Apocalypse Clock is a quantitative systemic-risk horizon model designed to estimate, compare, and visualize how major global threats may converge toward critical systemic pressure. It is not a prophecy, not a deterministic prediction, and not an empirical statement that collapse will occur in a particular year. Its purpose is more precise: to transform documented scientific and institutional evidence, uncertainty ranges, dependency structures, scenario assumptions, and explicit mathematical rules into transparent model-derived critical-horizon distributions.

The model starts from the premise that global systemic risks cannot be reduced to a single physical unit. Climate disruption, nuclear escalation, artificial intelligence capability risk, biodiversity loss, antimicrobial resistance, freshwater scarcity, cyber fragility, debt stress, geopolitical militarization, ocean degradation, governance fragmentation, and other civilizational-scale threats differ in mechanism, evidence base, time scale, reversibility, and causal structure. A simple ranking would therefore be analytically insufficient. Apocalypse Clock solves this by using a shared modelling grammar: every threat is described through the same set of systemic-risk dimensions, while still preserving its domain, process class, uncertainty range, source metadata, and dependency links.

The current model evaluates 23 systemic threats across three broad domains: civilizational, biospheric, and technological risk. Each threat is represented by eight parameters. Six of them are Multi-Criteria Decision Analysis dimensions: scale, urgency, acceleration, interdependence, irreversibility, and governance failure. The seventh parameter is an annualized effective systemic risk-growth rate. The eighth is a normalized destabilization threshold. In total, the current source-map contains 184 non-meta threat-parameter entries.

The six MCDA dimensions describe different aspects of systemic pressure. Scale estimates the potential magnitude of harm if a threat intensifies. Urgency captures near-term pressure and present-decade priority. Acceleration describes whether the underlying drivers, exposure, capabilities, or impacts are worsening. Interdependence measures how strongly a threat is coupled to other systemic risks. Irreversibility estimates the persistence of damage and the difficulty of recovery after severe disruption. Governance failure captures the risk that institutions, law, coordination systems, political structures, or collective-action mechanisms fail to contain the threat.

These dimensions are not treated as direct physical measurements. They are ordinal systemic-pressure scores. The growth-rate parameter is different: it represents an annualized effective risk-growth proxy, not raw indicator growth and not a direct probability of collapse. The threshold parameter is also distinct: it is a normalized destabilization threshold used by the model to estimate when a threat may cross from serious pressure into critical systemic relevance. The model therefore explicitly separates three epistemic layers: ordinal pressure scoring, growth dynamics, and threshold-crossing logic.

The data architecture is based on a flat source-map schema. Every parameter has a key of the form threat.metric, for example climate.scale, nuclear.threshold, ai.growth_rate, or water.interdependence. Each entry contains a central estimate, a lower estimate, an upper estimate, a source, a URL, an access date, an evidence-strength classification, and an explanatory note. This structure allows the model to validate completeness, trace assumptions, compare datasets, export results, and expose uncertainty rather than hiding it behind a single number.

The evidence layer is calibrated primarily from high-quality scientific and institutional sources. Preferred sources include peer-reviewed scientific literature, major assessment reports, official statistical datasets, intergovernmental assessments, and authoritative institutional publications from organizations such as the IPCC, WHO, WMO, IPBES, FAO, UNEP, UNESCO, the World Bank, SIPRI, UNHCR, V-Dem, ENISA, OECD, IMF, IEA, and comparable evidence-producing bodies. Popular media, unsourced claims, speculative commentary, and non-verifiable estimates are not used as primary calibration sources. Where no direct empirical measurement exists for an abstract systemic-risk dimension, the parameter is treated as evidence-anchored modelling judgment and documented through notes, ranges, and evidence-strength labels.

The model does not treat AI-generated numbers as evidence. AI systems may assist in structuring, auditing, comparing, or formatting source-map datasets, but the numerical parameters must remain anchored to traceable scientific, governmental, intergovernmental, or institutional evidence. This distinction is essential. Scientific sources strengthen the input layer, but they do not remove the need for a model layer, because no single dataset directly measures “civilizational destabilization” across all threat classes.

The computational pipeline proceeds in stages.

First, the source-map layer defines the initial threat state. For each threat, the model reads the six MCDA dimensions, the effective growth rate, the destabilization threshold, the uncertainty interval, the evidence strength, and the source metadata.

Second, scenario conditioning modifies the baseline parameter space. Scenarios can shift systemic-pressure dimensions, domain multipliers, growth assumptions, and governance-related stress. This means the clock does not claim to describe one inevitable future. It describes conditional model states under explicit assumptions.

Third, the model computes a weighted MCDA base-pressure score for each threat. In simplified mathematical form:

Sᵢ = Σⱼ wⱼ · xᵢⱼ

where Sᵢ is the base systemic-pressure score for threat i, xᵢⱼ is the normalized value of threat i on MCDA dimension j, and wⱼ is the corresponding dimension weight. The six dimensions are therefore transformed into a comparable baseline score.

Fourth, the model applies domain weighting. Threats are grouped into civilizational, biospheric, and technological domains. Domain weights allow the model or user to test how results change when one systemic layer is emphasized more strongly than another. These weights are normalized so that changing domain emphasis changes relative influence rather than simply inflating all risk scores.

