Characterization of materials is crucial for the quantification and prediction of their physical, chemical, and mechanical properties. However, as the complexity of a system increases, so do the challenges involved in elucidating its structure. While molecular simulation and modeling have proved invaluable as complements to experiment, such simulations now face serious challenges: new materials are being synthesized with ever increasing structural complexity, and it may soon prove impossible to generate models that are sufficiently realistic to describe them adequately. Perhaps, ultimately, it will only be possible to generate such models by simulating the synthetic process itself. Here, we attempt such a strategy to generate full atomistic models for mesoporous molecular sieves. As in experiment, this is done by allowing nanoparticles to self-assemble at high temperature to form an amorphous mesoporous framework. The temperature is then reduced, and the system is allowed to crystallize. Animations of atomic trajectories, available as Supporting Information, reveal the evolution of multiple seeds which propagate to form a complex framework. The products are polycrystalline mesoporous framework structures containing cavities connected by channels running along "zero", one, two, and three perpendicular directions. We suggest that it is easier to generate these model structures by attempting to simulate the synthetic process rather than by using more conventional techniques. The strategy is illustrated using ZnS as a model system. Further development of the mathematics of minimal surfaces will advance our understanding of these structures.