The core argument: physics describes reality as a discrete, finite-information, probabilistically-evolving state machine (Planck + Bekenstein + Bell). LLMs are the same class of mathematical object. We built one large enough to simulate thought, and it worked. The clockwork view predicts this shouldn't work. The probabilistic view predicts it must. The component ideas come from Wolfram, Wheeler, Lloyd, Jaynes, and the Platonic Representation Hypothesis from MIT. What I couldn't find anywhere is the specific synthesis: LLMs as empirical evidence for a probabilistic ontology. I searched extensively. If someone has already written this, I'd genuinely like to read it.