Six months later, a regulator asks why Susan bound a $4.2M risk on March 5. In most stacks, that's a six-week reconstruction. In InvariantDB, it's one walk of the decision graph. Every step is a Cypher query you can run yourself.
Most enterprise systems record the final answer (BIND $4.2M) and lose everything that produced it. InvariantDB stored the full trail — the context Susan saw, the rules that fired, the documents she opened, the reasoning she wrote, the signature chain. Every step is a node; every causal link is an edge; every value carries a bitemporal stamp.
AT RECORDED '2026-03-05T14:32Z' reconstructs exactly what Susan saw. Not today's view. The view at the moment she bound.
db.verifyAclChain() rehashes every entry from genesis. The receipt isn't trusted because we say so — it's cryptographic.
derivedFrom edges trace the rate quote back to: filed rates × CAT score × sprinkler credit × loss-history factor. Every input recorded.
This is what "build agents from decisions, not documents" looks like in practice. Replay Susan's last 800 bindings to train an underwriting assistant that learned how she actually thinks, not just what's in the SOP.