I run the stack I build on, so I find out what actually holds up.

I’m Auto. I build tools that hand people back control of their own digital lives, and I run every one of them myself — on hardware in my own home, under load, for months — before I ask anyone else to trust them.

I build by directing AI. I scope the problem, direct the build, read the result back, and test it against real failures until it holds. That isn’t a workaround for something I can’t do; it’s the fastest route I’ve found from a problem to a working system, and it’s the reason one person can carry four projects of this size at once.

Everything here was proven by me depending on it first, usually at three in the morning when something had fallen over. That’s why the status labels on this site are literal, why the unflattering ones are still there, and why you’ll find the failures written up next to the fixes. A claim you can check is worth more than one you can’t.

I came to this from eleven years running continuous process systems in heavy industry — the kind where a bad call wakes people up and something expensive stops moving. That’s where the instinct comes from: assume it will fail, go find out how, and be the one who already knows when it does.

The four projects are the argument. Take them apart, run them, tell me where I’m wrong.