Nine analyses. One clear verdict.
Trivium uses three AI models and three analytical personas to independently assess your dispute. When nine reasoning chains converge, you can trust the outcome.
How the 3×3 matrix works
Each dispute is independently analyzed nine times — not for redundancy, but for diversity of perspective.
Three distinct models
Your dispute is sent to three different language models from different providers. Each model has different training data, reasoning patterns, and biases — by design.
Three analytical personas
Each model analyzes the dispute from three jurisprudential perspectives: Strict interpretation, Consumer-Aware advocacy, and Precedent-Focused reasoning. That’s 3 × 3 = 9 independent assessments.
Consensus reveals truth
When diverse models and diverse perspectives agree, the signal is strong. When they disagree, the case has genuine complexity — and gets routed to human jurors at Kleros Court.
Transparency you can audit
Every score, every reasoning chain, every threshold is visible. No black boxes.
Each of the nine analyses produces a full reasoning chain you can read, challenge, and verify. This isn’t a score from a void — it’s a score with receipts.
When all nine analyses agree, you know the case is clear. When they diverge, the standard deviation and model spread tell you exactly where and why — persona bias, model bias, or genuine complexity.
API keys are stored locally and never leave your browser. Trivium has zero backend for the analysis pipeline — your data goes directly from your browser to the AI providers.
When the 3×3 matrix produces a split verdict, the case is routed to Kleros Court — a decentralized panel of human jurors. Ambiguity is handled honestly, not swept under a rug.