This talk is an interim report on a journey to enhance scientific programming in Python within the French social sciences. Scientific programming is a stepstone to develop open and reproductible data projects. In numerous field, tools and libraries were developed thanks to Python to become standards. Backed by detailed documentations, those tools diffuse beyond specialties and frontiers and opens new prospects. Nevertheless, social sciences in France seem to stand appart. One reason may be the precedence of R as a programming langage, with highly specialized libraries lacking sometime of documentation or update. Another one may be the diversity of traditions, data and approaches that can hinder shared practices and ease of use of existing libraries. Based on the experience of a small Python library - PySHS - and the organization of training sessions, I will argue that there is a need (and space) for an high-level and easy to use scientific programming framework in Python dedicate to social sciences that can bridge current practices with the state of the art libraries. Such library could in return help to clarify methodologies. This opens two main question that should be addressed : first, how to maintain the flexibility of scientific programming against overtly specialized tools ; second : how to build a meta-framework with other programming langages as R to avoid fragmentation.