This paper addresses the epistemological background of the constructivist artificial life (CLife) approach and sketches the cognitive architecture based on CLife. The approach aims at providing a new cognitive architecture which shows how an individual creates his or her own world. This is in a line with Jean Piaget’s “L’ intelligence… organise le monde ens’organisant elle-même”. Furthermore, the architecture maintains the explainability of cognitive processes at a functional level. What is the process by which we make sense of the world? If such constructions are to occur inside an organism there must be some type of separation between cognitive processes going on inside an agent and its sensor and effector surfaces (unspecifity of nervous signals). Thus we must not define apriori categories of perceivable objects and actions an agent can take. No new semantics will ever emerge in such systems. Furthermore, the usage of constructivist concepts implies an expectation-driven behavior: The behavior is controlled by schemata which, once invoked, ask for sensory or internal data only when it needs it: The agent neglects environmental events except for the demands of the current action pattern. The overall algorithm performs the process of assimilation in the Piagetian sense. Schemata have the following properties: (a) They are fuzzy to reflect the fact that in nature are no crisp entities and relationships. (b) By introducing functional couplings, simple schemata can be organized and combined into bigger chunks of knowledge. Couplings provide for increasing multimodal coordination by which an creature constructs the idea of invariant permanent objects.