Business Lunch Building an AI Chief of Staff with Obsidian and Claude Code
May 12, 2026
They explore turning note-taking systems into an AI chief of staff that stores memory locally and acts on workflows. Discussion covers plain-text markdown, Obsidian’s linked knowledge graph, and using Claude Code for transparent, editable memory. They cover automating setup with a single prompt, adding CLI “hands” for real operations, and the security trade-offs of autonomous agents.
AI Snips
Chapters
Books
Transcript
Episode notes
Store Context In Local Plain Text
- Store your business context locally as plain-text markdown to eliminate translation overhead.
- Use Obsidian to keep files on your hard drive so LLMs read raw text instantly without slow database/API conversions.
Markdown Is The LLM Native Format
- Plain-text markdown aligns with how LLMs process information, reducing parsing overhead and latency.
- Ryan Dice explains Notion requires a translation layer and database queries every time an AI reads content, adding friction.
Networked Notes Simulate A Chief Of Staff
- Obsidian's wiki-style linking creates a networked knowledge graph that an AI can traverse like a human chief of staff.
- The graph view maps notes as nodes and links, letting the AI follow preferences, past work, and project links in milliseconds.
