Glossary
What is LLM Wiki?
An LLM Wiki is a personal or team knowledge wiki structured for LLMs to read. Coined by Andrej Karpathy in 2024, the term describes a markdown-first, AI-readable wiki that gives LLMs context for tasks ranging from coding to writing to operations. The methodology emphasizes file-first knowledge that lives on disk in plain text, organized so an LLM can navigate it like a human navigating a wiki.
Last updated:
Origins
Where the term comes from
Andrej Karpathy described the LLM Wiki pattern in a 2024 GitHub gist and discussion that circulated through Hacker News and the LLM developer community. The core idea: instead of fragmenting your context across chat histories and per-tool memory, treat your knowledge as a wiki of markdown files that any LLM can read, navigate, and reason over. The methodology is open source and tooling-agnostic, which is part of why it spread quickly.
Capabilities
What LLM Wiki does
Personal AI context
Gives a single user or small team a markdown-based knowledge layer that any LLM can read.
File-first structure
Knowledge lives as plain markdown files on disk, not inside a vendor's database or cloud API.
Open methodology
No vendor, no proprietary system. Anyone can adopt the pattern using existing tools.
AI-readable navigation
Wiki-style links and structure let an LLM follow context across files the way a human follows links across pages.
Distinctions
LLM Wiki vs adjacent concepts
LLM Wiki is often confused with related but distinct ideas. Here is how it differs.
| Concept | What it is | How LLM Wiki differs |
|---|---|---|
| Context OS | A personal or single-team wiki for LLMs. DIY methodology. | A maintained system built on the same file-first principles, with a typed knowledge graph, MCP server, role-based access, and team layer out of the box. |
| Notion / Confluence | Document workspaces optimized for human reading. Files behind a cloud API. | Files canonical on disk. Markdown-first. No cloud-API tax. Designed for LLM consumption from day one. |
| Traditional wiki (MediaWiki, etc.) | Pages for humans, indexed for full-text search. | Files for both humans and LLMs, structured for AI navigation and retrieval. |
Who uses it
Who uses LLM Wiki
Solo developers, small technical teams, and founder-operators experimenting with personal AI workflows who want full control over their knowledge layer and have engineering time to build on top of the methodology.
FAQ
Common questions about LLM Wiki
Where does the term LLM Wiki come from?
Andrej Karpathy described the pattern in a 2024 GitHub gist that circulated through Hacker News and the LLM developer community. The core idea is to treat personal and team knowledge as a wiki of markdown files that LLMs can read and navigate.
How is an LLM Wiki different from a Context OS?
An LLM Wiki is an open methodology you build yourself, typically for personal or small-team use. A Context OS is the system that operationalizes the same principles for a company: typed graph, MCP server, role-based access, source attribution, and AI tool integration as a maintained product.
Can I build an LLM Wiki for my company?
You can, and many technical teams have. The work involves designing your own file structure, building your own retrieval and query layer, integrating with your AI tools (Claude, ChatGPT, Cursor, etc.), and maintaining the system as the team grows. A Context OS like DearTech-OS handles all of this as a productized layer.
Related terms
Keep going
See LLM Wiki in practice
DearTech-OS is a Context OS for founder-operators. Explore the product or talk through whether one is right for your team.