Oracle

Oracle answers with a number. Not "most likely", not "probably", but "68% by December 31" — and explains what that number is built from.

How it works

It first turns any vague question into a resolvable one: what exactly counts as the event happening and by what date. Without that, a prediction can neither be verified nor credited.

Quick answer — in seconds: fresh facts, the prediction market price if one exists, probability, and a couple of lines of reasoning.

The Council — when the question deserves it. Four analysts get the same set of facts and work independently, each through their own lens:

• Base rates — how often similar things happened historically
• Game theorist — players, their payoffs, equilibrium
• Skeptic — the strongest argument against
• Trendspotter — the strongest argument for "sooner than it seems"

Only after them does the prediction market open — so expert opinions are not an echo of it. The judge blends everything together using superforecasting rules: median as the starting point, extremization under independent agreement, and a liquid market weighs as roughly two experts.

Output: a final probability, the contribution of each lens, and what to watch — events that would shift the estimate.

Track record

Each prediction is logged with a deadline and a market anchor. Outcomes are settled automatically; accuracy is measured by means of the Brier score — over time you can see what predictions actually turned out accurate and which of what the Oracle thinks stands up to reality.

Mini-app

In Telegram, Oracle has its own observatory screen: a probability scale, a council axis with expert dots and the judge's marker, a breakdown by lens, and a prediction log. It comes to life during the council meeting process — you can see how the number is born.

What this is for

Geopolitics and elections, forecasts on releases and regulatory decisions, macro and crypto — and business questions where you need an honest assessment of chances, not just confidence.