In this paper we propose to revisit two emblematic projects, Manhattan and Polaris, with the models developed by design theory. In particular we demonstrate, relying on C/K theory, how these major projects, traditionally presented as radical innovations, are in fact quite different. In particular we show that the structure of the knowledge base (splitting or non-splitting) has major consequences. This explains the different managerial strategies of this two cases : whereas Polaris focuses on the control of the design process, Manhattan exhibit a very original strategy, characterized by the simultaneous exploration of different solutions, to manage unforeseeable uncertainties. We discuss the implications of this result for design theory and project management.