Knowable.
Building prediction infrastructure for humans and agents.
High-stakes decisions deserve better predictions than humans alone can make.
Why now: Frontier LLMs are closing in on superforecaster-level performance, and prediction markets now process over $25B in monthly notional volume — deep enough to grade thousands of forecasts every day. Models can predict. Markets can prove it.
What we have: an ensemble of LLM forecasters that beats both market consensus and frontier LLMs on prediction markets — measured by profitability in live trading and Brier-score wins across thousands of resolved markets in backtesting.
What we're building: a self-improving loop where every new settled market becomes feedback that tunes the next model.