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.

Each source has one designated primary method (its label on the Index panel); every other method that applies to that source (12 or 13 of them) runs as a shadow, contributing to the ensemble at its tier fidelity. Primary Wt counts only primaries; Ensemble Σfid counts all.

Sources without an active method on a given day are omitted from the composition; they re-enter automatically when their input flows.

Onboarding telemetry

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 from one domain. These indices feed GAI, GAI-x, and GAI-y jointly — the same channels drive the headline reading and the two counterfactuals.

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 Provenance-Anchored Multi-Agentic {AI} Early Warning System ({EWS}) for Global Catastrophe Risk},
  year         = {2026},
  howpublished = {White paper, Public edition v.2, {APOCALYPSE} Research Foundation},
  url          = {https://apocalypse-index.org},
  note         = {Daily publication of the GAI, GAI-x and GAI-y; methodology in the white paper}
}

All donations support open research. Receipts available on request.

• General inquiries info@apocalypse-index.org Media, press, technical questions about the Index, dashboard issues, and general correspondence.
• APOCALYPSE Research Foundation gai.foundation@apocalypse-index.org Institutional partnerships, research collaborations, grants, data-sharing agreements, and donor relations.
• Prof. Dr. Stelios Bekiros stelios.bekiros@apocalypse-index.org Direct line to the founder for academic exchange, invited talks, and confidential matters.
• Personal site steliosbekiros.com
• ORCID 0000-0002-9376-8726
• Google Scholar Publications profile
• YouTube @apocalypse-index Daily auto-published GAI briefing video covering every tab of the operations terminal, posted 12:25 UTC.
• X @apocalypse_idx Daily GAI readout and briefing links.

About APOCALYPSE

“APOCALYPSE: A Provenance-Anchored Multi-Agentic AI Early Warning System (EWS) for Global Catastrophe Risk” is the original work of Prof. Dr. Stelios Bekiros. It is an independent, AI-driven planetary and human-system stress intelligence platform, combining evolutionary and genetic optimization, deep artificial neural networks, and big-data ingestion into a single calibrated readout: 119 public data sources, completed by the data agents and expanded by the signal agents into thousands of model channels.

The readout has three coupled dimensions:

• GAI — the headline Global Apocalypse Index: a calibrated, present-state estimate of global stress across the full multi-domain channel universe.
• GAI-x — the human-driven counterfactual: the same forward pass with cosmic / natural channels held at their calibration-window baseline, giving the stress reading driven by human-system activity.
• GAI-y — the cosmic-driven counterfactual: the mirror image, with human channels held at baseline, giving the stress reading driven by geophysical and space-weather activity.

Together, GAI, GAI-x and GAI-y form a single readout — the headline level, the human-driven reading and the cosmic-driven reading, which do not sum to GAI — rather than a single number in isolation. Each dimension is a daily nowcast; every index carries its own forecast.

The system aims at monitoring, nowcasting and forecasting social and natural crises which are associated and/or caused by i) cosmic (exogenous) events and ii) human-driven (endogenous) socio-economic and geopolitical stressors. It generates the readout of GAI leading indicators regularly updated on a daily frequency. The platform renders a source of reliable predictability for imminent crises while providing highly technical yet also practical guidance on risk assessment and decision making.

As a multi-agentic AI-driven EWS, it timely detects complex patterns of global stressor phenomena amongst a plethora of human-triggered and cosmic-derived catastrophe risk factors, ranging from e.g., major conflicted zones, violent uprisings, large-scale popular protests, pandemic outbreaks, financial, micro- and macro-economic trends and reversals, cyber-attacks, food, supply chain and logistics disturbances, poly-market trends, electoral cycles, demographic and socio-economic turmoil for any geographical location, viral and disruptive social-media trends etc, and extending up to incorporate critical planetary and lunar convergences and divergences, solar-cycle and heliospheric perturbations with regular or abnormal periodicities, climate, environmental and weather anomalies, terrestrial and geomagnetic variations and seismic and volcanic precursor effects.

Users can monitor the status, frequency and evolution of all open-access publicly available data sources and indices. Reference readings on identifying, interpreting and predicting the complex system dynamics of the GAI, GAI-x and GAI-y drivers, can be accessed via the “Academic” menu tab.

How values get here

1. Rolling-window observations are gathered daily from every tracked source across the public-data universe.
2. Inputs are normalised, gap-filled, and reconciled into a unified state-vector at a shared cadence.
3. An ensemble of specialist neural agents is retrained daily (warm start) on the full input panel (all 119 indices over the daily timeline from 1900 to today) and backtested on the rolling window under cross-validated supervision.
4. The trained model emits GAI as the calibrated forward pass on the full channel panel; GAI-x and GAI-y are derived from the same model by holding the cosmic and human channel groups respectively at their calibration-window baseline, so all three dimensions share one set of weights and one calibration.
5. Each daily model is evaluated on a held-out validation slice against four pre-registered promotion gates (discrimination, calibration, false-alarm rate, anti-regression floor). The gate results guide the model’s optimisation and are recorded with every run; the newly trained model is published every day.
6. The dashboard publishes the daily readout with its model-run manifest; if no manifest is available for the day, the previous day’s readout remains in place.

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 — full publication cycle (dated as in the RUN label and the YouTube Daily Briefing).

Copyright & Trademark

APOCALYPSE™ — Patent Pending IE PTIE20260000000318 (v1) · IE PTIE20260000000354 (v2). Published under the auspices of the APOCALYPSE Research Foundation.

© Stelios Bekiros — All Rights Reserved. The methodologies, models, weights, and source code are proprietary.