This article examines the role of visualisations in astrophysics programming work, showing that visualisations are not only outputs for those producing them, but can help those developing them understand how to do their work. Studies of visualization in programming have mainly been of social and cultural factors influencing scientific research. We concentrate on the material aspects of scientific work, as of interest in their own right and on methodological grounds (since capturing the material practices of computer screen-work is an underexplored area). Using a ‘video-aided ethnographic’ method we analyse an episode of computational astrophysics involving the use of the Python programming language. We identify a selection of activities comprising the screen work of an astrophysics researcher to unpack how those activities contribute to the production of scientific knowledge.