A wiki that the model maintains itself

14 June 2026 Michael Engel Prompting Automation

Instead of giving an AI the same documents again and again, you have it write a wiki from them once and maintain it.

In depth: Knowledge base for agents – the LLM wiki after Karpathy

Andrej Karpathy describes the “LLM Wiki” in a freely readable text: raw sources that are never changed, wiki pages that the model writes from them and keeps up to date, and a rules file such as the CLAUDE.md (externally checked). New material isn’t just filed away but worked into the existing pages, and contradictions are flagged. Jack Roberts shows the setup with Claude Code (14 June). AI LABS adds how it stays manageable with many files: one topic per file, an index.md per folder, and Claude reads the indexes first (26 June). According to Karpathy, this holds about 100 sources and a few hundred pages.

Tool and alternative

Claude Code and a folder are enough; Obsidian for viewing is optional. Hermes from the video is not needed. The video title “Google Replaced The Karpathy’s LLM Wiki” exaggerates: Google’s Open Knowledge Format organizes the pattern, it doesn’t replace it.

Sources

Andrej Karpathy, “LLM Wiki”, April 2026, retrieved 11 Oct 2026
Jack Roberts, “Build a Hermes Knowledge Base That Self-Improves”, 14 June 2026 (3:49–6:31)
AI LABS, “Google Replaced The Karpathy’s LLM Wiki”, 26 June 2026 (5:20–5:52, 10:11)

Advertising in the source: Jack Roberts with self-promotion (community). AI LABS sponsored by Mobbin.

Want to try this with your team? That’s what the AI Lab is for. See the AI Lab