Grey Swan / Archimedes  ·  Fall 2026  ·  v12.0-R1

The productive engine is running. The institutions around it are not keeping pace.

The Fall 2026 run does not worsen the inherited DTR/LIR conditional probabilities. Instead, v12 asks an additional question: how likely is the world to follow the reform path at all? The answer is the central finding of this run.

Fall 2026
Current run
v12.0-R1
Baseline / transition
5 Sep 2026
Evidence cutoff
Global
Scope
Model overview Spring 2026 Fall 2026 2050 Settlement
The main Fall finding is not a new probability nudge. It is the new posture assessment.
The world still contains serious DTR-like reforms: grid investment, AI transparency rules, multilateral governance and other concrete measures. But the dominant global pattern is acceleration with fragmented governance, concentrated compute and weak diffusion of gains. v12 therefore initializes the global posture at 10% DTR and 90% LIR, with a reasonable DTR sensitivity range of 7–15%.
10%

Do The Right Thing

Material, coordinated reforms across multiple systems before pressures compound.

90%

Let It Rip

Acceleration continues while governance, distribution and institutional adaptation lag behind.

Three-run comparison
What changed from the October 2025 baseline to Spring and Fall 2026?
Choose an outcome. These are the directly comparable conditional probabilities. The flat Spring-to-Fall line is intentional: v12.0-R1 initializes the new measurement architecture without letting replacement indicators move the inherited odds on their first reading.
Do The Right Thing
Let It Rip
2030
2040
2050
New in v12

The posture-weighted real-world outlook

The conditional scenarios answer “what happens if leaders DTR or LIR?” The new blended view also asks “how likely are leaders to choose each posture?”

At a 10% DTR / 90% LIR posture, Lost Control dominates the blended outlook.
The result is 70.6% in 2030, 87.1% in 2040 and 94.4% in 2050. This does not mean a 94% chance of civilizational collapse. Lost Control means problems increasingly outpace institutional capacity, so inequality, concentration and legitimacy problems become harder and more expensive to reverse.
HorizonFull empowermentManaged disruptionUneven transitionLost control
20300.3%3.6%25.5%70.6%
20400.7%1.9%10.3%87.1%
20501.4%0.6%3.6%94.4%

Sensitivity check: even if DTR is raised to 15%, Lost Control remains 67.9% in 2030, 84.7% in 2040 and 92.1% in 2050.


What the Fall evidence says

The world economy is not collapsing. The problem is that productive capacity, institutional capacity and diffusion are not advancing together.

Energy and grids

Amber / improving

Major systems show better adequacy and storage, but connection queues, fast load growth and infrastructure constraints remain substantial.

Health security

Amber

Surveillance and preparedness capacity remain substantial, but the inherited Health Surveillance Flag is not yet cleared.

Education and skills

Amber-red

Teacher AI use is visible, but capability and assessment adaptation remain highly uneven and far behind deployment.

Trust and information

Amber-red

Operational AI accountability is becoming real in some jurisdictions, while the wider institutional trust environment remains weak and mixed.

Coordination

Amber

Shipping and supply networks are adapting better than governments are coordinating. Deliberation is growing faster than enforceable compact capacity.

Compute and AI access

Red

Advanced compute ownership remains highly concentrated and meaningful productive access is geographically narrow.


Growth is positive. Diffusion is not.

IMF growth remains resilient and global labour productivity is positive, but v12 classifies the current pattern as concentrated growth: the productive engine is functioning while gains, compute and capability remain unevenly distributed.

The model therefore refuses to equate GDP or productivity growth with Full Empowerment.

The Wealth-Diffusion Gate stays closed.

Household financial inclusion has improved, but labour’s share of global income remains weak and shared-prosperity progress is insufficient. The gate requires improvement across at least two dimensions without material deterioration elsewhere.

At Fall 2026, that test is not met.


Cross-cutting flags

Flags change how readily positive evidence can be credited.

Geopolitical Shock: active

Geopolitical risk remains elevated. No sustained reversal is visible.

Health Surveillance: active

The new baseline improves visibility but does not yet clear the inherited surveillance flag.

Economic Stress: inactive / watch

Leading indicators softened, but IMF growth evidence does not corroborate a broad downturn.

Participation Stress: inactive / watch

Distribution is weak, but access-to-work evidence has not deteriorated enough to trigger the new two-family rule.


What changed in the model itself?

v12 is a measurement upgrade, not a retrospective rewrite of the Spring result.

Institutional trust is restored explicitly

Government effectiveness, corruption, political freedom and direct survey trust sit in a structural Trust Basket, separate from operational AI accountability.

Growth and productivity are visible

A Macro Performance Panel tracks real GDP growth, GDP per capita growth and labour productivity, then reads them jointly with diffusion rather than assuming growth equals empowerment.

Compute is measured as productive infrastructure

v12 replaces broadband speed proxies with concentration of advanced compute and geographic availability of usable AI compute.

Participation and diffusion are strengthened

A Participation Stress Flag separates labour-market access from macro stress, while the Wealth-Diffusion Gate is tied to measurable household security, labour share and shared prosperity.

Posture likelihood is now explicit

The model finally distinguishes a desirable DTR pathway from the probability that leaders will actually choose it.


Evidence and downloads

The public page is deliberately readable. The full technical record is available separately.

The Fall run uses public-access evidence from established international statistical agencies, regulators, system operators and widely used independent datasets. Core sources include the IMF, World Bank, ILOSTAT, OECD/OECD.AI, WHO, IEA and major electricity-system operators, UNCTAD, Freedom House, Transparency International, public Gallup/OECD trust aggregates, the Caldara-Iacoviello Geopolitical Risk dataset and public Epoch AI compute estimates. Estimated datasets are explicitly flagged and given less evidential weight. No private telemetry or proprietary data are used.