Desktop Commander has published a practical guide on implementing the Zettelkasten note-taking method in Obsidian. Written by Karlina Sara Rozkalne, the guide argues that Zettelkasten's format of atomic, linked ideas maps almost perfectly onto how LLM knowledge bases need to be organized. The method, created by German sociologist Niklas Luhmann, helped him produce roughly 600 publications from around 90,000 handwritten notes over 40 years. The same structure that worked for a 20th century academic happens to be what AI tools need to work effectively with your notes.
The real value is the maintenance section. Most people abandon their Zettelkasten within three months because the system gets unwieldy. At 300+ notes, you start finding orphaned notes with no links, stale Maps of Content, inbox overload, and duplicate ideas. Desktop Commander offers a fix. The AI agent can audit your inbox, find orphaned notes, and suggest connections between ideas you might have missed. Think of it as an automated gardener for your knowledge graph.
The setup is deliberately minimal. Four folders: Inbox, Literature Notes, Permanent Notes, and Templates. Notes are stored as plain markdown files on your device. No vendor lock-in. Files that will be readable decades from now. Obsidian's bidirectional links and graph view handle the connection layer.
Hacker News commenters pushed back on whether this is worth the effort for non-academic work. Some pointed out that project-oriented work may not benefit from the method, and simple categorization with tags is often enough. That's a fair criticism. But for researchers, writers, and anyone building a knowledge base that AI agents need to interact with, the structure matters. Worth reading if you're thinking about how to organize information for both human and AI consumption. Karpathy's LLM Wiki Pattern offers another perspective on organizing knowledge for LLMs.