Fifth, the model applies dependency amplification. Systemic threats do not act in isolation. Climate stress can amplify water scarcity, food-system instability, displacement, and geopolitical conflict. Cyber disruption can interact with finance, infrastructure, supply chains, and epistemic systems. Governance fragmentation can weaken the capacity to respond to nearly every other threat. Apocalypse Clock therefore treats dependencies as computationally relevant. A threat with high network influence can receive greater systemic priority because of its ability to transmit or amplify destabilization.

Sixth, the model estimates threat-level threshold-crossing horizons. The question is not “when will the world collapse?” but rather: under this parameter set, when does a given threat approach a critical systemic-pressure threshold? The answer depends on the threat’s adjusted pressure, effective growth rate, threshold value, process class, and uncertainty range.

Different process classes are handled differently. Continuous threats, such as climate degradation, biodiversity loss, freshwater stress, or soil-system decline, can be represented through exponential pressure-growth logic. Event-like threats, such as nuclear conflict, engineered biological events, systemic cyberattacks, or severe geopolitical escalation, are better represented through event-arrival logic such as non-homogeneous Poisson or Bernoulli-style crossing processes. Regime threats, such as governance fragmentation, authoritarian drift, economic fracture, or advanced AI destabilization, may be represented through logistic transition or regime-shift logic. This distinction matters because ecological degradation, military escalation, institutional breakdown, and technological destabilization are not the same temporal process.

Seventh, the model propagates uncertainty through Monte Carlo simulation. Instead of calculating one deterministic result, it repeatedly samples plausible parameter values from uncertainty intervals. Bounded ordinal dimensions can be sampled with beta-type logic, while positive growth parameters can be sampled on log-normal support. Each simulation run recalculates threat scores, thresholds, growth paths, dependency effects, and crossing horizons. The final output is therefore a distribution of possible model-derived critical horizons, not a single unsupported date.

Eighth, the model converts threat-level horizons into system-level results through structural aggregation. The current model compares several aggregation logics. A compensatory aggregation rule estimates how total weighted systemic pressure accumulates across threats. A non-compensatory max-rule tests whether a dominant threat can drive the system-level horizon by itself. A graph-weighted aggregation rule gives additional importance to threats with strong network position and dependency influence. A dynamic-cascade rule estimates how interacting threats may propagate destabilization through coupled systems.

The large highlighted year is produced by the dynamic-cascade aggregation rule. It should be interpreted as the P90 upper edge of the dynamic-cascade critical-horizon distribution. P10, P50, and P90 are not empirical probabilities of collapse. They are quantiles internal to the model distribution generated by the selected source map, scenario, weights, growth rates, thresholds, and dependency structure. P10 represents the early edge of simulated critical-horizon outcomes, P50 represents the median model horizon, and P90 represents the upper-edge stress horizon under the selected assumptions.

The dashboard also produces secondary outputs that help interpret the result. These include a Global Systemic Index, threat-level horizon estimates, domain contributions, priority rankings, cumulative crossing curves, model-implied crossing probabilities by selected years, uncertainty intervals, dominant-driver diagnostics, and comparative structural outputs from alternative aggregation algorithms. These outputs are not decorative; they show why the headline year appears where it does and which assumptions exert the strongest influence.

A separate diagnostic layer examines model robustness. One-at-a-time sensitivity analysis tests how changes in individual parameters affect outputs. Sobol/Jansen-style global sensitivity analysis estimates how variance in outputs is distributed across input assumptions. SMAA-style weight robustness tests whether rankings remain stable under alternative weight configurations. Non-compensatory veto diagnostics test whether a severe threat can dominate even when aggregate scores appear moderate. Tail-dependence stress tests examine whether multiple threats can become dangerous together rather than independently. Network diagnostics such as eigenvector centrality estimate which threats are structurally influential within the dependency graph. Shannon entropy can be used to measure whether the risk landscape is concentrated in a few dominant threats or dispersed across many interacting threats. Joint-crossing diagnostics, including Poisson-binomial or multi-threat crossing logic, help estimate the probability that several threats cross relevant thresholds within the same time window.

The epistemic interpretation of the model is therefore deliberately cautious. Apocalypse Clock does not claim empirical predictive validity in the same sense as a model trained on many repeated historical collapses, because no adequate dataset of repeated global civilizational cascades exists. The correct scientific question is not whether the model can eliminate uncertainty. It cannot. The correct question is whether the assumptions are explicit, the data sources are credible, the scoring system is inspectable, the equations are transparent, uncertainty is propagated, sensitivity is tested, and the output is interpreted as a conditional model horizon rather than as a prophecy.

The strongest use of Apocalypse Clock is comparative and diagnostic. It helps identify which systemic threats appear most urgent under a given evidence map, which domains contribute most to total pressure, how much the result depends on particular assumptions, whether the risk landscape is concentrated or distributed, and how network interactions may shift the global horizon. It is a structured warning instrument: not a final prediction of the future, but a transparent computational framework for reasoning about interacting global risks under uncertainty.

In summary, Apocalypse Clock is a source-based, mathematically explicit, uncertainty-aware systemic-risk horizon model. It combines MCDA scoring, domain normalization, dependency amplification, process-specific threshold dynamics, Monte Carlo sampling, structural aggregation, dynamic-cascade modelling, and scientific robustness diagnostics. Its results should be read as model-derived critical horizons: analytical stress signals produced under explicit assumptions, not deterministic forecasts, not official scientific consensus, and not literal predictions of collapse.