Most software stores your operation as rows in tables that only that software understands. Discern stores it as a graph. Objects with fixed types, typed links between them, and facts that each carry a source, a time and a state. The vocabulary is small on purpose. A customer, an order, an asset, an event. Every industry renames the labels; the shapes do not change.
The graph sits on the appliance in open formats. Connectors propose facts into it from the systems you already run. A model may read it and propose. Only a policy or a person promotes. That one rule is what keeps a fluent guess from becoming the record.
The five pages that follow take one idea each. Objects, links, provenance, state and actions. Read them in order; each builds on the last.
Worked example · An order arrives
A system of record sends an order. It becomes an Order object with a stable id, typed attributes for the customer, the date and the items, and a link to the Asset it concerns. Nothing about it is guessed. Every attribute carries the system it came from and the time it arrived.
a worked example in generic terms; every operation renames the labels, the shapes do not change