A speculative-but-falsifiable framework treating matter, information, and intelligence as expressions of one measurable quantity — predictive organization — expressible in bits and joules. This package contains the full theory (research. md), a submission-candidate paper, eleven reproducible computational experiments (NumPy/Matplotlib), and historical-data analyses. Key results, all reproducible: (1) finite-horizon predictive information — not asymptotic excess entropy, and not merely the entropy rate — is the substrate-invariant measure of a finite agent's usable prediction (skill collapses onto it at R² = 1. 00) ; (2) that information is literally extractable work, so predictive skill is thermodynamic power — one bit ≈ 2. 87 zeptojoules at 300 K; (3) model-based agency is bounded by an empirically-measured predictive horizon, in discrete, learned, and continuous settings; and (4) suggestively, energy capture and stored prediction co-move across 165 nations today (r = 0. 95), five centuries of book production, and Morris's 16, 000-year record (rho = 0. 95), with energy leading. The measure and the bound are demonstrated at pilot scale and stated falsifiably; the grand unification of matter, information, and intelligence is offered as a labelled conjecture. Inspired by Yann LeCun's world-model / JEPA programme and the unification tradition in physics; neither endorses this framework, and any errors are the author's. Project site: https: //predict. digiphusion. com
K Schomaker (Wed,) studied this question.