Dual-Mode Adversarial Pipeline
6-cycle rapid mode delivers battle-tested COAs in 79 seconds. 14-cycle full analysis adds OPFOR inversion, collision detection, and hardening in under 3 minutes. Each cycle builds on previous outputs through sequential adversarial refinement.
OPFOR Cognitive Engine
Models the adversary as a probability distribution across doctrinal, adaptive, and opportunistic behaviors, not a single predictable reaction.
OPFOR Attack Inversion
The system switches perspective and generates 3 independent enemy offensive plans — how the adversary would attack YOU from their positions, blind to your intentions. These are not reactions to your plan. They are the enemy's best ideas.
Collision Detection & Hardening
Automatically identifies where your plan and the enemy's plan collide in space and time. Critical collisions receive minimal hardening patches — specific fixes that protect the COA without rewriting it.
Blue-Red-Blue-Red Adversarial Loop
COAs are attacked by a reactive Red Team, then adapted. Separately, independent enemy attack plans are generated blind. Collisions between both sides are detected and patched. Plans that reach the commander have survived two layers of adversarial testing.
COA Synthesis
Builds an optimal hybrid plan by combining the strongest phases from all generated COAs. Only proposed when it genuinely outperforms individual options.
Explainable Judge
Weighted criteria scoring with evidence citations, stability analysis, fragility assessment, and commander recommendation. Every verdict is auditable.
Doctrine Compliance
Every COA scored against FM 3-90, ATP 5-0.1, JP 3-0, JP 5-0 through doctrine RAG retrieval. Non-compliant elements flagged with specific references.
Commander Style Profiles
Same scenario, different philosophy. Aggressive, cautious, deception-first, doctrine-pure. The system adapts its planning to the commander's intent.
Decision Point Forecasting
Pre-identifies critical moments where the commander must decide, with pre-calculated options and consequences for each branch.
Interactive Tactical Map
SVG-based map with drag-and-drop units, terrain drawing tools, movement arrows, risk heat maps, and fog of war visualization.
Continuous Self-Improvement
The system generates training scenarios autonomously and learns from its own best outputs. Each generation of models outperforms the last.
Distilled Specialist Cascade
Task-specific distilled models for each pipeline stage — COA generation, red teaming, and scoring. Trained via frontier distillation with 93.3% success rate across 300 validated runs. Faster and more consistent than general-purpose models.
Monte Carlo Simulation & Decision Analysis
1,000-iteration stochastic simulation with perturbation modeling. Minimax regret analysis, sensitivity tornado diagrams, and payoff matrices. Every recommendation backed by statistical confidence intervals, not single-point estimates.
Multi-Domain Operations
Same pipeline architecture runs NTC land warfare, South China Sea naval operations, Baltic littoral multi-domain, and counter-UxS scenarios. Domain-specific doctrine and adversary models swap in automatically.
Adversary Doctrine Library
Structured doctrine profiles for Russia (land and naval), China (naval), Iran (naval), DPRK (naval), and generic peer adversaries. Each profile injects realistic OPFOR behavior, force structure, and tactical patterns into the pipeline.
