White paper
APOCALYPSE: A Multi-Agent Forecasting System for the Global Apocalypse Index
Public technical white paper describing the system architecture, the Global Apocalypse Index (GAI), and the daily publication cadence.
Download PDFGLOBAL APOCALYPSE INDEX SYS-01
GAI · INDEX
—
loading…
—
Provenance composition
— admitted
| Method | Count | Avg fidelity | Total weight |
|---|
Briefing
Loading…
GAI 7-day drift
—
forecast vs current (pp)
Anomalies (|z|≥2)
—
across all indices
Critical regime
—
|z|≥3 indices
Categories
—
domains tracked
Model inputs
—
channels trained daily
Public indices
—
on the dashboard
Real adapters
—
live today
Backtest AUC
—
walk-forward
—
| Index | Category | Value | Δ | Δ% | z | Regime | Trend (60d) | Source |
|---|
Provenance
Every admitted source is tagged with the method that produced today's value. Real publications carry weight 1.0; frequency-aligned upsamples 0.95; nowcasts 0.85; statistical-fidelity-gated synth 0.70. The composite GAI integrates each contribution at its calibrated weight.
Joining promotion countdown
Sources with at least one real observation but fewer than 30 in window. Each tile shows real-observation accrual progress and an estimated time-to-admission based on the source's native cadence.
Pending-bind accrual
Single-snapshot APIs accumulating a daily real-stamp counter. Each source becomes joining as soon as the stamp ticker reaches 1, then admitted on day 30.
Constellation
Each tile is a tracked index sized by its z-magnitude and shaded by regime. Click any tile to drill into the full chart.
Categories
Each category aggregates indices feeding the GAI from one domain.
Sources
APOCALYPSE Research Foundation
Founder: Professor Dr. Stelios Bekiros
Peer-reviewed publications, technical references, and white papers from the APOCALYPSE project.
Publications & documents
Published papers
Selected peer-reviewed work by the founder. For the complete list, see steliosbekiros.com or Google Scholar.
- Publications
- AI, ML & Big Data
- Quantum Computing
- Econophysics & Chaos Resources
- Blockchain — Blueprint for a New Economy
- Algorithms & Programming
Cite this work
@misc{bekiros2026apocalypse,
author = {Bekiros, Stelios},
title = {{APOCALYPSE}: A Multi-Agent Forecasting System for the Global Apocalypse Index},
year = {2026},
howpublished = {APOCALYPSE Research Foundation},
url = {https://apocalypse-index.org},
note = {Daily publication; see white paper for methodology}
}
Support APOCALYPSE Research Foundation
With your support we can continue this pioneering research project. Thank you.
PayPal
One-time donation. Accepts all major card types worldwide.
Donate via PayPalLink to be added.
Stripe
One-time donation with fixed amounts: €5 / €25 / €100 / Custom.
Donate via StripeLink to be added.
Ko-fi
Low-friction micro-donations and optional monthly support.
Support on Ko-fiLink to be added.
All donations support open research. Receipts available on request.
Contact
- Email info@apocalypse-index.org
- Personal site steliosbekiros.com
- ORCID 0000-0002-9376-8726
- Google Scholar Publications profile
About APOCALYPSE
APOCALYPSE is the original work of Prof. Dr. Stelios Bekiros. It is an independent multi-source intelligence system that estimates planetary and human-system stress and emits a single calibrated number, the Global Apocalypse Index (GAI).
How values get here
- Rolling-window observations are pulled from every tracked source.
- An ensemble of specialist neural agents trains and backtests on the rolling window.
- After inference, learned outputs (level, forecast, anomaly z, regime) are published.
- The dashboard refuses to publish without a fresh, validated model-run manifest.
Editorial principles
- Transparency of inputs. Every tracked index links back to its publicly citable publisher.
- No single-factor dominance. The Index is composite by design.
- Calibration over hyperbole.
- Not financial advice.
Cadence
- Daily 12:00 GMT — full publication cycle.
- On-event — out-of-band refresh on large disturbances.
Copyright
© Stelios Bekiros — All Rights Reserved. The methodologies and internal models are proprietary.