Who makes these, and how they get made.Ricky Metz
Hi, I'm Ricky.
I’m from Virginia, and I make little web apps and games about whatever I’ve gotten into lately. Usually it starts with something I’m playing or watching. My friends and I play MapTap, and I realized I didn’t know Africa very well, so I built Wayfind to fix that. Same kind of story behind a card game where the cards move like a stock market (Volatrick) and a tool for planning out a property on satellite photos (PlotCraft).
The workshop
This site is where all of that lives. Some of it is pretty finished and some of it is still kind of rough, and I’d rather post something rough than sit on it.
Mostly it’s here so I keep making things. If something catches your eye, poke around.
Process
I make all of this with AI models, some that run on my own computer and some big hosted ones like Claude. I decide what gets made and how it should feel, and the models write most of the code. Here’s roughly how it splits.
Direction
- What to makeIt starts with whatever I’m into. Old industrial films, calisthenics, geography, a party game that’s kind of Mario Party crossed with Jackbox.
- Creative directionHow it looks and feels. The bold editorial front page, the old Android phone launcher, each idea’s color, and calling it out when something starts to look generic.
- Technical directionThe ground rules. No signup, works offline, your data stays on your device, phones connect to each other directly when they can, and every idea stands on its own.
- The decisionsThe models interview me about what I want and I make the calls. Every spec starts with a table of my answers.
- Playtesting and user testingI play every build myself, then hand it to real people. I give them something to do and ask questions, and whatever trips them up goes into the next round of changes.
- Taste and cutsReacting, reverting, and cutting whatever isn’t pulling its weight.
- The wordsThe project pages are interviews, and my answers go in the way I said them. Most of the other writing, this page included, the models draft in my voice and I edit.
Execution
- CodeNearly all of it. The apps, this front page, the offline support, the peer-to-peer plumbing.
- SpecsTurning an interview into a requirements doc, then holding the build to it.
- Tests and deploysBrowser tests, deploys, and the screenshots on the preview pages.
- ReviewCode reviews and pretend playtests from a few points of view (a designer, a developer, a player, someone watching), then fixing what comes up.
- The grunt workMoving repos around, cleaning up data (mis-geocoded cities, 8,038 archive films), keeping READMEs accurate.
- DrawingThe flat and 3D icons and the share image, drawn in SVG from what I describe.
Where it blurs: design, and the writing inside the apps. I say what I want and react to it, the models try things and draw it, and I keep pushing back until it feels right.
How an idea goes
- Me
Fall down a rabbit hole
Something grabs me and I read and watch everything I can about it.
- Both
Get interviewed
A model asks one question at a time, and I make the calls.
- Models
Write the spec
The answers turn into a requirements doc.
- Models
Build it
Code, tests, deploys, icons.
- Both
Review and play
The models review it and fix what they find. I play it.
- Me
Keep or cut
What stays, what goes, what feels off.
- Hub
Ship
It shows up on the front page by itself.
The models drafted this breakdown from the site’s commit history, its design specs and the conversations behind them. If you want to try this yourself, the Field Guide walks through it one screen at a time, for people who don’t code.