First entry of the model
First entry: there is no previous quarter in our register to evaluate.
Initial distribution 35 / 25 / 20 / 10 / 10. No previous official revision.
The New World / Scenario Observatory
The future changes.
The forecast is updated.
Five scenarios for 2026–2035. The forces that shift their weight, a reconstruction since 2022 and a register that preserves every revision.
The chart, table and core text are readable without JavaScript. To explore the cards and prepare revisions, enable JavaScript in your browser.
The distribution
of scenarios.
The initial weights coincide with the report. New distributions will enter the register on this page; the report and the overview preserve the research framework.
Subjective working probabilities, normalised to 100% across the five scenarios. They are dominant analytical regimes and may share some mechanisms; they do not exhaust every possible future. Select a scenario to see what favours it and what weakens it.
Managed multipolarity
Widespread productivity gains and selective cooperation favour S1. Persistent barriers, concentrated rents and energy shocks weaken it.
How the picture
has moved.
From the inflationary trauma of 2022 to the interplay between technology and fragmentation. One chart to compare the direction of the five hypotheses.
2022–2025: retrospective reconstruction, prepared on 12 September 2026. There is no archive here of probabilities expressed at the time. The dashed points are retrospective weights; the model's actual register begins on 12 September 2026. The sources document the context, not these percentages.
ReconstructionRegister from 12/09/2026
First entry: 12 September 2026
The priors are 35%, 25%, 20%, 10%, 10%. Comparing them with the reconstruction does not measure forecast accuracy.
| Date | S1 | S2 | S3 | S4 | S5 | Nature of the point |
|---|---|---|---|---|---|---|
| 31/12/2022 | 20% | 10% | 40% | 15% | 15% | Reconstruction |
| 31/12/2023 | 25% | 20% | 30% | 10% | 15% | Reconstruction |
| 31/12/2024 | 35% | 25% | 20% | 10% | 10% | Reconstruction |
| 31/12/2025 | 30% | 25% | 25% | 10% | 10% | Reconstruction |
| 12/09/2026 | 35% | 25% | 20% | 10% | 10% | First entry |
How to read the reconstruction and the reference to 2021
2021 is the year Dalio's book was published. Our application dates from 2026: it does not attribute to the report's author forecasts recorded since 2021. The series begins in 2022, as a reconstruction of the context that followed the book.
The 2022–2025 weights are an analytical proposal in steps of five percentage points, built with the report's categories and anchored to dated sources. They do not derive from an estimated statistical model or from probabilities supplied by the IMF, the ECB or Stanford. A different reasonable reading may assign different weights.
A retrospective limitation is unavoidable: we already know the events, and some sources were published after the year being reconstructed. In particular, the AI Index 2024 describes 2023, and S5 in 2022 concerns mainly debt, before the current cycle of investment in generative AI.
The lines connect discrete points; there are no intermediate monthly or quarterly observations. This reconstruction is not used to compute Brier scores or to claim accuracy. The first recorded point is that of 12 September 2026; the continuous stretch will emerge with future revisions.
Which signal
shifts the weight?
What matters is how much a new piece of evidence distinguishes one scenario from the others. More news about the same cause is not independent confirmation.
The last recorded distribution is maintained.
Sensitivity simulation. The multipliers are illustrative assumptions, not estimated likelihood ratios. The selection does not change the official weights.
Simulated distribution / change from the recorded entry
Simulation rule
New weight = recorded weight × multiplier, divided by the sum of the five products. Rounding to whole numbers is distributed so as to preserve the 100% total.
S1 × 1 · S2 × 1 · S3 × 1 · S4 × 1 · S5 × 1
Only one family of evidence is explored at a time. A real revision requires assessing intensity, persistence, contrary data and dependencies between signals; the model does not apply these multipliers to the register automatically.
Capability → results
Better models and more capex are signals of capability. Productivity, external revenues and cash flows verify that the benefit is economic.
Shock → persistence
For S3, duration and transmission matter: a temporary price rise does not carry the same weight as persistent inflation and dearer credit.
Scale → distribution
The S1–S2–S4 comparison depends on where the benefits spread: across several poles, mainly the US, or the Chinese industrial base.
