So far I have only talked about problems. I believe Cyc falls short both in expressive power and inference efficiency. This means it fails both on epistemological and heuristic grounds. I have tried to argue that the problems were typically due to insisting on some limiting principles like explicit and uniform representation, and relying mainly on deductive inference.
I should make clear that although Cyc seems to have different representations and inference engines at the heuristic level, they are just reimplementations of the general representation and inference framework of the epistemological level. The specific modules are separated on syntax based rather than domain based concerns. They do not implement different frameworks in the sense I will describe below.
I believe that these limiting factors were borrowed from a formalism that was not intended to be a theory of thinking, and buried into the foundational hypotheses by the fathers of AI. In this section I will try to demonstrate what it means to break these limitations.