Notes on wikis that AI agents build and maintain
5 posts on LLM wikis, retrieval-augmented generation, and keeping a knowledge base current. These are the ideas behind the XKnow Knowledge Base — a pre-compiled LLM wiki you can own.
What Is an LLM Wiki? The Knowledge Base AI Agents Maintain
An LLM wiki is a knowledge base AI agents compile and maintain from raw sources: cross-linked Markdown, an index, and ingest, query, and lint operations.
Karpathy's LLM Wiki: How It Works and How to Build One
Karpathy's LLM Wiki pattern: an agent compiles raw sources into a living markdown wiki. The three layers, ingest, query, lint, and how to build one.
LLM Wiki vs RAG: Two Ways to Build a Knowledge Base
RAG retrieves chunks by similarity and answers fresh each time; an LLM wiki compiles linked pages once. Compare both knowledge base designs.
How to Add Embeddings to a Markdown Knowledge Base (No Code)
Add embeddings to a Markdown knowledge base without code: model choices, chunking defaults, recall tests, and a symptom-to-layer fix table.
Knowledge Base Freshness: Scoring and Automated Updates
Score and refresh an LLM knowledge base: freshness inputs, an automated lint pass, contradiction handling, and a monthly review checklist.
Skip the build — start from a wiki that holds together
The XKnow Knowledge Base is an LLM wiki you own: hundreds of linked SEO and SaaS notes with an index, a changelog, and a graph — delivered as Markdown, ready for an agent to read and extend.