The low-temperature dynamics in amorphous silicon occurs through a sequence of discrete, activated events that reorganize the topology of the network. In this review, we present some recent work done to understand better the nature of these events and the associated dynamics in a-Si. Using the activation-relaxation technique (ART), we generated more than 8000 events in a 1000-atom model of a-Si, providing an extensive database of relaxation and diffusion mechanisms. The generic properties of these events, such as the number of involved atoms and the activation energies, were investigated and found to be in agreement with experimental data. As it turns out, the bond-transposition mechanism proposed by Wooten, Winer and Weaire (WWW) some time ago plays an important role in the events generated by ART. We have therefore turned to an optimized version of the WWW algorithm to generate the best overall configurations of a-Si available today. We discuss the details of the optimization and present the structural and electornic properties of the resulting models.