Cernalus is a research initiative exploring how frontier artificial intelligence can be applied to build a general forecasting engine: a single system capable of reasoning about future outcomes across a wide range of domains.

Our approach

Forecasting has traditionally relied on specialised models, each designed, tuned and evaluated in isolation for a single domain. Advances in frontier AI create the opportunity for a more general approach: one engine that can take a well-defined question, gather and weigh the relevant evidence, and produce a calibrated probability.

Calibrated by reality

The value of a forecast lies in its accuracy over time. Cernalus is designed around a continuous cycle of prediction, resolution and assessment. Every forecast is scored against the observed outcome, and those results inform the ongoing refinement of the underlying models. Rigorous evaluation is central to the system, not an afterthought.

Applications

We believe a general forecasting engine has broad application across finance, policy, science, risk and industry: wherever important decisions depend on a well-informed view of what is likely to happen next.

Current stage

Cernalus is an independent research project in the early stages of development.

Contact

We welcome conversations with researchers and collaborators. hello@cernalus.com