How we know what we show.

Project Janus describes ten futures for human civilization, each worked through so its economy, politics, technology and environment fit together. They are possibilities, not forecasts. The researchers don’t say how likely any of them is, and nothing on this site—order, color or layout—hints at it either.

1. A published paperEvery value starts at a specific table or figure in a Project Janus paper.
2. Careful copyingWe copy each value with its units and precision, and keep blanks blank instead of turning them into zeros.
3. Automatic checksBefore the site is built, checks reject any value that is missing, malformed or without a source.
4. The sitePages draw every scientific number from that checked dataset. None are typed into the design by hand.

The Observatory doesn’t simulate a telescope. For each of the ten futures, it shows what the observing study found five proposed instruments could detect, looking at a world like ours from another star.

When an instrument has nothing listed, that only means the study lists no sign it could pick up. It never means the civilization has no technology.

Distance and observing time stay fixed at the study’s own choices, because there’s no tested way to vary them. Atmosphere values come straight from the study’s tables; we don’t show spectra traced from figures in the PDF.

A separate Janus study runs each future 200 times over 1,000 years to see how often it collapses and how it recovers. The authors published their code, but without a license that would let us reuse it, so we link to it rather than copy it.

To explore the model ourselves, we rebuilt it from the equations in the paper. Our version matches 29 of the 29 results the paper states in its text, within set tolerances. For 11 more, the paper only shows a chart, so there is no exact number to check against. It’s our own rebuild, not the authors’ code, and it never replaces the paper’s numbers.

In the downloads, every value carries a label saying where it came from.

reported

Stated directly in a paper.

transcribed

Copied from a paper into our dataset, with units and precision kept.

derived

Calculated by us from published numbers, the same way every time.

reimplemented

Produced by our own rebuild of a published model.

editorial

Our explanation or grouping.

interpretive

Artwork imagined from a scenario’s description.

fictional

A creative work set inside a scenario.

model_generated

Written by an AI. Never used as fact.