Extended Inertia: The Loophole in Laplace's Demon and the Case for Leaving the Middle of the Curve

I call extended inertia the hypothesis that every state of the universe follows uniquely from the preceding one, as though inertial motion extended to all interactions. I formalize this hypothesis as a deterministic flow and recall the conditions under which it remains compatible with quantum mechanics in light of Bell's theorem. I then show that two limits on prediction survive even if the hypothesis holds. The first is a predictability horizon that grows only logarithmically with measurement precision in chaotic systems. The second is a loophole in Laplace's demon: no correct prediction of a binary choice can be announced to an agent who responds by contradicting it, although a self-consistent prediction exists when responses vary continuously. Building on these limits, I state a decision-theoretic corollary: when a goal lies above the expected outcome and losses are bounded, strategies that widen the distribution of outcomes raise the probability of reaching it. I propose a tabletop protocol for measuring the predictability horizon and list testable consequences. The results drawn on here are established; the contribution is to connect them within a single framework.