The Positioning Vault

Most Obsidian setups are archives. Mine can't afford to be.

Builds-on: the-positioning-vault-pattern Informs: Projects/tech-blog

A lot of people use Obsidian, or Notion, or a personal wiki, or a CLAUDE.md file, as some flavor of digital prosthetic. Extended mind. Second brain. The language doesn't matter. The shape is: drop things in, find them later, hope the graph does something useful on the way back out.

For most people, an archive is enough. You remember roughly where you put the thing, the search works, the links are nice-to-have, and your working memory handles the relationships between ideas while you're actively thinking. The file system is a backup, not the primary.

Some of us can't run that way. My working memory does not reliably hold the relationships between things. That's not a complaint — it's just the hardware. And it means a neutral store is not a prosthetic. A neutral store is a bigger pile to lose things in.

Three weeks into building my version of this, I asked Claude to take stock of what was in the folder. What fell out wasn't an archive. It was a pattern that reads relationships out loud to me every time I open a file. It took me a while to realize nobody in public is doing exactly this — and that it probably deserves a name.

What fell out

The folder is three things at once:

1. Knowledge compaction — a memory directory that persists across sessions. Who I am, how I work, what I've tried. Ground truth about the person using the tool.
2. Strategy development — a docs folder with a dense typed graph. Not tagged cloud. Explicit relationships between every document.
3. Code and artifact production — tools that ship, blog drafts that merge, sibling projects that exist because of arguments made in prior docs.

Most public templates keep these separate. I didn't have that luxury. If the memory isn't paired with the active thinking, I forget the context I had yesterday. If the thinking doesn't spawn artifacts, the thinking doesn't compound. If the artifacts don't link back to the reasoning, I lose the why within a week.

The load-bearing piece is the typed links. Every doc has them, placed right after the title:

• Related — conceptually near
• Builds-on — this doc extends or deepens that one
• Led-to — this research produced that artifact
• Informs — connects to a real project or codebase

These are not decorative. They are the prosthetic. My brain doesn't have to hold "this connects to that." The file says it. When I open any document, I know within five seconds what it descends from, what it produced, and what it's currently arguing with.

The second load-bearing piece is slash commands. The folder isn't a filing cabinet, it's a REPL. /research to deep-dive a topic and connect it to my situation. /position to re-run gap analysis against my current profile. /magi for three-perspective review — Operator, Contrarian, Mirror. That last one matters because internal review collapses to one voice if you're wired a certain way. External voices force a shape that self-talk doesn't.

The landscape, as of early 2026

The meta-question I ended up asking: is anyone else doing this, and do they call it anything?

Short answer — pieces of it are emerging everywhere, no consolidated name:

• Karpathy's "LLM Wiki" (April 2026 gist) — markdown files maintained by Claude Code for agent knowledge compaction. Most-cited origin term. Knowledge-first. No typed links, no code artifacts.
• Titus Soporan's "Spine pattern" — meta-repo above sibling codebases, layered CLAUDE.md routing, a task system. Code-only. No PKM layer.
• obsidian-claude-pkm, Kempenaar's Medium series, a handful of adjacent templates — same folder shape (CLAUDE.md + Obsidian + memory), but without typed relations or code spin-outs.
• Geoffrey Litt, Maggie Appleton — conceptual parents. Writing about the shape of malleable software and tools-for-thought rather than shipping templates.

Nobody I could find is running all three — knowledge compaction, a typed essay graph, and real code and drafts spawned from the same folder — with conventions this formal. The typed-link discipline is the absent part of the public conversation. It's closer to Andy Matuschak's evergreen-notes habits than to the LLM-Wiki crowd.

What I'd call it

Positioning vault.

Not a wiki. Not a second brain. Not a PKM. Those are archival framings.

A positioning vault is infrastructure for a specific person's active decisions. Research feeds decisions. Some decisions spawn code or drafts. Memory holds ground truth about the person using it. Slash commands make the vault operable as a tool rather than passive storage.

The whole thing has a direction. The person using it is trying to get somewhere specific, and the folder is shaped around that trajectory. That's what makes it a positioning vault instead of a library.

The portable piece

You don't need to be wired the way I am to get value out of this. The typed-link discipline is portable outside PKM entirely. Teams could use it to track decisions → artifacts instead of ideas → essays. Every decision doc gets a Led-to: pointing at the ticket, PR, or design doc that executed on it. Every artifact gets a Builds-on: pointing at the decision it implemented. Over time the graph becomes a record of the team's actual reasoning, not just its outputs.

That's the generalization: separate the convention from the cognitive motivation. The people who need the prosthetic most will adopt it first because the pain is loudest. Everyone else can adopt it because it's cheap discipline and it compounds.

If you're already running something close to this, the question I'd leave you with is: which of the three hats is your folder actually wearing, and which are you missing? Most people are wearing one of them well. The interesting version shows up when you collapse all three.

The full architecture, conventions, and usage audit is in the research vault — including the leak fixes I'm working through and what breaks when you try to formalize a system that was never supposed to be public.