LLM Wiki: Your Mind, Structured by AI
summary
LLM wiki structures chaotic sources into a searchable Obsidian-style knowledge base by applying four automated operations — group, ingest, improve, lint — converting raw bookmarks, PDFs and transcripts into standardized markdown, extracting concepts and linking pages. A prototype using web clipper and an LLM (Gemini) runs 20 queries against raw files versus the wiki, cuts token cost by more than half, fails to answer 7 raw queries but answers all from the wiki, and supports re-ingesting high-quality LLM answers to compound knowledge.
AI-written summary · about
chapters (17)
- 0:00LLM Wiki introduction
- 1:04Traditional knowledge problems
- 3:10How RAG works
- 5:15Inefficiencies of processing full documents repeatedly
- 7:24LLM Wiki workflow
- 9:11Three-step ingestion
- 12:43Standardizing wiki pages
- 14:56Querying the wiki
- 16:02Performance benefits
- 17:46Better answers with structure
- 18:48Preserving valuable responses in the wiki
- 19:52Four core operations recap
- 20:34Linting to prevent hallucinations
- 22:22LLM Wiki use cases
- 25:13RAG vs LLM Wiki
- 26:14Implementing LLM Wiki
- 26:56Closing and resources
related talks
16mLLMs are not the product
30mBuild an AI knowledge fabric for your organisation
32mTruth-seeker: Using LLM agents to build and verify knowledge bases
46mLLM—Search Engine Killer?
47mGive Your LLMs a Left Brain
33mDesigning LLM Applications for Massive Datasets
13mStrategies for Enhanced Personal Knowledge Management
37mExpanding the Capabilities of LLMs Through Community Collaboration