Andrej Karpathy proposes an AI-maintained markdown knowledge base system that replaces RAG pipelines to solve context-limit problems in AI development.

Former Tesla AI Director and OpenAI co-founder Andrej Karpathy shared a new 'LLM Knowledge Base' architecture on X to address the stateless nature of AI development

The system uses an evolving markdown library maintained by the LLM itself, eliminating the need to reconstruct context when sessions end or token limits are hit

Karpathy's approach is simpler and more elegant than traditional enterprise solutions like vector databases and RAG (Retrieval-Augmented Generation) pipelines

The LLM acts as a full-time researcher, maintaining a persistent record of projects to solve the 'context-limit reset' frustration that developers face during extended coding sessions