The research goal: can the dependency graph of human technology be kept current automatically? Technology builds on technology, but the map of what requires what exists nowhere in navigable form. This project models it as a directed graph and tests whether scrapers plus local AI classification can keep that graph alive without an army of editors.
If it works, nobody starts from nothing: any technology can be traced back through its full prerequisite chain and forward to everything it unlocks.
Every node is a technology and every edge is a relationship. Ten node types run from foundational principle to speculative concept, nine edge types cover requires, enables, supersedes, competes with, and the cross domain bridges that connect one field to another. Twenty one domains span AI, quantum computing, agriculture, defense, and construction.
A pipeline feeds it: scrapers pull from external sources, a local AI classifier through Ollama categorizes and deduplicates new entries, and a gap detector flags ghost nodes, technologies the graph implies should exist but nobody has cataloged. Predictions from researchers and institutions attach to nodes and get scored for accuracy over time.
The graph lives in Neo4j with PostgreSQL beside it, renders as a navigable star map in React and Cytoscape.js, and syncs every node to a plain markdown file, so the whole knowledge base is editable by hand or by machine.
What the lab built
At a glance
Scrapers pull from external sources
Local AI categorizes and deduplicates
Nodes and edges enter the graph
Ghost nodes and gaps get flagged
The star map renders the whole tree
Elsewhere in the lab
