8.2
Obsidian As The Durable Brain
What you need for this one
- Download ObsidianFree. Take the Universal build for Mac.
Why this exists
Last lesson found the gap: you have durable knowledge and nowhere durable to put it. This lesson fills it.
The specific tool matters less than the properties, but you need to actually pick one, because "I'll organise my notes properly at some point" is where durable knowledge goes to die.
The idea
Plain files in a folder you own. Everything else is preference.
That's the requirement, and it's short for a reason. Whatever holds your durable knowledge has to be:
| Property | Why |
|---|---|
| Plain text | Readable in fifty years, by anything |
| Local files | Not hostage to a company or a subscription |
| Searchable | Findable in seconds, not browsable |
| Readable by Claude | It's a folder — so it's a project |
That last one is the reason this fits everything you've learned. A vault of markdown files is a project. Point Claude Code at it and every skill from Level 1 onwards applies — it reads the files, it works in the folder, it can search across everything you've written.
Obsidian is a good default because it's exactly this: a folder of markdown files, with a decent interface over the top. But the folder is the thing. The app is a viewer.
Structure loosely, and less than you want to. The instinct is to design a beautiful hierarchy before writing anything, and it fails every time — you'll spend an afternoon on folders and never fill them. Three or four folders, start writing, let the structure emerge from what you actually accumulate.
What goes in, based on 8.1:
- How you work — triage rules, the daily loop, escalation format
- What you've learned — things you worked out that cost you time
- Why decisions were made — the reasoning that evaporates
- What spans systems — knowledge with no single project to live in
Write it when you learn it, not later. Later is a lie. The moment you've solved something is the only moment you understand it well enough to write it down usefully, and the understanding fades faster than you expect.
The test of a note is whether it helps you in six months when you've forgotten the context. Write for that person. They know less than you do and they're in a hurry. That means: say what the problem was, not just the answer.
Do
- Create a vault — a folder for markdown files. Obsidian is a fine default; the folder is what matters.
- Make three or four folders, no more. Something like: how-i-work · what-i-learned · decisions · systems.
- Move your durable artifacts in — from 8.1's audit, the ones that were durable rather than project-level.
triage.md,daily-loop.md, the escalation format. - Write three new notes from things you learned in this course that aren't written anywhere yet.
- For each, write the problem, not just the answer. "Sessions get worse over time and here's why and what to do" beats "use /compact".
- Point Claude Code at the vault and ask it something that requires reading across several notes.
- Add a
CLAUDE.mdto the vault — what this is, what's in each folder, how you name things. Level 3, applied to your own brain. - Write one note titled "what I'd tell someone starting this job" and put in it the five things you wish you'd known in week one.
Step 8 is the most valuable note you'll write. It's also the one you can only write now, while week one is still recent enough to remember and far enough away to see.
Done when
Never do
Don't design the structure before you have content. You'll build a beautiful hierarchy and never fill it. Three folders, start writing, let it grow.
Don't write only the answer. In six months you won't recognise which problem it solved, and an answer you can't match to a problem is the same as no note at all.