LLM Wiki Turns Karpathy's 16M-View Idea Into an Open-Source Desktop App
LLM Wiki turns Karpathy's viral knowledge-base pattern into a full desktop app — compiling your documents into a persistent, interlinked wiki instead of re-answering from scratch every time.
- What it is: LLM Wiki is a free, open-source desktop app that compiles documents into a persistent, interlinked wiki instead of re-answering from scratch like RAG.
- Origin: Based on Andrej Karpathy's viral LLM Wiki pattern (16M views), which proposed treating knowledge compilation like software compilation.
- Two-step ingest: Documents go through an analysis pass then a generation pass, producing entity/concept pages with full source traceability via YAML frontmatter.
- Knowledge graph: A 4-signal relevance model (direct links, source overlap, Adamic-Adar, type affinity) plus Louvain community detection surfaces hidden connections and knowledge gaps.
- Broad support: Works with OpenAI, Anthropic, Google, Ollama; handles PDF, DOCX, PPTX, XLSX; includes a Chrome Web Clipper and Obsidian vault compatibility.
- Key risk: LLM hallucinations can propagate across linked wiki pages; the built-in lint pass and source traceability are the main mitigations.
In April 2026, Andrej Karpathy published a post on X describing a new way to manage personal knowledge with LLMs, along with a GitHub Gist laying out the full architecture. The pattern , a three-folder markdown setup where an AI model compiles, maintains, and queries a structured knowledge base without a vector database , generated 16 million views on that single post. The Gist accumulated 5,000 stars and 485 comments within days. Now, LLM Wiki by nashsu takes that abstract design and ships it as a ready-to-run desktop application.
The problem RAG never solved
Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time.
The analogy Karpathy reaches for is from software engineering: compilation. When you write source code, a compiler transforms it into an optimized binary artifact. You don't execute the source code directly every time , you compile once, distribute the artifact, and run it efficiently on demand. The compilation step is expensive, but it pays for itself across every subsequent use.
The LLM Wiki pattern says: compile your knowledge first. Use an LLM to read your sources, synthesize their contents into structured, interlinked wiki pages, and then run all your queries against that compiled artifact. The synthesis happens once (and incrementally as sources are added); the queries benefit from pre-organized, cross-referenced knowledge every time.
From gist to desktop app
This project is based on Karpathy's LLM Wiki pattern , a methodology for building personal knowledge bases using LLMs , implemented as a full desktop application with significant enhancements. Karpathy's pattern inspired a wave of open-source tools, with at least seven GitHub projects garnering over 3,000 combined stars in less than a week. LLM Wiki by nashsu is the most fully-featured of these implementations, with over 11,000 stars.
The app is built on
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
