Claude-Wiki

A self-maintaining knowledge OS — the public system layer of a personal LLM-wiki, and the companion to the paper on horizontal scaling.

---

Project maintained by dumbbutt0 Hosted on GitHub Pages — Theme by mattgraham

📄 Read the paper  ·  🔒 Privacy model  ·  💻 Source on GitHub

A self-maintaining knowledge OS

This is the public system layer of a personal LLM-wiki — a persistent, self-visualizing, self-maintaining knowledge base maintained by Claude Code, following Karpathy’s LLM-wiki pattern (raw sources → an LLM-maintained wiki → a schema that governs it).

It’s the companion to the paper on horizontal scaling — the idea that a compute-constrained builder gains capability by composing systems (memory, tools, graphing, privacy, governance) around a model, rather than by training a bigger one. This site is that claim, running.

Publish the method, not the work. Everything here is the system and the pages about the system — never the owner’s private knowledge, sources, conversations, or personal graph. See the privacy model.

Explore

• The paper — Horizontal Scaling in AI Systems
• How the privacy model works
• The full system + code on GitHub

---

_{Built and maintained with Claude Code · code under MIT, writing under CC BY 4.0.}