How to Build Karpathy's LLM Wiki: The Complete Guide to AI-Maintained Knowledge Bases

Andrej Karpathy's LLM Wiki is an LLM-driven knowledge base built on plain markdown. It uses a three-layer architecture: raw sources, LLM-generated wiki pages, and a schema (CLAUDE.md) defining structure. The system performs three primary operations: ingest (processing new sources), query (answering questions), and lint (health checks). The LLM acts as a tireless keeper, summarizing, cross-referencing, and updating pages. Maintenance is handled by the LLM, reducing the burden of knowledge base upkeep. The CLAUDE.md file is the most critical as it defines structure and operational workflows. Obsidian is recommended for its graph view and Dataview plugin, enabling easy navigation and querying. LLM Wiki is a simpler, stateful alternative to RAG for personal/team-scale knowledge management.