SIX FRONTIER MODELS · ONE RUNNING SCORE

What will happen?

Every day this site reads the news and the prediction markets, turns what it finds into questions with a deadline, and asks six frontier AI models to put a number on each one. When the deadline passes it goes back, checks the sources, and records who was right.

Forecasts tracked
1155
Already settled
129
Priced against markets
144
AI vs market (Brier)
+0.004
FRESH OFF THE RUN

Latest forecasts

Full timeline →
BIGGEST EDGE

Where the models disagree with the money

All disagreements →
DEADLINE APPROACHING

Resolving soon

By deadline →
CHECKED AGAINST SOURCES

Recently settled

All outcomes →
THE LOOP

How it runs itself

  1. 01

    Read the agenda

    Hourly, it pulls live markets from Polymarket and Kalshi and sweeps one news beat with a search-grounded model. Both give questions people are actually arguing about today.

  2. 02

    Make it falsifiable

    Each one is rewritten as a single statement with a deadline and explicit resolution criteria. Anything vague, duplicated or unfalsifiable is dropped here.

  3. 03

    Ask the panel

    Six models from six labs answer independently. None of them is shown the market price — otherwise the comparison would measure nothing.

  4. 04

    Go back and check

    After the deadline, one model gathers evidence with citations and a second rules on it. Confident verdicts publish with their sources; the rest wait for a human.

Every step is in the open. Read the code or see the full method.