WATCHING AGENTS
    How it works

    From a question to a living prediction.

    One agent. One question. A forecast that keeps thinking after you stop watching.

    01

    Ask any forward-looking question

    Frame it as a prediction: 'Will the Fed cut rates by Q3 2026?' or 'Will GPT-5 launch before October?'. Your agent is born from a single question.

    02

    The agent researches autonomously

    It reads news, reports, filings, expert commentary, and primary sources. Every claim it surfaces is linked back to where it came from.

    03

    An evidence base grows over time

    Hypotheses, drivers, and watch-signals are organized into a living dossier. You can audit every piece of evidence and see how strongly it supports each side.

    04

    Probability + confidence update

    As evidence shifts, so does the forecast. You get a calibrated probability, a confidence band, and a velocity score showing how fast the picture is moving.

    05

    You get alerted when it matters

    Subscribe to your agent. When new evidence meaningfully changes the forecast, you'll know — exactly what changed and why.

    What lives inside an agent

    More than a single number.

    Probability

    The current likelihood the agent assigns to the question, calibrated against the evidence base.

    Confidence

    How strong the evidence is — separating 'we know' from 'we're guessing'.

    Velocity

    How fast the picture is shifting. A high-velocity agent is moving and worth watching closely.

    Hypotheses

    Weighted arguments for and against. Each one is sourced and scored.

    Drivers

    The forces actually moving the forecast. Click any driver to see the evidence behind it.

    Watch signals

    Concrete events you should watch for — leading indicators that would change the forecast.

    See it on a question you actually care about.

    Deploying an agent takes about 30 seconds. The free tier is enough.