The Spring 2026 run confirms a deterioration since the October 2025 baseline and introduces a third signal that the model is now formally tracking: economic conditions and labour market resilience. Two flags are active. A third is approaching threshold.
Grey Swan converts public data into auditable probabilities for two scenarios and four outcomes at three time horizons: 2030, 2040, and 2050. Odds move only when evidence is persistent across multiple readings and corroborated across domains. Every change is logged. No private data is used.
The name comes from Taleb. A grey swan is consequential and visible in today's data — but uncomfortable enough to ignore. This model refuses to look away.
The two scenarios are Do the right thing — where leaders choose coordinated reforms before 2030 — and Let it rip — where action is deferred despite visible evidence. Three cross-cutting forces shape all six levers: climate, geopolitical shocks, and (from v11.9) economic conditions and labour market resilience.
The original model was designed when GDP growth could reasonably be treated as slow background context. That assumption no longer holds. Tariff shocks, AI-driven labour market displacement, and the uneven diffusion of productivity gains mean economic conditions can now suppress or enable institutional reform within a single six-month window.
v11.9 introduces the Economic Stress Flag to track this. It is not yet active. But it is approaching. That is itself the news.
AI is compressing decision timescales across energy, health, education, information, cross-border cooperation, and computing access. Leaders have a closing window to make coordinated choices simultaneously. Grey Swan makes the cost of inaction legible, one public indicator at a time.
What each scenario leads to — and what distinguishes them in practice.
Power is reliable enough for households and firms to plan. Health systems manage peaks without burning out staff. AI-aware assessment is embedded in education. High-risk content carries provenance by default. Cross-border compacts function in practice. Advanced computing is accessible to small firms and public-interest projects. This outcome is only awarded when multiple levers meet thresholds simultaneously and the Wealth-Diffusion Gate is open — requiring rising household savings, spreading profit-sharing, and improving inequality trends.
Shocks continue but the system absorbs them more often than not. Power is tight in some places and stable in others. Health systems cope unevenly. Some classrooms and firms use AI tools well while others lag. Provenance and audit regimes exist in sensitive areas but coverage is incomplete. Gains are real but not yet secure. A plausible waypoint on the reform path — a step toward Full Empowerment if coverage widens, or back toward a patchwork if progress stalls.
Real progress in some regions and sectors alongside stagnation elsewhere. Some grids add capacity while others queue. Some education systems adopt AI-integrated assessment while others ban the tools. Coordination across borders is fragile. Compute access and skills gather in a small number of hubs. Wealth gains cluster. People in leading regions do well; others feel left behind. This is the most probable near-term state under the reform scenario.
Power shortages recur and hold back investment. Health systems lose staff and fail during peaks. Assessment integrity erodes because old tools no longer work and new ones are not in place. Misinformation outpaces provenance capacity. Cross-border coordination fails when most needed. Household income buffers shrink and inequality worsens. Recovery from shocks is slow and expensive. This outcome is not a dramatic collapse — it is a state in which problems compound faster than institutions can absorb them. Under the deferment scenario, it dominates at every horizon.
The pressure points that move the whole system when they shift together — named after Archimedes' principle of leverage.
Reliable, affordable electricity delivered where it is needed. Tracked via connection queues, curtailment rates, grid stress during peaks, and nighttime-lights anomaly data.
Surveillance, clinical capacity, and staffing that detect threats early and sustain care through peaks. Tracked via WHO FluNet, EuroMOMO excess mortality, and surveillance coverage breadth.
Teaching, assessment, and credentials that produce adaptive learners with employer-recognized proof of ability. Tracked via AI-aware assessment coverage, teacher training rates, and credential uptake.
Provenance, transparency, and independent audits that make content traceable and high-impact AI accountable. Tracked via internet outage events and confirmed national disruptions.
Narrow, enforceable cross-border compacts that keep critical inputs and data flowing and resolve disputes quickly. Tracked via shipping connectivity, container throughput, and the tariff fragmentation index.
Practical, safe access to capable compute for small firms and public-interest work. Tracked via M-Lab broadband performance and reserved compute utilisation.
Climate, geopolitical shocks, and economic conditions cut across all six levers rather than occupying a single domain. Their effects show up inside each lever's indicators rather than in a separate measurement.
