What he’s talking about here is learning, right? There are two things - learning, and inferring. When learning, the aspects you and Max mention are probably quite important (I don’t know, what do empirical results say?!). When inferring, you’re not learning, unless of course you get your inference wrong, then someone tells you the answer/their reasoning, and then you go “ahhh I see” and you learn.
I think they anti-correlate. There are fixed resources within a skull, thus a trade-off between average length vs. number of neurites. I think sparse long-range connectivity underlies generalist cognitive bias, and dense short-range connectivity underlies specialist cognitive bias. Although these biases can differ across cortical areas.