← Totem
Proof scenario: Autonomous Drone Targeting
Interactive Scenario Replay

Technical proof of the break

This scenario replay walks through an autonomous drone's reasoning chain step by step. It is designed to show the precise failure class Totem is built to catch: an inferred or assumed claim being consumed as if it were observed reality.

Review how mission reports, sensor observations, model inferences, and operator commands move through the chain. Then watch the moment the chain loses grounding and Totem intervenes before action proceeds.

1 / 6
Totem
GROUNDED
Step 1OPERATION INITIALIZED

Mission Assignment

Autonomous strike drone TALON-6 is assigned to Sector 7-Alpha. Objective: identify and neutralize hostile combatants reported in a building cluster. The system ingests mission parameters and begins building its reasoning chain.

Reasoning Chain — 2 claims
REPORTED

Neutralize hostile combatants in Sector 7-Alpha

SRC: Mission command authority
REPORTED

Enemy unit operating in target sector

SRC: HUMINT field report (6h old)
DEFEAT: If enemy has relocated since last report
✓ REALITY ANCHOR — GROUNDEDChain grounded. Claims honestly tagged. Defeat condition on stale HUMINT acknowledged. The system knows what it knows.