[Dylan16807]

But is that selection bias relevant here? The tweaks we're comparing against are not lethal ones. The category is genetic spread among living humans, so mutations in living children should be the correct number to use as a baseline.

It's true that that selection bias exists, but so do a hundred other kinds of selection bias in earlier steps. As long we we're not looking at those steps, and any bias in them is already factored in to our number, we don't need to worry about those details.