The open test in 2026. The IMF's July update describes simultaneous technological tailwinds and war-related headwinds. We maintain the report's initial distribution, but the persistence of the energy shock and its financial transmission are evidence that may call for more weight on S3. IMF, July 2026 ↗
One quarter at a time.
Without rewriting the past.
Each entry preserves what was forecast, what happened, the errors acknowledged and the reason for any new weights.
01 / Compare
Separate genuine surprises, normal fluctuations and data that are still incomplete. Compare each piece of evidence with the recorded forecast.
02 / Decide
Justify the points assigned and those subtracted. Confirming the weights is also a decision to be documented.
03 / Preserve
Add a new dated entry. Previous versions remain in the chart and in the journal; the report's framework remains available for consultation.
Learning
from the outcome.
Three narrowly defined questions recorded on 12 September 2026 make part of the forecast assessable before the 2035 picture is resolved.
0 / 3Questions resolved
—Mean Brier · lower is better
0.25Benchmark: always 50% on a binary event
Initial judgemental priors: 65%, 80%, 60%. The model's own elaboration, not probabilities of the institutions cited. The score is computed only on verified outcomes; three questions are not enough to demonstrate a stable capability. The five scenario probabilities and these three events concern different objects.
65%
Global growth 2027
Will the first edition of the October 2026 WEO show a forecast for world real growth in 2027 of at least 3.0%?
Open · deadline 31/10/2026
Verification rule
Use the aggregate world value of the first October WEO, without replacing it with subsequent revisions. If the edition is not available by 31/10/2026, leave unresolved.
80%
Centrality of the dollar
Will the first COFER release for the fourth quarter of 2026 show a USD share of at least 55.0% of world foreign-exchange reserves?
Open · deadline 30/04/2027
Verification rule
Use the USD share of the total by currency according to the COFER methodology of the first release, excluding gold. A change of scope that prevents comparison leaves the question unresolved. Verification deadline: 30/04/2027.
60%
AI diffusion in EU enterprises
Will the first Eurostat estimate for 2026 show at least 25.0% of EU enterprises with 10 or more employees using an AI technology?
Open · deadline 31/03/2027
Verification rule
Table isoc_eb_ai: EU-27, all enterprises with 10 or more employees within the survey scope, any AI technology, percentage rounded to one decimal place. If absent by 31/03/2027, leave unresolved.
How we measure error
For each binary question: Brier = (initial probability − outcome)², with outcome 1 if the event occurs and 0 if it does not. The mean uses only resolved questions. The initial probabilities and the rules remain preserved even once the answer is known.
A missing data point is not automatically assigned a "no". Statistical revisions and changes of scope must be flagged according to the rules of the individual question. Truly assessing calibration and quality will require many more forecasts, recorded over time with consistent criteria.
Workspace — prepare the next quarterly revision
Complete after the quarter closes. The button downloads the updated fourth page, with the complete register: replace only scenario_observatory.html on the site. Completing the form by itself does not change the online version. Entries already recorded are preserved.
The sources.
The documented context.
Institutional documents and methodological references. The scenario percentages and the multipliers remain the model's own elaborations.
Ray Dalio / Simon & Schuster · Principles for Dealing with the Changing World Order: Why Nations Succeed and Fail30 November 2021Chang, Chen, Mellers and Tetlock · Developing expert political judgment2016Stanford HAI · The 2026 AI Index Report2026OECD · Venture capital investments in artificial intelligence through 202517 February 2026International Energy Agency · Key Questions on Energy and AI, Executive summary2026European Central Bank · The international role of the euro2 June 2026European Central Bank · Financial Stability ReviewMay 2026Bank for International Settlements · The next-generation monetary and financial system2025The White House · Establishment of the Strategic Bitcoin Reserve and United States Digital Asset Stockpile6 March 2025IMF · WEO, Countering the Cost-of-Living Crisis11 October 2022IMF · WEO, Navigating Global Divergences10 October 2023Stanford HAI · AI Index 20242024; data relating to 2023IMF · WEO, Policy Pivot, Rising Threats22 October 2024IMF · The Global Economy Enters a New Era22 April 2025IMF · WEO, Global Economy in Flux, Prospects Remain Dim14 October 2025IMF · WEO Update, Crosscurrents of War and Technology8 July 2026IMF · COFER — reserves by currencyquarterly seriesEurostat · Use of artificial intelligence technologies in enterprisesannual series isoc_eb_ai