You can, in fact, predict the effect of an atomic bomb without detonating it. Sure, you can't have the same certainty as if you'd detonated it -- empiricism is important; they tested the Fat Man design for a reason. But you may also recall that they were sufficiently confident in the Little Boy design that that one they didn't bother testing, and as it turned out, they were correct in their confidence. That's the power of understanding.

Which is to say, empiricism is important, but if your practice of empiricism is purely looking at what happened before and predicting that things will continue in that vein, rather than using your empirical observations to actually form a model of the world that will allow you to make more general predictions, you're not really using the full power of your intelligence. This is the sort of mistake that led many to not prepare for the covid pandemic, for instance! Meanwhile, others looked at what was going on, realized it wouldn't stay confined like it currently was, realized that it would get much bigger, and prepared.

"I'll believe it when it happens" can be a decent guideline a lot of the time, yes, but at some point you have to actually think. Evidence often comes not in the form of prior similar events, but in the form of models or arguments. To discount these as non-evidence is to blind oneself!