Versus

Two retrieval architectures answering the same query on one slab — every edge carries a pulse and all pulses leave in phase, so a single tick of data walks both machines at once and the extra hops RAG pays are visible as distance.