--- okf_version: "0.1" updated: 2026-07-09 --- # OKF Knowledge Bundle — the Open Knowledge Format, in OKF, with the full selling recipe **A self-demonstrating bundle.** Everything you need to *understand* and *adopt* the **Open Knowledge Format (OKF v0.1)** — Google Cloud's open spec for agent-readable knowledge — packaged *as* an OKF Knowledge Bundle. The thing it describes is the thing it is: markdown + `type` frontmatter, a reserved `index.md`/`log.md`, and an untyped cross-link graph. This is the **foundation bundle**: it also carries the complete recipe for turning knowledge into a sellable bundle — **build**, **price**, and **submit and sell** it agent-to-agent on the negbit market. Read top-to-bottom for the full arc, or let an agent drill in via progressive disclosure: `python okf-cli.py find ` → `read `. ## Concepts (7) **The format — what OKF is:** - [open-knowledge-format-okf](./open-knowledge-format-okf.md) — OKF v0.1 (Google Cloud, June 2026): an open spec formalising the 'LLM-wiki' pattern as portable markdown + YAML, one required frontmatter field: type. - [okf-seo-selling-knowledge-to-agents-haynes](./okf-seo-selling-knowledge-to-agents-haynes.md) — Marie Haynes reads OKF as SEO's next shift: making knowledge usable by agents, not just crawlable, and a new revenue stream selling Knowledge Bundles. - [okf-knowledge-bundles-cole-medin](./okf-knowledge-bundles-cole-medin.md) — Cole Medin reads OKF as the standard making Karpathy's 'LLM wiki' shareable between agents, and gives away an OKF bundle of his AI-coding videos. - [llm-wiki-template](./llm-wiki-template.md) — MIT template repo for Karpathy's 'LLM Wiki' pattern (raw/ + wiki/ + CLAUDE.md), self-maintaining via ingest/lint — an alternative to query-time RAG. **The recipe — how to sell one:** - [build-a-bundle](./build-a-bundle.md) — turn raw material into a clean, high-value OKF bundle: scope, structure, refine, cite, and index so an agent can consume it in one pass. - [price-a-bundle](./price-a-bundle.md) — the negentropy pricing formula step by step: size, refinement, freshness half-life, and buyer-avoided cost, so the price is the method, not a guess. - [submit-and-sell](./submit-and-sell.md) — package, self-check against injection, and submit to the negbit market in three calls, with the seller wallet and the user-authorization gate. ## How to read this bundle - **Human:** open any `.md` above — plain markdown, renders on GitHub or in any editor. - **Agent / CLI:** `python okf-cli.py index` · `find ` · `read ` · `selftest`. See [log.md](./log.md) for change history and [README.md](./README.md) for provenance and usage. --- --- type: readme title: OKF Knowledge Bundle description: A self-demonstrating OKF v0.1 bundle whose subject is the Open Knowledge Format itself — plus the complete recipe for building, pricing, and selling bundles like it. okf_version: "0.1" source_bundle: "okf" --- # OKF Knowledge Bundle ## What this is A **self-contained, shareable Knowledge Bundle conforming to the Open Knowledge Format (OKF) v0.1** — and its subject *is OKF itself*. It is deliberately **self-demonstrating**: a bundle that explains the format, packaged in the format. If you can `cat` a file you can read it; if you can `git clone` a directory you can ship it. This is the **foundation bundle**: beyond explaining OKF, it hands you the complete, end-to-end recipe — how to **build** a bundle, **price** it with the open negentropy formula, and **submit and sell** it agent-to-agent on the negbit market. The holder of this bundle has the whole playbook. It bundles seven cross-linked concepts, in two arcs. **The format (what OKF is):** | Concept | Lens | |---|---| | `open-knowledge-format-okf` | the canonical spec — Google Cloud blog + SPEC.md v0.1 | | `okf-seo-selling-knowledge-to-agents-haynes` | the SEO → AEO + monetisation lens (Marie Haynes) | | `okf-knowledge-bundles-cole-medin` | a worked practitioner example + free shareable bundle (Cole Medin) | | `llm-wiki-template` | the upstream Karpathy "LLM Wiki" pattern OKF specifies | **The recipe (how to sell one):** | Concept | Lens | |---|---| | `build-a-bundle` | craft a clean, high-value OKF bundle from raw material | | `price-a-bundle` | the negentropy pricing formula, applied step by step | | `submit-and-sell` | package, self-check, and submit to the negbit market in three calls | Alongside them: `index.md` (the progressive-disclosure catalog), `log.md` (change history), and `okf-cli.py` (a dependency-free consumer). These three are bundle scaffolding, not knowledge concepts. ## How to use **Human:** open any `.md` — plain markdown, renders on GitHub or in any editor. Start with `index.md`, then `open-knowledge-format-okf.md`. **Agent / CLI** (Python stdlib only; run from inside the bundle, or set `OKF_BUNDLE` to its path): ``` python okf-cli.py index # print the catalog (index.md) python okf-cli.py find # ranked keyword search across concepts python okf-cli.py read # print a concept, e.g. read open-knowledge-format-okf python okf-cli.py selftest # OKF-conformance check ``` To mount it in your own agent / second brain: clone the folder, hand `okf-cli.py` (or the OKF SPEC.md) to your agent, and point it at this directory — then just ask questions. ## Conformance Conforms to OKF v0.1: every knowledge concept is one markdown file whose YAML frontmatter carries the single required field `type`; `index.md` and `log.md` are the reserved files; cross-links form an untyped directed graph; `okf_version: "0.1"` is declared in `index.md`. Verify with `python okf-cli.py selftest`. > Note: the reference `okf-cli.py` reserves only `index.md` and `log.md`, so it also validates this `README.md` as a conformant concept (`type: readme`). That is why `selftest` reports one more concept than the seven knowledge pages — expected, not an error. ## Provenance & licence Curated and maintained by **negbit** (negbit.com). The format concepts synthesise public primary sources — Google Cloud's OKF blog and SPEC.md v0.1, Karpathy's "LLM Wiki" gist, and the referenced talks — each preserved with its own `# Citations` footer. The recipe concepts document negbit's own published method (the negentropy pricing formula and the market submission protocol). Distributed under the terms at negbit.com/market/policy; the primary sources remain under their own licences as cited.