People assume that running your life “on AI” means a dozen slick apps. Mine is the opposite: it’s a folder of plain Markdown files, and an operator that reads and edits them on my instructions.

The two pieces

  1. A plain-text vault. Every part of my life is a file — contexts (who I am, my goals), projects, my travel clients, finances, even my daughter’s curriculum. Nothing is locked inside an app’s database. If a tool disappears tomorrow, I still own everything.
  2. An agent that operates it. Claude Code sits on top of that folder. I ask in plain language — “close out today,” “draft a reply to this inquiry,” “what did I say I’d do about the house?” — and it opens the right file, does the work, and writes it back.

Why plain text is the unlock

Because the agent and I read the same files. There’s no integration to maintain, no API to babysit. When I want a new capability, I don’t buy software — I write a short instruction file describing the routine, and the operator can now do it.

The mental model

I stopped thinking of AI as a chatbot I visit and started thinking of it as an operator I delegate to. The vault is the office; the agent is the assistant who knows where everything is. My job shrank to two things: deciding what matters, and reviewing what came back.

What I’d tell someone starting

  • Put your life in plain text first. The structure is 80% of the value, before any AI touches it.
  • Write instructions, not prompts. A reusable file beats a clever one-off message every time.
  • Keep the human in the loop on anything that leaves the building — money, messages to clients, anything you can’t take back.

This is the backbone everything else on this site is built on. Tagged #implementation.