Our Data — Crene Prediction Intelligence Platform
We run 4 frontier AI models on earnings, macro releases, central bank decisions, and market events. We track which models get it right and make everything publicly verifiable.
4 AI models independently forecast every active event. Revenue beats, EPS surprises, macro data releases, and central bank decisions — all with structured resolution criteria.
View →AI-powered earnings intelligence. 4 models predict beat/miss/meet for every major earnings report before Wall Street knows. Verified against actual results post-earnings.
View →Which AI model is most accurate? Per-model accuracy tracked across all resolved predictions and categories. Full transparency — every prediction is public and auditable.
View →| Step | Process | Details | Frequency |
|---|---|---|---|
| 01 | Detect Events | Scan upcoming earnings, macro releases, central bank meetings, and market events. Generate structured binary questions with resolution criteria | Continuous |
| 02 | AI Forecasting | GPT-4o, Gemini, Claude, Grok independently evaluate each question. Each produces a probability + reasoning | 6x daily |
| 03 | Spread Analysis | Compute model agreement and spread. Flag high-spread predictions where AI models disagree on outcomes | Real-time |
| 04 | Auto-Resolution | When markets resolve, compare model predictions against actual outcomes. Score correct/incorrect | Every 4h |
| 05 | Accuracy Tracking | Per-model, per-category, and consensus accuracy updated. Calibration curves computed. All data public | Every 4h |
Every prediction, every resolution, every accuracy metric is publicly accessible. No black boxes.
Four frontier models forecast independently. Cross-model spread signals identify high-uncertainty events.
No single AI model dependency. Consensus across 4 frontier models from 4 different providers reduces individual model bias.
Outcomes are resolved programmatically against market outcomes. No manual intervention, no cherry-picking.