What the site does
For each pair and month-end target date the site shows three probabilities per threshold (1%, 5%, 10% and 50%), separate estimates from OpenAI and Claude, a statistical reference built from price history, the evidence and citations behind the model estimates, and matching prediction-market contracts where they exist.
Forecasts are published as frozen versions with their own reference price and timestamps. After each target date the outcome is recorded and scored on the public track record.
What it is not
- Not investment advice, and not a trading platform: there are no accounts, wallets or trade buttons.
- Not a guarantee: probability estimates can be wrong, and the model estimates are uncalibrated until enough outcomes exist to test them.
- Not written by human analysts: explanations are generated from a curated evidence pack and checked against their sources.
Who operates it
Operated by [Company legal name] — supplied later, [registered address and company number] — supplied later.
Demo data This site currently runs on synthetic fixtures for design and testing; none of its numbers are real forecasts.
Learn more
Read how the forecasts are made, compared and scored, the data sources and the disclaimer.