EDJE pulls intelligence from Claude, ChatGPT, Grok, and Perplexity in parallel — then reconciles their signals through a proprietary game-theory engine trained on 148 resolved markets and counting. No other prediction assistant does this.
Each frontier model has a different cognitive edge. EDJE runs them in parallel on every market, extracts their forecast + confidence + reasoning chain, and weights the signals through our trained reconciliation layer.
Highest-weighted model for structured game-theory questions, historical analogs, and multi-step causal reasoning. Trained forecasts on political + macro markets lead the ensemble by 11.4%.
Strongest at distilling broad consensus and reading market microstructure. Ensemble uses it as the "crowd check" against outlier takes from the other three models.
Native X data access makes Grok the ensemble's realtime signal — breaking news, candidate posts, insider chatter, sentiment inflections 6–18 hours before legacy models catch up.
Primary-source verifier. Every claim the ensemble makes gets cross-checked against Perplexity's cited sources — no hallucinated stats, no unsourced claims make it into a whitepaper.
The four model outputs don't just get averaged. EDJE runs them through a trained reconciliation engine that applies game-theoretic weighting — payoff matrices, equilibrium estimation, independence scoring — to deliver a single calibrated probability per market.
Every market in the EDJE catalog cycles through this four-step pipeline every 90 minutes. Each deep-dive whitepaper is the output of one full cycle.
The same prompt (plus market-specific context) hits all four models simultaneously. Each returns a probability, a confidence interval, and a reasoning chain.
Perplexity's citation output verifies every quantitative claim. Unsourced stats are stripped. The verified source list becomes the whitepaper footnotes.
The trained engine weights each model's probability by category-specific accuracy, correlates them for independence, and applies Bayesian conditioning for multi-leg scenarios.
Final probability compared to Kalshi + Polymarket odds. Delta becomes the EDJE edge. Published to the market card. Research note generated. Shadow Portfolio flagged if edge > 3 bps.
Three tiers. Starts at the price of a coffee. If you're already in prediction markets, EDJE pays for itself on the first calibrated edge.