LLM Wiki: Let Your Agent Keep Its Own Notes (and When Real Memory Wins)
Karpathy's LLM wiki lets an agent maintain its own markdown notes. I ran one against octobrain memory on the same sources. Here's where each one wins.
This article discusses Andrej Karpathy's LLM wiki concept, a method for maintaining an AI-powered knowledge base with interlinked markdown pages, and compares it with Octobrain memory for managing AI agent knowledge.
What happened
The article describes the author's testing of Andrej Karpathy's LLM wiki concept and Octobrain memory for managing AI agent knowledge. The author set up both systems, tested them with the same sources, and compared their performance in handling contradictions and retrieving information.
Why it matters
For AI developers and technology professionals, understanding the trade-offs between different knowledge management systems is crucial for optimizing AI agent performance and maintaining accurate information. The article highlights the importance of conflict resolution and efficient information retrieval in AI systems.
- Use an LLM wiki for human-readable, maintainable knowledge bases with automatic conflict resolution.
- Opt for Octobrain memory for faster lookups and efficient fact retrieval across large datasets.
- Implement conflict resolution mechanisms to ensure accurate information in AI systems.
- Combine both systems for different use cases: LLM wiki for human reading and Octobrain for agent memory.
- Maintain version control with git to track changes and revert if necessary.
The bigger picture
The article provides insights into effective knowledge management strategies for AI agents, helping developers choose between different systems based on their specific needs. It emphasizes the importance of maintaining accurate and up-to-date information in AI workflows.
The article compares Andrej Karpathy's LLM wiki and Octobrain memory for managing AI agent knowledge, highlighting their strengths and weaknesses. It suggests using the LLM wiki for human-readable knowledge bases and Octobrain for faster, more efficient agent memory, depending on the specific needs of the project.
Technical details
- LLM wiki: A knowledge base of interlinked markdown pages maintained by an AI agent.
- Schema file: Defines operations (ingest, query, lint) and conflict resolution rules.
- Octobrain memory: A memory server that indexes documents and stores distilled facts with relationships.
- Conflict resolution: The LLM wiki automatically flags contradictions, while Octobrain requires manual 'supersedes' edges.
- Obsidian: A markdown editor used for viewing and navigating the wiki.
FAQ
What is an LLM wiki?
An LLM wiki is a personal knowledge base of interlinked markdown pages that an AI agent writes and keeps current from source documents you curate. It was described by Andrej Karpathy in an April 2026 gist, involving raw sources, an agent-owned wiki, and a schema file with ingest, query, and lint operations.
Is Karpathy's LLM wiki better than RAG?
The LLM wiki is better when a human reads the result and the collection is moderate, with a comfortable range of around 100 sources and hundreds of pages. Retrieval over indexed documents (RAG) is better for fast lookups across large or growing collections.
Can I use an LLM wiki with Obsidian?
Yes, that's how Karpathy runs it: the agent in one window, Obsidian on the same folder in another. Obsidian provides the graph view, Web Clipper for turning web articles into markdown sources, and plugins like Dataview for querying page frontmatter.
Is an LLM wiki a good AI second brain?
For knowledge you want to read and think with, yes. It's the most transparent AI second brain setup, with the agent handling maintenance. For an agent's working memory during tasks, a memory server is faster and searches by meaning across everything.
What is the main advantage of the LLM wiki over Octobrain memory?
The LLM wiki automatically flags contradictions during ingestion, while Octobrain requires manual 'supersedes' edges to manage conflicting facts. The wiki is also more portable and human-readable, using plain markdown files.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by Zentask Blog on Oct 13, 2026.
Comments
Sign in to join the discussion.