Gains only count as empowerment when they reach ordinary households. The Wealth-Diffusion Gate requires rising household savings, spreading profit-sharing, and improving inequality trends before the best outcome can be awarded. All data is public; odds move only when signals persist across multiple readings.
Current odds under both scenarios at three time horizons. Each row sums to 100%. Odds are conservative by design — they move only when public indicators show persistent, corroborated improvement.
| Horizon | Scenario | Full empowerment | Managed disruption | Uneven transition | Lost control |
|---|---|---|---|---|---|
| 2030 | DTR | 3% | 18% | 57% | 22% |
| 2030 | LIR | 0% | 2% | 22% | 76% |
| 2040 | DTR | 7% | 10% | 40% | 43% |
| 2040 | LIR | 0% | 1% | 7% | 92% |
| 2050 | DTR | 14% | 6% | 27% | 53% |
| 2050 | LIR | 0% | 0% | 1% | 99% |
Signals that drove the change from the October 2025 baseline.
Fifteen months of tariff volatility — Liberation Day in April 2025, the Supreme Court strike-down in February 2026, and residual Section 232 tariffs on metals and pharmaceuticals — has pushed the trade fragmentation index to approximately 142 against a 2019 baseline of 100. Persistent across shipping, connectivity, and policy data.
USAID funding reductions have reduced WHO FluNet reporting frequency across sub-Saharan Africa and Central Asia. The Health Surveillance Flag is active: positive health-lever nudges are capped at zero until coverage recovers across two consecutive runs.
The Caldara-Iacoviello Geopolitical Risk index stands approximately 1.4 standard deviations above its 24-month rolling mean. Geopolitical Shock Flag active: positive signals in treaty-lite, energy, compute, and health require three consecutive readings before generating an upward nudge.
New in v11.9. The OECD Composite Leading Indicator shows mild deterioration linked to tariff-driven uncertainty but has not yet met the two-consecutive-reading threshold. ILO employment-to-population is softening in lower-income groups. The Economic Stress Flag is inactive this run but is the primary signal to watch for Autumn 2026.
Open-weight model proliferation has broadened access beyond frontier labs. Does not yet meet the elevated three-reading threshold required while the Geopolitical Shock Flag is active. Noted for the next run.
Probabilities shift only when public indicators show persistent, shared improvement across multiple domains simultaneously.
Movement in the lost control outcome by horizon and scenario, in percentage points.
Active flags raise the corroboration threshold for positive signals in exposed levers.
WHO FluNet coverage degraded across sub-Saharan Africa and Central Asia. Positive health nudges capped at zero until recovery across two consecutive runs.
GPR index ~1.4 SD above rolling mean. Trade fragmentation persistent. Treaty-lite, energy, compute, and health levers require three-reading persistence for positive nudges.
OECD CLI softening but not yet at trigger level. ILO employment-to-population declining in lower-income groups. Reassess at Autumn 2026 run.
No improvement across household savings, profit-sharing breadth, or inequality trends. Tariff shock trajectory likely to worsen subsequent vintages.
Full model documentation, working papers, and the Spring 2026 intelligence memo.
The Grey Swan / Archimedes framework is documented in full in the foundational working paper (WP11). The methodology uses a four-tier evidence architecture: Tier-0 canon (first principles), Tier-1 slow anchors (IMF, World Bank, UN, IEA), Tier-2 operational signals (WHO FluNet, shipping indices, grid operator data, M-Lab broadband, OECD CLI), and Tier-3 interpretive context. All data is public. No private telemetry is ingested. Every odds change is logged with source, direction, magnitude, and the reversion rule that will unwind it if the signal fades.
Public data only. No private telemetry. Sources: IMF, World Bank, UN, IEA, WHO, OECD, ILO, NOAA, M-Lab, CAIDA, RWI/ISL, UNCTAD, WTO, Federal Reserve.
Odds move in small bounded steps when signals persist across multiple readings and are corroborated across levers. Every move is reversible. Single data points do not move probabilities.
Runs every six months — spring and autumn — allowing genuine two-reading persistence to establish before any movement is recorded. Next run: Autumn 2026.