Loading…
Loading…
The actual setup. What runs, what broke,
and the rules that keep it sane.
I've used AI every day since November 2022. At some point it stopped being a tool I open and became a system that runs whether I'm at the keyboard or not.
This page is the current state of that system. Not the demo version, the real one. It changes as the setup changes, which is often, so the date above matters. If the page and reality disagree, reality won and I haven't updated yet.
The biggest upgrade in how I work had nothing to do with models. I stopped typing to AI and started talking to it. Dictation is faster than my fingers, and the thinking comes out less filtered, which turns out to be exactly what the AI needs. I wrote about the switch in Why I Stopped Typing to AI.
The thing is, the AI doesn't know what you know. Every new session starts from zero unless you fix that yourself. So I built a second brain. A plain text wiki that holds my projects, my plans, what happened last week, what I'm trying to learn. Every AI session I start reads it first, and updates it before it ends.
The structure is simple on purpose. One hot page with recent state, one log of what happened, one index of everything else. Markdown files in a git repo. No database, no app, nothing to maintain. The boring format is the feature. Plain text survives every tool change, and there will be tool changes.
An old MacBook Pro at home runs an AI session around the clock with the second brain as its only context. I message it from my phone and it answers, whether my laptop is open or not. Scheduled jobs are moving onto it too. Morning briefings, research digests, the kind of background work an assistant would do if I had one.
Getting it stable taught me more than any tutorial. My favorite lesson so far, the messaging API only allows one listener per bot. I learned that the way I learn most things, by accidentally running two and spending an evening wondering why nothing answered.
An autonomous system that can change itself unsupervised is not a system, it's a liability. So mine has one hard rule. It can propose anything, it can apply nothing. Improvements to how it works arrive as pull requests I review on my phone. Most days I approve. Some days I don't, and those days are the reason the rule exists.
Same thinking on access. The always-on machine can read my notes repo, answer my messages, and that's about it. I looked hard at the popular self-hosted agent frameworks and walked away. Giving a 24/7 process broad access to your digital life is a decision you only get to regret once.
None of this is about writing code faster, even if it does that too. It's about having a thinking partner with perfect memory across everything I care about. Work, the band, the house, learning. And it's practice. Every piece I set up teaches me how these systems behave in the real world, where the pitfalls are, what I'd do differently. The way I see it, that knowledge is the actual product. The setup is just where I earn it.
Loaded into every AI session I start. The memory layer.
I message it from anywhere, it answers from home.
An old MacBook Pro that never sleeps. Final cutover in progress.
Scans my field overnight, opens a summary I read over coffee.
Drafts from my own notes and ideas. I edit, I publish. Always.
Updated by hand when the setup changes. No automation here.