The range is the answer
A point estimate for a decade out is a guess wearing a suit. Every output is a distribution with its width reported.
Everything runs in your browser. Nothing you build is sent anywhere.
Why this exists and who builds it.
Futurisca started from a simple frustration. Ask almost any tool what happens if a large economy restricts an export, and you get either a paragraph of hedged prose or a single number with no reasoning attached. Neither helps you think.
The alternative is not a better guess. It is a model you can interrogate: one that returns a range rather than a point, that clusters its futures into families you can name, that shows the terms behind every number and lets you keep opening them until you reach the assumption underneath, and that tells you plainly where it is extrapolating past anything history can check.
That is the whole product. A state of the world, a coupled model over it, and an interface built around the question why rather than the question what.
It is built and maintained by Lewis De La Paz.
A point estimate for a decade out is a guess wearing a suit. Every output is a distribution with its width reported.
If a number cannot be opened into the terms that produced it, it should not be on the screen.
An estimate presented as a measurement is the one thing a world model must never do.
The limits page is not a disclaimer. It is a feature.
Before a single equation, two hundred and thirty eight country records: population, output, prices, employment, debt, trade, energy, emissions, technology, security and stability, every field stamped with where it came from and when it was checked. A model over a world you do not trust is an exercise in decoration.
Countries are joined by a bilateral trade network, by shared goods, by alliance and bloc membership, and by the prices of energy, food, chips and freight that everybody faces at once. Most of what the instrument shows you is second order: the country nobody named that moves because of who it sells to.
A single run was never the product. The ensemble came next, then the separation of what the shocks explain from what the coefficients explain, then the clustering into families, then the sensitivity pass, then the historical comparison. Each of those is a different question about the same run.
Growth, inflation and unemployment were rewritten as sums of named terms so that every number could be opened. This cost performance and it is the reason the causes page exists. A number you cannot open is a number you cannot argue with.
The map, the globe, the network, the palette, the sentence box. Built last deliberately: an interface designed before the thing it is an interface to ends up shaping the model to suit the screen.
The engine is JavaScript and it runs in your browser. A scenario about a company, a border, a client or an unannounced policy is never uploaded, because there is nowhere for it to be uploaded to. Saved scenarios, if you save any, go into your own WordPress database on your own site.
Every source it draws on is open and unauthenticated. The optional written briefing can use a model provider if you configure one, and if you do not, the briefing is still written, deterministically, by the instrument itself.
There is no research team behind the coefficients. They are published elasticities with published ranges, cited on the model page, chosen by one person reading the literature. The sensitivity pass exists partly so that you can see which of those choices your answer actually depends on.
A wrong base year figure propagates into every year after it. If you know a country record is out of date or wrong, that correction is worth more than a feature request, and the provenance marks are there partly so you can find them.
Corrections to the world dataset are especially welcome.
Get in touch