BREAKING
Karpathy shares "LLM Wiki"
How LLM Wiki works
1
Drop raw sources
↓
2
LLM builds wiki
↓
3
Cross-linked pages
Ingest, Query, Lint
1
Ingest sources
↓
2
Query the wiki
↓
3
Lint health check
LLM Wiki vs conventional RAG
Conventional RAG
NotebookLM
●
Rediscovers on every query
●
Nothing accumulates
LLM Wiki
Persistent
●
Compounding wiki layer
●
Local-first, low overhead
0
s
schema setup
0
sources in index.md
The next step after RAG
AI NEWS BLITZ
Andrej Karpathy has shared a pattern where an LLM builds a wiki from your raw sources.