← Field Notes

Finally, an Open Standard for the Karpathy LLM Wiki is HERE

This video introduces the Open Knowledge Format (OKF), a proposed standard for structuring personal knowledge bases built with LLMs. It addresses the lack of interoperability between different LLM wiki implementations, e

5 min read

Finally, an Open Standard for the Karpathy LLM Wiki is HERE

Source: Finally, an Open Standard for the Karpathy LLM Wiki is HERE — Cole Medin (https://www.youtube.com/watch?v=T33iI6izAKw) · published 2026-07-02
Speaker(s): Cole Medin
Relevant to: BexarByte, Bexar Labs, Agentic OS, Cross-business

TL;DR

This video introduces the Open Knowledge Format (OKF), a proposed standard for structuring personal knowledge bases built with LLMs. It addresses the lack of interoperability between different LLM wiki implementations, enabling agents to share and consume knowledge bases consistently. The video demonstrates how OKF can be used to build, refactor, and share knowledge bundles, with a practical example of an AI coding knowledge bundle. This is especially relevant for AI-native operations like Bexar Labs and Agentic OS, where structured knowledge sharing and agent interoperability are key.

Key ideas

  • [00:00] Andre Karpathy's LLM wiki concept gained popularity with a GitHub gist, but lacks a standard for sharing between users.
  • [01:21] The main problem is that each person's LLM wiki is structured differently, making it hard to share or for agents to understand.
  • [02:12] Google's Open Knowledge Format (OKF) provides a standard for organizing information and metadata in LLM wikis.
  • [03:23] OKF standardizes two things: how information is organized (entity documents, concepts) and the metadata fields used in each document.
  • [04:00] A traditional Karpathy wiki includes an index file, metadata, and links to related concepts, but lacks a consistent structure across users.
  • [07:00] OKF's spec.md file can be used in a coding agent to build or refactor a wiki, teaching the agent the standard format.
  • [11:03] OKF enables sharing of knowledge bases, which is especially useful for content creators and teams.
  • [12:45] The video includes an example of an OKF bundle containing AI coding YouTube videos, demonstrating how to use the format in practice.
  • [16:26] The agent can search through the bundle, progressively drilling into concepts and videos to answer questions.
  • [17:34] While some criticize OKF as too simple, the speaker argues that its minimalism is a strength, enabling broad adoption and interoperability.

Tools, services & specific callouts

Tool / service What it is How it's used in the video Use-case for my ventures
Open Knowledge Format (OKF) A standard for structuring personal knowledge bases using LLMs Used to build and share LLM wikis with consistent metadata and organization Enables Bexar Labs and Agentic OS to build interoperable knowledge systems for agents, improving collaboration and knowledge sharing
GitHub gist A simple markdown file for building LLM wikis Used as the original LLM wiki template by Karpathy Can be adapted to use OKF for better standardization and sharing
Posthog Analytics platform for understanding user behavior Used in the speaker's open-source AI coding harness builder, Archon Useful for BexarByte and Bexar Labs to track user interactions with knowledge systems and improve agent performance

How AI / automation is used

  • Generative AI is used to build and maintain personal knowledge bases, extracting key information from documents and integrating them into structured wikis.
  • AI agents are trained using the OKF spec to understand and consume knowledge bases in a standardized format.
  • The agent can search, traverse, and answer questions based on the structured knowledge graph, using metadata like tags and types for categorization.
  • The agent can be instructed to refactor existing wikis into the OKF format, using sub-agents to process large knowledge bases efficiently.

SEO / GEO / marketing / growth

  • Content creators can package their knowledge into OKF bundles, making it easier for others to consume and integrate into their own knowledge systems.
  • This approach can be used to grow an audience by offering curated knowledge bundles related to AI, coding, or other topics.
  • The video promotes Posthog as a tool for analytics, which can be used to track how users interact with knowledge systems and improve them over time.

Infrastructure & hardware

  • None in this video.

Notable quotes

"You can take this file, copy it, paste it into your coding agent, and ask it to build you an LLM wiki and it's going to be able to just oneshot it." — [00:20]
"There's no standard. And so, there's really not a way to share your LLM wiki with someone else." — [01:32]
"This is the future of personal agents. It's like what MCP did for agent to tool communication, this OKF is doing for agent to knowledge base communication." — [09:47]
"If everybody has the same standard for how they are building up their own personal knowledge base, everyone can share ideas more." — [10:23]
"The fact that there's not much substance here might seem counterintuitive, but I think that is actually a good thing." — [18:41]

Actionable takeaways

  • [Bexar Labs] Implement OKF as a standard for internal knowledge sharing and agent interoperability.
  • [BexarByte] Use OKF to create standardized knowledge bases for civic tech projects, enabling easier collaboration and data sharing.
  • [Agentic OS] Integrate OKF into the agent stack to ensure consistency in knowledge representation and retrieval.
  • [Cross-business] Use OKF to standardize knowledge management across all ventures, improving agent training and knowledge sharing.

Open questions / to verify

  • Is the Open Knowledge Format (OKF) an official Google product or a community-driven initiative?
  • What is the exact version or release date of the OKF spec mentioned in the video?
  • Are there any known implementations of OKF beyond the example provided in the video?

Filing metadata

  • Suggested title: Open Knowledge Format (OKF) as a Standard for LLM Wikis
  • Primary venture: Agentic OS
  • Secondary ventures: Bexar Labs, BexarByte, Cross-business
  • Type: Research
  • Keyword tags: LLM wiki, knowledge base, Open Knowledge Format, AI agents, knowledge sharing, metadata standard, personal knowledge management
  • One-line index hook: Open Knowledge Format (OKF) provides a standard for structuring LLM wikis, enabling agents to share and consume knowledge consistently.

AI-assisted summary of a YouTube video — source. Generated by a local model and human-reviewed; verify specifics against the original before relying on them.