I’m publishing how I work for one reason: public things are easier to refine. I can reach this page from anywhere, link to it, get feedback on it, and git keeps every version. This page will change. That’s the point.

The brain holds the code

My base is an Obsidian vault: my second brain. My code repositories live inside it. Notes and code sit side by side, so an idea can point at the code that implements it, and a decision in the code can point back at the reasoning behind it.

Repos and ideas connect and develop together. I call the result 100x engineering. I can’t measure the multiplier, but I can feel the difference.

Workstreams, each with a diary and documents

The structure I’ve landed on, for now:

workstream/
├── diary/
│   └── week/
│       └── day
└── documents/
    └── persistent references, notes, anything

• Diary: what happened, day by day, grouped in weeks. Time-ordered. I build with a day-by-day philosophy: evolve a little, every day, and write it down.
• Documents: what should stay true beyond any single week. Reference documents, business rules, notes. A local mirror of the repos, organized by workstream instead of by codebase.

It’s the same idea as the two caching lanes on this site: separate what changes every day from what should stay stable, and treat each one accordingly.

Me as the orchestrator, AI as the pair

The highest orchestration point in this system is me. I decide what matters, what ships, and what it means.

Claude is my pair. It works inside the vault, so it can read the notes and the repos together. A CLAUDE.md file at the root of the brain holds how I think and build, so every session starts from my principles instead of from zero. Every session ends with a receipt: one real thing done.

This site is an example. In one day of that loop, a lost essay was restored, five essays were rewritten or written, a first visit got about half as heavy (and repeat visits stopped re-downloading 3 MB per page), and every change went live through the pipeline.

What I stopped doing

I used to sync the whole vault between two computers with git, to keep both brains aligned. It brought more cons than pros, so I stopped. The code repos still live in git. The brain doesn’t need to.

That’s part of the system too: update your beliefs, including beliefs about your own tools.

Still evolving

One caveat from my own essays: telling isn’t doing. A published workflow only counts on the days I actually use it.

This structure will keep changing, and every change will follow the same rule: refine, never distort.