Summary
Personal knowledge management (PKM) at scale becomes tractable with LLMs: search over your notes semantically, summarize collections, extract connections you missed. The trick is treating AI as a query layer over your data, not a replacement for your thinking.
Objectives
- 01Design an AI-augmented PKM system.
- 02Distinguish query use from generation use.
- 03State the failure mode of PKM without discipline.
The Lesson
The stack
Notes in Markdown (Obsidian, Logseq, or plain files). Embedded with an embedding model. Indexed in a local vector store (pgvector, LanceDB, Chroma). Queried through a chat interface that shows sources. The AI never writes to your notes without your review.
The discipline
PKM decays without curation. AI can help retrieve and synthesize; it cannot decide what is worth keeping. The productive pattern: write notes yourself, use AI to search, summarize, and connect — never delegate the writing itself.
Key Ideas
- AI is a query layer over your notes, not a writer of them.
- Curation is human work; retrieval and synthesis can be delegated.