Descriptions:
Matt Wolfe walks through the complete construction of an AI-powered second brain knowledge management system built on OpenAI’s Codex, Obsidian, and a Chrome web clipper — moving well beyond passive note storage to create an active, queryable knowledge layer. The system is organized around three pillars: a wiki and knowledge base populated from YouTube transcripts, articles, tweets, and podcast notes; a CRM for storing structured information about contacts met at events or on calls; and a journaling interface that queries the knowledge base in real time, surfacing relevant stored content to inform and ground AI responses to daily journal entries.
The technical build covers configuring an Obsidian vault with a RAW intake folder, setting up the Obsidian Web Clipper Chrome extension to capture full page content (including YouTube transcripts with channel metadata), and instructing Codex to process raw files in the background. Codex automatically summarizes ingested content, extracts named entities — people, companies, tools, ideas, and themes — and builds out a navigable wiki structure with cross-linked Markdown files. The demonstration shows Codex processing two source files in roughly three minutes and correctly organizing them with front matter fields including source URL, channel name, and tags.
For practitioners who have found traditional note-taking apps to be information graveyards, the video provides a concrete, replicable architecture that turns passive archiving into an interactive knowledge system. Wolfe also shows how the journal layer ties everything together — prompting the AI with current challenges and having it retrieve grounded, personalized guidance from the accumulated knowledge base rather than generating generic responses.
📺 Source: Matt Wolfe · Published May 06, 2026
🏷️ Format: Hands On Build
