One theme of OB and LW has been to take the fuzzy complexity of the real world and show that in principle at some level it's related to something precise. We can't actually do Bayesian math in our heads for real-world calculations, but just knowing how to work the ideal case protects us against certain real-world errors. Likewise, Eliezer's meta-ethics reduces morality down to some horrendously complex thing that we can never calculate, but it's nice to know that morality does reduce to something when we're wondering whether it exists at all, whether it's all relative, or so on.

The real positivists thought they could reduce all language to their positivism and spent thirty years trying. I don't think I'm going to do that in a few days of posting about stuff on a blog. But if I can sketch a few really-large-scale things and then let people's common sense fill in the blanks, that'll still be better than nothing.