This paper explores the advantages and highlights the potential limitations of using off line information to generate approximations of dynamic systems. The procedure is presented in the context of existing approaches motivated by studies in chemical kinetics. Store and retrieve approaches utilize previously generated information about the kinetic evolution of the reaction system in order to build implicit approximations. The unstructured data reside in memory and computational algorithms, detecting nearest neighbors, are used to define a region around a novel query point. The properties of the query point are determined by linear regression of the stored information. The examples we considered are designed to address the merits of the approach as well as the fundamental issues that warrant further research. (C) 2004 Elsevier Ltd. All rights reserved.