I'm only now replying to this, since I've only just figured out what it was that I was groping for in the above.

The important thing is not compression, but integration of new knowledge so that it affects future cognition, and future behaviour. The ability to change one's methodologies and approaches based on new knowledge would seem to be key to rationality. The more subtle the influence (ie, a new bit of math changes how you approach buying meat at the supermarket) then the better the evidence for deep integration of new knowledge.