Trace what a machine knew, why it acted, and what happened afterward.
A verifiable graph connecting external evidence to machine decisions, handoffs, actions, outcomes, reviews and corrections. Everything on this page is sanitised or synthetic demonstration data. References in; verifiable evidence out.
The explorer.
Select a decision (in any view), and this view shows the shortest verifiable path back to its evidence: decision → receipt → observation → source, with the integrity state of every hop.
| Type | Label | Created | Integrity |
|---|
Trust boundary
A verified graph relationship proves that the referenced evidence and identifiers satisfy the implemented integrity checks. It does not, by itself, prove objective truth, completeness, safety, causation, that every relevant event was observed, or that the resulting decision was correct.
Integrity (bytes unaltered), attribution (which key signed), provenance (which source published), freshness (how old at use), corroboration (who else agreed), reported action (what a system says it did), independently observed outcome (what a third party saw), completeness (what was not captured), causation (what actually caused what) and objective truth are ten different things. The inspector states which of them each record establishes.
Tamper laboratory.
Each button copies the loaded demonstration graph, changes exactly one thing, and re-runs the structural and integrity checks on the copy. You see which check fails, which relationship is no longer supported, and which downstream decision is affected. The original fixture is never altered.
Where to next.
Developer documentation · verify a receipt · the product tour · bring us one AI decision · fixture manifest.