The collapse, four systems become one pass
You are looking at Bitcoin value-flow (the §234 collapse query)188.97M nodes · 684.24M edges in one Semurg substrate. Ask it to run graph + vector + predicate in one pass and it runs the real query live on this box, on data you can download and re-run yourself ▸.
Click an example in the agent rail, or re-run collapse. Every run is measured end to end, on this box, on the real dataset. Nothing is pre-computed.
"Transactions near this vector, matching this predicate, within K hops of this node." A conventional stack needs a vector store + a search engine + a graph database + glue, four systems, four copies, network hops. The substrate does it in one pass over one copy. Measured live.
the best-of-breed stack
NO SINGLE ENGINE ANSWERS THIS. The composite query needs a vector store AND a search engine AND a graph database AND glue to stitch them, four systems, four copies of the data, and a network hop between each.
semurg, one engine
4 systems become 1 pass. Graph hop, vector Hamming and the predicate are evaluated on one container touch each, in one pass over one copy of the data.
Four systems, four copies and three network hops collapse into a single pass.
The §234 collapse query on the live Bitcoin value-flow graph. Measured live on this box, fused vs stitched cost shown. Re-run "collapse" yourself.
ask the agent to "collapse", the one-pass composite query renders here
Honest framing: the win here is not "faster at one thing." Each leg (a vector store, a search engine, a graph database) can be fast on its own. What no best-of-breed stack avoids is the four copies and the network hops between them, and at cold or large scale that is four disk sweeps where the substrate does one. Every other category demo here is a face of this one substrate.