Kausable

Kausable develops causal foundation models trained to infer underlying structure and dynamics rather than relying on statistical correlation, targeting forecasting and reasoning in complex systems using synthetic causal priors. It shipped TipPFN, a forecasting model for critical transitions, in Q2 2026 after publishing the CausalDynamics benchmark at NeurIPS 2025.

Ecosystem functionThe clearest attempt to replace correlation learning with models that learn reusable structural priors — a genuinely different bet from scaling.

Business modelCausal foundation models