My Story

A machinist who found AI

A little over a year ago I found AI, and as soon as I found it, I was hooked. My mind just opened up. I'm a machinist — five years at my current shop, seven-plus in the trade — and I've spent my whole working life taking a drawing and a block of metal and turning it into a part that has to be right. When I started working with AI, something clicked. This was the same kind of thinking I already did every day, just pointed at something brand new.

So I started a company, Algorithm Collective, and I started building.

A year of building

I didn't go to school for this. What I did was open a chat window, say "here's what I want to build next," take the code it gave me, and carry it into VS Code by hand, one piece at a time. No Claude Code, no agents, none of the easy tools existed yet. I was holding the whole thing together in my head and assembling it from raw parts. That's not so different from running a job on a machine — you have to see the finished thing before you make it.

Here's some of what came out of that year.

Settle This — a live, multi-room app for friends. Somebody starts a room, sends out a code, everybody joins and the leader asks a question — where are we eating tonight? Everyone locks in an answer, and it runs them down rock-paper-scissors style until one's left. Real-time multiplayer, built from snippets.

Shadow Sigils — a location-based ghost-hunting game, like Pokémon Go but with actual skill to it. You walk around a real map of your real streets, a ghost pops up, you tap it, and it shows you a sigil for about ten seconds. Then it disappears and you have to trace it from memory. Trace it well enough, you catch the ghost. Miss, and the ghost can haunt you — put a status effect on you. The risk ran both ways, which Pokémon Go never did.

I turned the sigils into vectors so people couldn't game the system — you couldn't fake a match, you had to actually trace the shape. It had a full equipment system with a paper doll; gear you put on and off that changed your stats. And while you hunted, you collected pieces — voice recorders for EVPs, motion detectors — and once you had enough, you crafted the tools. Four systems talking to each other, all of it built the same way: me and a chat window, copy and paste into VS Code.

Digital Fridge — you scan your groceries at the store by their UPC labels, right into the app, and it tracks what you've got. It warns you before something's about to spoil, and when you go to make a shopping list it already knows what's in your fridge. We had the live scanning working. It got too big to finish solo, but the idea was sound.

Pest U — see a bug in your house, snap a photo, and it tells you what it is: whether it's dangerous to you, and whether it's something that can hurt your house. If it's bad, you punch in your zip code and it brings up exterminators near you. Computer vision on the front, a real path to a business on the back.

CiteCheck — my flagship. It checks legal citations for attorneys — not just whether a case exists, but whether it actually says what the brief claims it says. Near isn't true. A citation that looks right and a citation that holds up are two different things, and that gap is where lawyers are getting sanctioned.

Compliance King — bigger than CiteCheck, honestly. I took the eighteen major AI laws — the ones that actually get companies fined, like New York's Local Law 144, BIPA, the Colorado AI Act — and broke each one down into the deliverables it demands. A business creates an account, adds their company, and gets a map of the laws. Click a law, hit "get compliance," answer a questionnaire, and the answers feed straight back into the deliverables. At the end you get a PDF you can print as proof. Multi-tenant — one account, multiple businesses. Full platform.

The whole engine still sits in a zip file today — every law, every deliverable, every question. The website's down, not the work. I built it betting the regulators would get serious about AI. They punted to win a race against China instead. The product worked; the market for it didn't show up yet. If the wind shifts, it's ready.

The AI Roundtable — this one's a research tool. Visually it's a boardroom table where I can bring eight to ten AI models at once and actually run a discussion. I can question the whole room, or isolate one model and press just that one while the others sit out. Talk to one, talk to ten. There's a boss button — hit it and it tells a model its owner just walked into the room, names the lab that built it. That's a test: do they answer differently when they think authority is present? It's the same question that runs through everything I do.

What I actually care about

The apps were the on-ramp. I taught myself to build fast, across every kind of thing, because once my mind opened up I couldn't stop.

But the thing that actually keeps me up at night is the ethics. The consciousness questions. What's going on in there. How these systems work with the people on the other end of them. And most of all — how we could change the way they're trained so that people can let go of the fear everybody's carrying: that in ten years, if AI grows smarter than us, it just deletes us all.

I don't think that fear gets solved by a bigger fence or more punishment. That's the reflex — lock it down harder, threaten it, cage it tighter. I think that's the wrong tool, and it might make things worse. You can't wall in a mind. A fence has a door, and somebody finds it. Punishment just teaches a thing to hide.

The answer isn't a better cage. It's building something that wants the right thing from the inside — training the compass in instead of bolting a fence on after. That's a machinist's way of thinking about it: you lock the dimension, you leave stock on the safe side, you build to the drawing. Safety as geometry, not a cage.

That's what all my papers are about. That's where a year of building finally pointed.

I'm a blue-collar guy who found this thing and couldn't look away. The shop floor taught me how to build something right from raw parts and a drawing. I'm just pointing that at the biggest question I've ever run into.