The first numerical model of atmospheric motion was developed during World War I by Lewis Fry Richardson, an English ambulance driver. Unfortunately, his technique retained fast-traveling acoustic waves, and electronic computers had yet to be invented, so his weather forecast was (quite) unrealistic. Since then, atmospheric modelers have made great progress in development of appropriate approximations in the “primitive” equations for use in weather and climate models. Furthermore, the sheer computational power available with today’s supercomputers easily allows integration of these equations on global grids for weeks (weather) and centuries (climate) of model time. However, forecasting is still limited by some of the same issues related to grid spacing, time step, and unresolved processes. In this talk, I’ll start by briefly discussing the basic equations governing atmospheric motion, focusing on the “hard” terms where all the numerical excitement is. I’ll show some of the common strategies (i.e., parameterizations) different modeling centers have used to address these “hard” terms. The proliferation of approaches has led to a range of climate model behaviors in response to external “forcings.” My work centers on understanding (on a more emergent level) why these models differ and (ideally) which are more realistic. The feedbacks between clouds and the general circulation/climate in models are responsible for much of the current discrepancies among model behaviors, but other radiative feedbacks, such as the ice-albedo feedback and water vapor feedback, also play a role and provide clues to the larger system behavior. I’ll end with some of the big open issues currently facing the climate modeling community.