# karpathy-llm-wiki: Build a Karpathy-Style LLM Knowledge Base

This repository profile is provided by osrepos.com, an open source repository discovery platform.

Source: osrepos.com
Repository profile: https://osrepos.com/repo/astro-han-karpathy-llm-wiki
Generated for open source discovery and AI-assisted research.

karpathy-llm-wiki is an Agent Skills-compatible tool that implements Karpathy's LLM Wiki idea, allowing users to build a durable knowledge base. It enables LLMs to maintain structured wiki pages by ingesting sources, compiling knowledge, and answering questions with citations. This project offers a robust alternative to traditional RAG for compounding knowledge over time.

GitHub: https://github.com/Astro-Han/karpathy-llm-wiki
OSRepos URL: https://osrepos.com/repo/astro-han-karpathy-llm-wiki

## Summary

karpathy-llm-wiki is an Agent Skills-compatible tool that implements Karpathy's LLM Wiki idea, allowing users to build a durable knowledge base. It enables LLMs to maintain structured wiki pages by ingesting sources, compiling knowledge, and answering questions with citations. This project offers a robust alternative to traditional RAG for compounding knowledge over time.

## Topics

- llm-wiki
- knowledge-base
- agent-skill
- karpathy
- python
- productivity
- rag-alternative
- ai

## Repository Information

Last analyzed by OSRepos: Wed Sep 09 2026 12:13:25 GMT+0100 (Western European Summer Time)
Detail views: 1
GitHub clicks: 2

## Safety Notice

OSRepos shares public repositories for knowledge and discovery only. Review source code, dependencies, licenses, and security implications before running or installing anything.

## Content

## Introduction

**karpathy-llm-wiki** is an open-source project by Astro-Han that brings Andrej Karpathy's innovative LLM Wiki concept to life. This repository provides a reusable skill compatible with Agent Skills tools like Claude Code, Cursor, and Codex, enabling you to build and maintain a dynamic, LLM-driven knowledge base. It allows your coding agent to ingest raw sources, compile durable knowledge pages, answer questions with citations, and lint the wiki for consistency. This project effectively packages Karpathy's LLM Wiki idea into an installable skill for enhanced knowledge management.

## Why Use It & Benefits

The `karpathy-llm-wiki` offers a compelling alternative to traditional Retrieval-Augmented Generation (RAG) systems, focusing on compounding knowledge over time. Instead of re-searching raw documents for every query, an LLM wiki maintains structured, curated markdown pages. Knowledge synthesis happens during ingest and maintenance, leading to durable cross-links and evolving summaries.

Key operations include:

*   **Ingest**: Collects source material into a `raw/` directory, triages it, and then creates or updates relevant wiki articles in `wiki/`.
*   **Query**: Searches the compiled wiki pages to provide grounded answers, complete with citations linking back to your markdown knowledge.
*   **Lint**: Performs health checks on the wiki, verifying index integrity, links, and overall consistency, with auto-fixes for common issues.

This approach ensures that your knowledge base improves and grows smarter with each new piece of information. The workflow is based on a production knowledge base, demonstrating its readiness with 94 wiki articles and 99 source materials maintained daily since April 2026.

## Installation

To get started with `karpathy-llm-wiki`, you can easily install it using `npx`:

bash
npx add-skill Astro-Han/karpathy-llm-wiki


This skill is designed to work seamlessly with any tool that supports the [Agent Skills open standard](https://agentskills.io){:target="_blank"}.

## Examples

The quick start guide outlines three core interactions with your LLM wiki:

1.  **Ingest your first source**: Provide a URL, file, or pasted text to the skill.
    
    "Ingest this article: https://example.com/attention-is-all-you-need"
    
    The skill processes the source, stores it in `raw/`, and then compiles or updates the relevant knowledge pages in `wiki/`.

2.  **Ask your wiki a question**: Query your knowledge base for information.
    
    "What do I know about attention mechanisms?"
    
    The skill searches your curated wiki and provides answers, citing the markdown pages where the knowledge resides.

3.  **Keep the wiki healthy**: Regularly lint your wiki to ensure its integrity.
    
    "Lint my wiki"
    
    This command checks for broken links, missing index entries, stale cross-references, and other related issues, helping to maintain a robust knowledge base.

## Links

*   [karpathy-llm-wiki GitHub Repository](https://github.com/Astro-Han/karpathy-llm-wiki){:target="_blank"}
*   [Agent Skills Standard](https://agentskills.io){:target="_blank"}
*   [Karpathy's LLM Wiki Idea](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f){:target="_blank"}