Recursive CTE demo: WITH RECURSIVE, one pass not n passes
You are looking at Bitcoin value-flow graph1.8M nodes · 7.37B value-flow edges in one Semurg substrate. Ask it to expand the recursive CTE as a k-hop fund-trace 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 change the depth (trace 5, trace 7), or the seed tx. Every run is measured end to end, on
this box, on the real dataset. Nothing is pre-computed.
The live Bitcoin value-flow graph stored as 64-byte containers. Following a coin through the chain, fund after fund: the same WITH RECURSIVE walk. We run it as a k-hop BFS, the edges scanned once, each hop one frontier, no per-level table re-scan, no UNION-de-dup materialized. Measured live.
the recursive-CTE SQL leader
Re-scan every level. WITH RECURSIVE re-joins the base table once per level and materializes a UNION working set it de-duplicates each iteration, so a deep coin trace pays for the same value-flow rows again and again.
semurg, one engine
Parents scanned once. The walk is a k-hop BFS: each level is one frontier off a single resident pass, zero CTE re-scans, no UNION spool to de-dup.
The per-level table re-scan does not get faster, it is gone.
Live Bitcoin value-flow graph on this box, parsed straight off the P2P network. Watch the CTE re-scan count read 0. Re-run it yourself.
ask the agent to "trace 5" (a 5-hop fund-trace) or "trace 7" (a deeper coin follow)
Honest caveat: the win here is raw, not only collapse. Because the value-flow edges are traversed exactly once and each frontier feeds the next, depth tracks the distinct rows reached, not the number of passes. Cold and warm are both shown, tier for tier, never our-hot against their-cold, and the same substrate also serves key-value, full-text and vectors.