Feedback Loop is GNI's reality validation hub — the place where predictions meet outcomes and where GNI learns from what actually happened in the world. Every MAD debate produces directional predictions (BULLISH or BEARISH) with a specific verify date, allowing GNI to score its own accuracy objectively using the GPVS standard. The GPVS (GNI Prediction Validation Standard) tracks each agent's directional accuracy across three time horizons: 7-day, 30-day, and 180-day predictions. As predictions verify over time, source trust weights are automatically adjusted — sources whose articles led to accurate predictions gain higher weight in future pipeline runs. This self-correcting feedback mechanism is what elevates GNI beyond a static intelligence system into a continuously learning autonomous platform.
All MAD agent predictions with verify dates, confidence levels, and GPVS accuracy scores — the complete prediction ledger. Each prediction is tagged with the agent that made it (Bull, Bear, Black Swan, or Ostrich) and the specific time horizon for verification. Active insights are surfaced automatically when predictions reach their verify date and reality confirms or contradicts GNI's forecast.
When predictions verify accurately — which agent was right, which was wrong, and what patterns emerge in their accuracy over time. The GPVS scorecard in full detail.
Surprise outcomes, model recalibration events, and source bias corrections. Tracks how GNI improves its prediction accuracy over time through autonomous self-correction.
When current escalation matches historical validated sequences, GNI predicts what comes next. The most advanced predictive capability in the GNI roadmap.