Emergent phenomena share the fascinating property that they are not obvious consequences of the design of the system in which they appear, a characteristic equally relevant when attempting to model them. We describe several systems that exhibit surprisingly rich emergent behavior, each studied by molecular dynamics (MD) simulation. Modelling self-assembly processes relevant to virus growth reveals the ability to achieve complete, error-free assembly, where, paradoxically, high yields are due to reversible bond formation. In the case of fluids studied at the atomistic level, not only can complex hydrodynamic phenomena in rotating and convecting fluids -the Taylor-Couette and Rayleigh- Benard instabilities - be reproduced within the limited length and time scales accessible to MD, but there is even quantitative agreement. Finally, studies of granular mixtures show behavior that, in the case of a rotating drum, reproduces the familiar counterintuitive axial and radial segregation, and in the case of a vertically vibrated layer, indicates a novel form of horizontal segregation. While there are limitations to the MD approach, both conceptual and computational, the present results offer a tantalizing hint of what can be accomplished.