The development of flexible, transferable problem-solving skills is an important aim of contemporary educational systems. Since processing limitations of our mind represent a major factor influencing any meaningful learning, the acquisition of flexible problem-solving skills needs to be based on known characteristics of our cognitive architecture in order to be effective and efficient. This paper takes a closer look at the processes involved in the acquisition of flexible problem-solving skills within a cognitive load framework. It concludes that (1) cognitive load theory can benefit from putting more emphasis on generalized knowledge structures; (2) there are tradeoffs between generality and power with respect to specific versus generalized knowledge structures; (3) generalized knowledge structures of "medium" generality are essential for flexible expertise; and (4) cognitive load theory could provide a valuable framework for considering essential attributes of flexible expertise.