This paper aims at assessing the importance of the initial technological endowments when firms decide to establish a technological agreement. We propose a Bertrand duopoly model where firms evaluate the advantages they can get from the agreement according to its length. Allowing them to exploit a learning process, we depict a strict connection between the starting point and the final result. Moreover, as far as learning is evaluated as an iterative process, the set of initial conditions that lead to successful ventures switches from a continuum of values to a Cantor set.Financial support of 2001SGR-00162 and BEC2000-0172 (Xavier Martínez-Giralt) and from the European Community Marie Curie Fellowship (IHP programme) under contract n. HPMF-CT-2000-00855 (Rosella Nicolini) is gratefully acknowledged.