How it was built
Each exploration began as a single research question and was developed into a weighted knowledge graph: concepts as nodes, relationships as edges, every claim carrying its sources. Crucially, each new exploration was conditioned on the corpus that already existed — so the graph has cross-references a cold-start query cannot reproduce. The result was then read back as a whole, which is how the five cross-sector patterns were found: they live in the connections between explorations, not inside any one of them.
What's free here
All of it. PlexusGraph is a publication, not a product with a locked door. The point of view and the curation are the work; the data is meant to be used. So the merged graph is downloadable, and every exploration ships with an LLM-ready context pack you can drop straight into your own AI.