Every major crisis — Sri Lanka's default, Lebanon's collapse, the Arab Spring — was preceded by weak signals visible months in advance that no analyst caught in time. Traditional early warning systems rely on lagging indicators that confirm crises already underway, and domain-specific models miss the cross-domain interactions that produce the most dangerous cascades. Early Warning Analyst (EWA) deploys five adversarial AI agents powered by Claude Opus 4.6 with real-time web search to autonomously discover, validate, stress-test, and synthesize weak signals of emerging crises for any country. Signals surviving adversarial challenge are grouped into Constellations matched against historical pre-crisis Fingerprints and scored into risk bands. A live streaming dashboard lets analysts watch the agent debate unfold and run What-If scenarios to test hypothetical shocks against the risk landscape.
Team
Placement
Participant
