CROWNLESS
CARDINAL ENGINE / ACTIVE  ·  ARBITER INTERFACE v0.1
CROWN
LESS

You are an Arbiter. You do not fight. You decide who does.

Field broken soldiers into Fractures. A deterministic engine resolves what you set in motion. Manage what returns.

scroll
Simulation
Deterministic
same input, same output
AI in combat
Never
heuristics only
Backend
Node.js + SQLite
server-controlled
Narrative
Groq LLM
post-simulation only
How It Works
THE
CYCLE

Seven phases. Same structure every session. What changes is the state you carry into it.

01
Evaluate

Review Stray status and available Fractures. Behavioral flags determine viability. Assign roles.

02
Call

Optionally run the Calling. Candidates carry partial profiles. You learn what was hidden once they are in a Fracture.

03
Log State

Before dispatch, each Stray's drift, morale, fatigue, and complacency are captured as simulation inputs.

04
Enter

The Fracture initializes. Partial intel returns. The deployment window closes.

05
SimulateENGINE

Cardinal Engine runs. No input from you. The log fills. You wait.

06
Resolve

Outcomes distribute. Resources, XP, narrative fragments, state changes per Stray.

07
Recover

Strays return to Bastion. Facilities determine what heals and what compounds into the next cycle.

Behavioral System

Four Variables. One Decision Function.

DRIFT0–1.0

How far a Stray has diverged from their assigned baseline. High drift restructures the decision tree. Role assignments deprioritize.

MORALE0–100

Psychological state. Below 30, retreat thresholds rise and aggressive actions deprioritize. Above 80, a performance multiplier applies.

FATIGUE0–100

Accumulates each Fracture. Does not clear without Bastion rest. High fatigue degrades reaction scoring — the timing heuristic.

COMPLACENCY0–1.0

Builds from repetition. Complacent Strays apply cached evaluations. They fail systematically against novel compositions.

LLM handles
+Stray dialogue and personality
+Post-mission narrative fragments
+World event flavor text
+Character memory generation
+Rare dynamic world events
LLM never touches
×Combat target selection
×Action evaluation or scoring
×Threat prioritization
×Progression or reward calculation
×Economy or resource logic
Cardinal Engine / Design Philosophy
"Systems first. Content second. No dice in the critical path. No language model deciding outcomes. The engine is honest. What you build into it is what you get back."