THE ANALYTICAL ENGINE
CORVUS ingests data from 25 institutional sources across 217 countries, processes it through a 10-phase analytical pipeline containing 7 proprietary algorithms, and produces mechanism-based risk assessments, probabilistic forecasts, contradiction detection, and irreversibility analysis.
It is not a news aggregator or a chatbot wrapper. It is a quantitative analytical engine that models geopolitical risk through state vectors, phase transitions, eigenvalue decomposition, Bayesian networks, and information entropy.
7 PROPRIETARY ALGORITHMS
Each algorithm produces a 0-1 risk score with confidence, plus Tier 1 features: structured sub-outputs unique to that algorithm. All 7 feed into a second-pass interconnection fabric.
Cascade Fragility Index
Models nations as nodes in a weighted dependency network (trade 30%, debt 25%, energy 25%, military 12%, diaspora 8%) and computes how geopolitical shocks propagate through the system.
A 0-100 score measuring each country's capacity to absorb shocks before they cascade to neighbors. Includes Monte Carlo simulation (200 runs per country across 10 shock magnitudes).
The only algorithm that models international systemic risk as network-level contagion.
Temporal Entropy Divergence
Measures information environment turbulence across 6 streams: event entropy, narrative divergence, media attention anomaly, theme shift, emotional intensity, and Wikipedia public attention.
Classifies each country's information environment as CONSENSUS, CONTESTED, NOISY, or FRACTURED. Measures within-bloc vs. between-bloc divergence across 7 information blocs.
Detects when facts are contested and narratives are fragmenting. The information-warfare signal.
Economic Stress Tensor
Models economic vulnerability as a 5-dimensional stress tensor (currency, debt, trade, capital, commodity) with eigenvalue decomposition to identify the dominant failure mode.
Identifies whether a country faces Currency Crisis, Debt Crisis, Trade Shock, Capital Flight, or Commodity Collapse. Computes mode transition probability and time-to-binding in days.
Uses advanced linear algebra to predict crisis type transitions before they occur.
Regime Phase Transition Detector
Models governance stability as a dynamical system. Detects critical slowing down, variance amplification, and flickering: early warning signals that appear 6-12 months before regime collapse.
12-observable state vector with autocorrelation-based critical slowing down detection. Variance amplification threshold at 1.5x. Calibrated against 10 historical regime transitions.
Applies complexity science, not expert opinion, to regime stability.
Bayesian Conflict Onset Network
Estimates P(conflict | observable signals) with 4-dimensional commitment threshold detection (military, rhetorical, political, diplomatic). Supports counterfactual conditioning via causal graph surgery.
Tracks actor commitment across 4 dimensions with 5-tier classification. Hidden Markov Model state transitions: STABLE / ESCALATING / CRITICAL / CONFLICT / RECOVERY.
The only system that supports counterfactual analysis: 'what if we remove this signal?'
Linguistic Escalation Gradient
Analyzes diplomatic language (0-35 pts), bilateral actions (0-30 pts), media rhetoric (0-20 pts), and travel advisories (0-15 pts). First systematic quantification of rhetorical escalation.
Computes when actors can still safely de-escalate vs. when they are locked into confrontation. Multilingual keyword detection across 10 languages.
Quantifies the gradient from diplomatic euphemism to explicit threat, independent of military action.
Predictive Risk Oracle
Fuses structural trajectory, momentum/velocity, precursor pattern matching, and causal trigger modeling into probabilistic multi-horizon forecasts at T+7, T+30, T+90, and T+180 days.
Detects when a country's future genuinely splits into multiple plausible paths. Identifies the specific trigger variable that determines which path materializes.
The only algorithm that explicitly models fork points in geopolitical futures.
THE IRREVERSIBILITY ENGINE
Answers the most critical question in geopolitical risk: “Can this situation still be recovered, or has it passed the point of no return?”
Irreversibility is NOT about risk level. A country at score 60 that has exhausted its policy tools, closed all diplomatic exits, and locked actors into committed positions is worse than a country at score 80 with reserves, open channels, and flexible actors.
The first is irreversible. The second is recoverable.
UNIFIED INTERPRETATION
Three components consume all upstream outputs and produce unified interpretation. Read-only: they never modify algorithm outputs.
Compresses all 7 algorithms + Tier 1 features into a single state classification per country. 16-state taxonomy across 7 families: STABLE, STRESSED, CRISIS, TRANSITION, IRREVERSIBLE, AMBIGUOUS, RECOVERY.
Detects hidden divergences between algorithm signals. 6 contradiction types, each mapped to historical precedent.
Uses temporal dynamics (velocities and accelerations across all algorithms) to classify movement type. Includes hysteresis to prevent classification flicker.
6 ANALYTICAL FRAMEWORKS
LLM integration with conditional routing based on each country's system state. Not a bolted-on chatbot, but structured analytical frameworks with specific triggers and outputs.