Glossary

Every term used in the product, defined once. Skim this when you hit a word you do not know. For the walkthrough, see Getting Started.

#Persona

The main output of the Diagnostic. A markdown file that describes how an organisation decides: its principles, win conditions, red lines, and how it reads others. Load it over any language model and the model reasons like that organisation. Written from the underlying fingerprint.

#Strategic fingerprint

The score layer beneath the persona. The Diagnostic’s reading of an organisation across the strategic Ingredients, with its coherence and cluster proximity. The numbers the persona is written from; in day-to-day use you work with the persona, not the raw fingerprint.

#Counterfactual

A “what if” question. The simulator lets you fork a scenario at any tick to run an alternative timeline against the main one.

#Cluster

A family of archetypes. Five families: conservative, aggressive, culture, system, independent. The Diagnostic places you nearest one (or two, hyphenated).

#Three measurement registers

The three question types in the Diagnostic. Preference (sliders, stated position). Observed behaviour (multi-select, what you actually see). Forced choice (short scenarios, what you do when you cannot do both).

#Engine primitives

The building blocks of every Simulate scenario: Actor (player), Resource (something they spend or accumulate), Capability (a flag that gates actions), Action (what they submit each tick), Tick (one unit of game time), Phase (within-tick sub-step), Resolver (function that turns submissions into outcomes), External event (scheduled exogenous shock).

#Four resolver classes

The four resolution mechanics the engine ships with. allocation_share (budget allocated across categories; share is proportional). arbiter (Claude or the operator narrates the outcome). matrix_payoff (game-theory matrix; pair-wise strategies). auction (sealed-bid; highest bidder wins).

#Event-sourced history

The engine never stores game state directly. It stores a log of events; any state is computed by replaying the log up to a tick. This is why forking, rewinding, and exporting an audit trail are cheap.

#Tick

One unit of game time. Atlas Regional has six quarterly ticks (Q1 to Q6); other scenarios use monthly, weekly, or event-driven ticks. Tick zero (T+0) is the starting state, before anything has been played.

#Phase

A sub-step within a tick. Most scenarios have one phase per tick (e.g. allocation). Some have several (e.g. negotiation, then arbiter resolution, then payoff).

#Branch

A named timeline. main is the default branch every scenario starts on. fork-1, fork-2 etc are counterfactual branches created by clicking Fork at a tick.

#Twin briefing

The persona in its system-prompt form, used when an agent plays an actor in a wargame or answers in the twin chat. Same document as the persona; “twin briefing” is the older name. Contains the overview, principles, win conditions, red lines, and posture notes. Produced by the Diagnostic; can also be authored or edited directly in Studio.

#Arbiter

The decision-maker for arbiter-class phases. Defaults to Claude; the operator can override. Trust mode (ask / session / always) controls how often the operator is asked to confirm.