Summary
Cole Medin demonstrates a practical implementation of Google’s Open Knowledge Format (OKF), a new standard for building shareable, agent-queryable knowledge bases. Medin frames OKF as doing for agent-to-knowledge-base communication what MCP did for agent-to-tool interaction — a universal format that lets anyone build and share structured knowledge, not just consume it personally. In this video, he has packaged his entire YouTube channel — 200 videos — into a single OKF bundle queryable by any AI agent or second brain.
The walkthrough covers how the knowledge base is structured: an index file with thematic overviews, concept and entity pages derived from markdown transcripts, and raw timestamped transcript files for source citation. Medin shows the system answering complex multi-video questions — such as his complete end-to-end AI coding agent workflow — with accurate source citations and video timestamps, all triggered by a single agent prompt that clones the GitHub repository.
He also explains the pipeline used to build such a knowledge base from scratch: extracting YouTube transcripts as markdown, processing them through an LLM Wiki structure, and organizing by themes and entities — a process replicable for any YouTube channel or content corpus. The video includes a sponsored segment from SerpApi for integrating live web search into AI agents. Essential viewing for developers building personal AI second brains or looking to make large video archives searchable through conversational agents.
📺 Source: Cole Medin · Published July 30, 2026
🏷️ Format: Hands On Build
