For decades, one of the overarching objectives of self-assembly science has been to define the rules necessary to build functional, artificial materials with rich and adaptive phase behavior from the bottom-up. To this end, the computational and experimental efforts of chemists, physicists, materials scientists, and biologists alike have built a body of knowledge that spans both disciplines and length scales. Indeed, today control of self-assembly is extending even to supramolecular and molecular levels, where crystal engineering and design of porous materials are becoming exciting areas of exploration. Nevertheless, at least at the nanoscale, there are many stones yet to be turned. While recent breakthroughs in nanoparticle (NP) synthesis have amassed a vast library of nanoscale building blocks, NP-NP interactions in situ remain poorly quantified, in large part due to technical and theoretical impediments. While increasingly many applications for self-assembled architectures are being demonstrated, it remains difficult to predict-and therefore engineer-the pathways by which these structures form. Here, we describe how investigations using liquid-phase transmission electron microscopy (TEM) have begun to play a role in pursuing some of these long-standing questions of fundamental and far-reaching interest. Liquid-phase TEM is unique in its ability to resolve the motions and trajectories of single NPs in solution, making it a powerful tool for studying the dynamics of NP self-assembly. Since 2012, liquid-phase TEM has been used to investigate the self-assembly behavior of a variety of simple, metallic NPs. In this Account, however, we focus on our work with anisotropic NPs, which we show to have very different self-assembly behavior, and especially on how analysis methods we and others in the field are developing can be used to convert their motions and trajectories revealed by liquid-phase TEM into quantitative understanding of underlying interactions and dynamics. In general, liquid-phase TEM studies may help bridge enduring gaps in the understanding and control of self-assembly at the nanoscale. For one, quantification of NP-NP interactions and self-assembly dynamics will inform both computational and statistical mechanical models used to describe nanoscale phenomena. Such understanding will also lay the groundwork for establishing new and generalizable thermodynamic and kinetic design rules for NP self-assembly. Synergies with NP synthesis will enable investigations of building blocks with novel, perhaps even evolving or active behavior. Moreover, in the long run, we foresee the possibility of applying the guidelines and models of fundamental nanoscale interactions which are uncovered under liquid-phase TEM to biological and biomimetic systems at similar dimensions.