wiki_indexed	2026-05-24T02:24:11Z
wiki_hash	9498ea72bd72bd588e132ec326c2626742b3b2575bad8b3cd64e8e8490b47061
wiki_role	wiki

wiki-ai-thinkers

An Obsidian vault that collects authoritative primary sources from prominent AI researchers, founders, and policy thinkers - and renders them as a queryable knowledge graph.

Each thinker has a profile.md with structured positions on AGI timelines, takeoff dynamics, alignment, economic impact, and policy. Each source has a source.md with verified citation metadata, a one-line summary, and links back to the thinker. Two browser artefacts (graph.html, positions.html) read the vault directly via the File System Access API - no build step, no server, no Claude API calls.

What is in here

• 15 thinkers under 20_People/<slug>/ with profile.md and a curated source set.
• Hundreds of source records under 10_Sources/Paper|Talk|Article|Video/<slug>/ - every record names the thinker as a primary author or named speaker.
• Auto-generated Maps of Content under _wiki/MOC-*.md and curated MOCs under 50_MOCs/.
• A locked schema under 90_Templates/ and 99_Meta/schema.md.
• Two browser artefacts under .workspace/ - a force-directed graph and a positions matrix.

Architecture

wiki-ai-thinkers/
├── 10_Sources/             Citation records grouped by source type
│   ├── Paper/<slug>/source.md
│   ├── Talk/<slug>/source.md
│   ├── Article/<slug>/source.md
│   └── Video/<slug>/source.md
├── 20_People/<slug>/
│   ├── profile.md          Structured positions, affiliations, role
│   └── positions.md        Optional rich position essays (currently: dario-amodei only)
├── 30_Concepts/            Concept scaffolding (alignment, safety, governance...)
├── 40_Domains/             Domain scaffolding (AI, economics, policy...)
├── 50_MOCs/                Hand-curated Maps of Content
├── 90_Templates/           Schema templates - source.md, person.md, positions.md...
├── 99_Meta/                Vault schema, build scripts, automation index
├── _wiki/                  Auto-generated MOCs from the wiki skill
└── .workspace/             Browser artefacts (graph.html, positions.html)

The browser artefacts

After cloning, double-click START_HERE.html at the vault root - it opens in your browser with step-by-step launch instructions for your OS. The short version:

• Windows: open .workspace/ in File Explorer (Win+E) and double-click serve.bat. (VS Code / Cursor sidebar opens .bat files as text; use the editor terminal: .\serve.bat.)
• macOS / Linux: cd .workspace && ./serve.sh.
• Any platform: python .workspace/serve.py.

Then click Connect vault and pick this folder. Read-only. The browser remembers the folder via IndexedDB, so subsequent visits one-click reconnect.

Requirements: Python 3, Chrome 86+ or Edge 86+ (File System Access API; Firefox and Safari are not supported).

Schema

Defined in 99_Meta/schema.md. Templates in 90_Templates/. Every record carries YAML frontmatter - source.md and profile.md are the two schemas the browser artefacts depend on. If you fork this vault to track a different roster of thinkers, keep the frontmatter keys intact and the artefacts will work unchanged.

How sources are added

A scheduled job runs weekly on the maintainer's machine. It dispatches a per-thinker researcher subagent that finds new primary-author papers, talks, articles, and videos, writes one source.md per accepted candidate, and appends wikilinks to the corresponding profile.md. Per-thinker state lives under 99_Meta/wiki-thinkers/ (excluded from the published repo). Clones of this repo get a snapshot, not the live researcher.

A note about cached PDFs

80_Attachments/ and 20_People/*/papers/ are excluded from version control. They hold cached PDFs of the source set - useful for offline reading but large, redundant (every source has a canonical_url in its frontmatter), and outside the scope of what should ship in a public repo.

To rebuild caches locally, walk the canonical_url field in each 10_Sources/**/source.md file.

Licences

This repo is dual-licensed.

• Content (10_Sources/, 20_People/, 30_Concepts/, 40_Domains/, 50_MOCs/, 90_Templates/, 99_Meta/, _wiki/, and all other markdown) is released under Creative Commons Attribution 4.0 International (CC BY 4.0). See LICENSE-content.txt. Attribution requested; commercial use permitted; no share-alike requirement.

• Code (.workspace/, serve.py, serve.bat, serve.sh, and all JavaScript / Python / shell files) is released under the MIT License. See LICENSE.

The licence that applies to a file is determined by which of the two categories above it falls into. Files under .workspace/ are code even if they happen to be HTML.

Acknowledgements

Source records cite their primary authors and venues directly in YAML frontmatter (authors, venue, year, canonical_url, doi). Citation metadata follows the New Zealand Law Style Guide where applicable; otherwise standard scholarly conventions.

The vault layout draws on Karpathy's "LLM Wiki" pattern (see kaparthy-pattern.txt in earlier history) - separating raw sources from per-entity profiles from auto-generated indexes.

Not a contribution project

This is a personal research vault that happens to be public. Pull requests adding or modifying thinkers will not be merged. Fork freely and run your own roster.