Paper — Calibrated World Models for AI Agents

A working paper laying out the SF thesis: prediction markets aggregate structured probabilistic beliefs on future events in near-real-time, in a way no other public signal does. The paper formalizes the world-model interface, measures ground-truth fit against held-out settled markets, and discusses calibration drift between model-released-in-month-N and world-state-in-month-N.

Licensed CC-BY-4.0. Draft version — comments welcome at patrick@simplefunctions.dev.