Back to the catalog

soros-wiki — OKF bundle

The collected essays of George Soros (1984-2023) as an Open Knowledge Format v0.1 bundle.

Open source Repository Open in the app JSON README (API)

About

# soros-wiki — OKF bundle

An LLM-navigable wiki built from all **243 English essays** published at
[georgesoros.com/essays](https://www.georgesoros.com/essays/), spanning
1984-2023. Raw text was scraped via the site's WordPress API and synthesized
into interlinked source / concept / entity / theme pages.

**Counts.** 12 themes · 66 concepts (39 stubs) · 799 entities (724 stubs) · 243 sources. **Total: 1120 pages. Zero dangling links.**

Stub pages are nodes referenced across the corpus but not yet given a dedicated
synthesis. Each stub aggregates its back-references (where it is mentioned +
surrounding context) so it is an informative node, not an empty placeholder.

## Themes (12)

- [Brexit](/themes/brexit.md)
- [China](/themes/china.md)
- [Migration](/themes/migration.md)
- [Open Society](/themes/open-society.md)
- [Philanthropy](/themes/philanthropy.md)
- [Reflexivity and Markets](/themes/reflexivity-and-markets.md)
- [The Euro Crisis](/themes/euro-crisis.md)
- [The Future of Euro

Details

Kind
OKF bundles
Topic
Productivity
Publisher
profishdan
Origin
okf_github
Category
dados
Version
0.1
Stars
24
Forks
2
Last push
2026-07-17T17:30:17Z
Repository state
ativo
Language
HTML
Added
2026-09-09 19:04:11
Updated
2026-09-09 19:04:11
Origin id
profishdan/soros-wiki:index.md

README

# soros-wiki

An [Open Knowledge Format](https://github.com/GoogleCloudPlatform/knowledge-catalog)
(OKF v0.1) knowledge bundle built from **all 243 English essays** of George
Soros published at [georgesoros.com/essays](https://www.georgesoros.com/essays/),
spanning 1984–2023.

It is an LLM-navigable, Obsidian-viewable wiki: a directory of interlinked
markdown files. Nothing depends on a network service.

![Link graph of the wiki — 1120 nodes, 9546 edges, color-coded by page type](graph_sc.png)

*The full link graph (from [`viz.html`](viz.html)): themes and major concepts
as hubs, sources in gray, entities in green, concepts in blue, themes in red.*

## State

- **1120 pages** — 12 themes, 66 concepts, 799 entities, 243 sources
- **13,610 internal links · 0 dangling links**
- **763 stub pages** (39 concepts, 724 entities): nodes referenced across the
  corpus but not yet given a dedicated synthesis. Each stub aggregates its
  back-references (where it is mentioned + surrounding context), so it is an
  informative node, not an empty placeholder.

Start at [`index.md`](index.md) — the catalog of every page.

## Layout

```
soros-wiki/
├── index.md       # reserved: catalog of every page
├── log.md         # reserved: build history
├── README.md      # this file
├── raw/           # 243 scraped essays (clean markdown, faithful to source)
├── sources/       # one page per essay (summary, key claims, quotes, links)
├── concepts/      # recurring ideas (reflexivity, open society, boom-bust, …)
├── entities/      # people, orgs, countries, instruments named across essays
├── themes/        # 12 top-level topic clusters
└── _build/        # the scripts + intermediate data used to build the bundle
```

## How it was built

1. **Scrape** — all 243 posts in the English `essays` category pulled via the
   site's WordPress REST API, rendered HTML converted to markdown into `raw/`.
2. **Source pages** — each essay synthesized into a `source` page against a
   controlled vocabulary (`_build/vocab.md`) so cross-link slugs stay consistent.
3. **Graph** — full pages authored for the most recurrent concepts (27),
   entities (75), and 12 themes; stubs auto-generated for every other
   referenced node, each with back-references.
4. **Validate** — `_build/validate.py` confirms 0 dangling links and that every
   non-reserved file carries a `type`.

See [`log.md`](log.md) for details. Build scripts live in `_build/` (run
`reindex.py` after editing pages to regenerate `index.md`; `validate.py` to
re-check link integrity).

## House style

- Links are **bundle-root-absolute**: `[reflexivity](/concepts/reflexivity.md)`.
- **Path = identity** — the file path is the canonical key; no slug table.
- Frontmatter: required `type`, plus `title`, `description`/section content,
  `tags`, `timestamp`. Sources also carry `url`, `year`, and `raw_path` (the
  local raw essay, an OKF extension field).

## Conventions for a few key page types

- **source** — `## One-liner`, `## Summary`, `## Key claims`, `## Entities`,
  `## Related essays`, `## Quotes`, `## Source`.
- **concept** — `## One-liner`, `## Definition`, `## Origin`, `## Recurrences`,
  `## Adjacent concepts`.
- **entity** — intro, `## Soros's take`, `## Mentioned in`, `## See also`.
- **theme** — `## Position`, `## Key essays`, `## Core concepts`, `## Key entities`.

## Filling out stubs

To replace a stub with a real synthesis: read its `## Referenced in` section,
follow the back-references into the source pages (and via each source's
`raw_path:` into the raw essay), and rewrite the body. The stub already has the
right frontmatter and inbound links — only the body needs work.

## Viewing

- **viz.html (most reliable)** — open `viz.html` in any browser. A
  self-contained Cytoscape graph of all 1120 nodes / 9546 edges, color-coded by
  type. No server, no network, no Obsidian. Regenerate with `_build/viz.py`.
- **Obsidian** — open the folder *itself* (`wiki-soros`) as a vault. Each page
  ends with a plain `**Connections:**` line of `[[wikilinks]]`
  (`_build/obsidian_links.py`) because Obsidian's graph counts wikilinks, not
  the `/`-rooted markdown links used in the body. If pages still look isolated,
  fully **quit and reopen Obsidian** (its link index is cached) and make sure
  the vault root is this folder, not a parent.
- **GitHub** — absolute markdown links render directly.
- **OKF graph visualizer** — point Google Cloud's `knowledge-catalog`
  `visualize` at this folder for an in-browser force-directed graph.

More