neighbors, who stayed in one place, were locked out. Not for lack of ideas. Because execution was the gate, and the gate was shut.

That gate is opening.

---

A few weeks ago, on a Sunday afternoon, my buddy Blake and I got on a call with no plan. I’d been telling him about Claude Code, which is an AI that writes software from a conversation in your terminal, and we’d been saying for a while that we should build something together. We finally had the afternoon free.

Blake mentioned Uno. That was it. We were off.

I pulled up Excalidraw and we started sketching. What screens do we want? What should the game look like? How does a turn work? Just rough doodles, nothing serious. Then I tossed those sketches into Claude Code, gave it the context of my server setup, and told it to ask us questions about what we wanted.

We went back and forth for a while. Blake and I talking through game logic, the AI asking clarifying questions, planning out the architecture, writing code. Neither of us typed a single line. We were just describing what we wanted, answering questions about trade-offs, making decisions. The code happened as a side effect.

Around that time our friend Drew hops on. Drew’s a graphic designer. We tell him we’re making Uno and share the Excalidraw. He doesn’t miss a beat, pulls up Illustrator, starts designing pixel art cards.

Three friends on a Sunday call. One doing design, one talking through logic, one managing the AI and the infrastructure. Nobody “coding.” A multiplayer card game materializing in front of us.

But we weren’t building regular Uno.

When we were kids, we played Harry Potter Uno. It had an invisibility card. It had draw three instead of draw two. The rules were different in small ways that mattered to us, the way things from when you’re twelve just do. You can’t find that game anywhere online. No company would ever build it. There is no market for Harry Potter Uno with our specific house rules. That version only exists if we make it ourselves.

And we always knew that. We’d known it for fifteen years. “It’d be cool if there were a good Uno online.” And then you move on with your day, because what are you going to do about it? Learn to code?

Now you can do something about it. On a Sunday. Without learning anything.

We told the AI to build the engine so it’s extensible. Not just Uno, a card game engine that supports different rule sets. Harry Potter Uno out of the gate. It figured out the architecture, built it, and when things broke we described the problem and it fixed it. I don’t even remember how it’s architected. WebSockets, React, something. Doesn’t matter. It works.

I had to leave before we got it deployed. That evening, a couple more prompts to finish, a couple more to set up auto-deploy. Now Blake pushes changes from his laptop and my server picks them up automatically. Anybody with the link can play.

Blake is not a software engineer. He has never been a software engineer. He’s pushing deployable code to a live multiplayer game from his laptop.

A week or so later I built a multiplayer drawing game. Same thing. Idea to working product in a few hours. I’d always wanted a better Scribble. Now I have one.

Could either of these handle a million users? No. I want to serve like twelve. My friends, maybe their friends. These things take next to nothing to run.

---

There’s a quote I’m going to butcher, but the gist of it is: humans discovered an invisible field that permeates and governs the entire universe, and the first thing we did was put music on it.

We built the most powerful AI tools in history, and the first thing through the gate was a card game three friends played as kids.

That’s not a frivolous use. I think it might be the point.

We’ve been eating slop out of a trough for decades. The products that won at scale are the only ones that exist. You use Google’s version of email. Spotify’s version of a music library. Discord’s version of a chat room. Not because these are the best possible versions for you. They can’t be. They were built for everyone, which means they were built for no one in particular. They just won the scale game and nothing else survived.

The whole structure existed because building software was expensive. You needed engineers, infrastructure, capital, and scale to justify all of it. You couldn’t build the thing yourself, so you used what someone else built for everyone. The gate was shut, and the only software that got made was software built by people who could get through it. People in cities, at companies, building for strangers.

That lock is breaking. And what’s coming through looks nothing like what the tech industry has been building, because it’s coming from people who were never part of that industry and never wanted to be.

What they have instead is context.

Not taste, not some founder’s singular vision. Context. The specific, deeply personal knowledge of what these people in this place actually need. The knowledge that your friend group played Harry Potter Uno with house rules. The knowledge that your neighborhood could use a better way to coordinate the Saturday farmers market. Stuff that only exists because you’ve been somewhere, with the same people, for a long time.

No company can have that. No amount of user research or analytics will surface it. It’s the one input the AI can’t supply. The AI builds whatever you describe, but it can’t know what to describe. That part comes from being present, in a place, with people, over years.

And this is still early. What took us a Sunday afternoon will keep getting easier. In a couple of years, you won’t need the afternoon. You’ll mention offhand to your agent that you wish there were a better version of some game you used to play, and it’ll come back a few minutes later with a working draft. The distance between wanting something and having it is collapsing.

When it does, cloud doesn’t die. It just retreats to the stuff that genuinely needs scale. Training runs, inference, big data. But the apps? The things you actually use? Discord doesn’t have to be run by Discord. Your email doesn’t have to be Gmail. Those can live on a box in your house, maintained by an AI that understands both the code and what you want from it.

---

There’s a reason the first thing through the gate is coming from rural Kentucky and not San Francisco.

If you live in San Francisco, scale is the only lens you have. The city only exists because of economies of scale. The most valuable companies on earth, all building for a billion users they’ll never meet. When that’s your whole world, personal software doesn’t make sense. Why build for twelve people?

Rural communities rejected that a long time ago. Otherwise they wouldn’t still be here. They chose small on purpose. And the context has always been thick in places like this. People know each other. There are third spaces. You see the same families at the school, the restaurant, the gas station. What was missing was the ability to act on any of it.

I lived in Austin for a while. Could have stayed. There’s no reason to anymore. I can live on a few acres, have a yard, raise my kid here. I was thinking about the phrase “touch grass” the other day and realized I’d forgotten that’s not a trivial thing for most of the country. For me it’s daily. I walk around barefoot when the weather’s nice.

And I can build software on a Sunday afternoon that my friends across the country can use by that night, running on a box in my house that’s smaller than the router it’s plugged into.

I eat at the Mexican restaurant owned by Jose, who sends his kid to the same public school as everyone else in the county and brings his food truck to the elementary school games on weekends. I know the people here. When I build something, I build it for people I can name.

That’s what software should be. Something you made on a Sunday for the people you know, running on a box you can touch, in a place you chose because the grass is right outside.