A Living Knowledge Graph, Built from the Literature

This index is part of an ongoing effort to construct a continuously evolving biomedical knowledge base — inspired by Andrej Karpathy’s vision of an LLM Wiki: a structured, machine-readable encyclopedia that grows organically from primary scientific literature, powered by large language models.

Every day, the latest PubMed abstracts across seven research domains — Immunotherapy, Gene Editing, Genomics, Mitochondria, Longevity, AI & Biotech, and Drug Discovery — are automatically ingested, parsed by an LLM, and written into a PostgreSQL knowledge graph. Named entities (diseases, proteins, genes, drugs, and technologies) are extracted alongside their relationships, synonyms, and key facts, yielding a database that currently spans over 4,500 entries, 3,100 inter-entity relations, and 1,100+ source papers.

The ranking above serves as a real-time signal of research momentum: entities accumulating the most citations tend to mark the active frontiers of biomedical inquiry — areas of unresolved challenge, emerging consensus, or rapidly shifting clinical interest.