The paper intends to gather some elements for an epistemological history of French didactics, by underlining conscious and unconscious connections of early French didactical works with both Piaget’s genetic epistemology and Wittgenstein’s analysis of language. The core idea is that French didactics genesis lies in a mix of borrowing, critics and radical transformations of tools, concepts, methodology and theoretical framework of psychology of education. Piaget introduced specific apparatuses (“dispositifs”) to study the genetic development of child cognitive functions. Such apparatuses created artificial environments for a given child, who was supposed to interact with them by using specific cognitive functions. Such a postulated univocal correspondence between one apparatus and one set of connected cognitive functions paved the way for a systematic psychological study of child knowledge or intellectual abilities through his interactions with selected artificial environments. This correspondence has been interpreted by Brousseau as a reversible one; namely as a way for children to learn specific knowledge by interacting with selected “handmade” environments. In the seventies, this interpretation fuelled the building of numerous “didactical situations” (school artificial environments), the interaction with was supposed to trigger pupil learning of specific knowledge. This “learning by interaction with artificial environments” enables teachers to pass over the paradox of the ‘contrat didactique’, first formulated by Brousseau. A paradox that appears to be a reformulation of a Wittgenstein’s linguistic analysis, which states that from learning a given syntax rule one can not infer any knowledge about the class of linguistic situations in which this rule could be used.