This paper focuses on the organization of design processes and the difficulty of simultaneously achieving control and exploration while aiming to achieve radical innovation. After a first generation of works that tended to oppose NPD processes (with controlled convergence and very limited exploration) to Innovation processes (with poorly controlled convergence and random (uncontrolled) exploration, the new generation of works proposed ways to combine control and convergence either through concept shift or through stable architectures. Relying a generic analytical framework (design space / value management) it appears that each model makes restrictive hypotheses (respectively smart leadership or stable architecture) to address two critical questions: Q1. How can one increase the efficiency of exploration? Q2. How can one ensure forms of cumulative convergence? Relying on the ame analytical framework we analyze two cases that explore the unknown in a controlled way and still don't correspond two either of the two models. We show that these two anomalies and the two models actually have two critical features in common: a focus on generative constraint and a logic of cumulative design rules. As a consequence these two features might generic to several processes where teams have to explore the unknown and still have to keep a rigorous control of exploration and convergence.