Parameterization implementation and coupling issues can be hard to understand (or even identify) based on runs from the fully-coupled GCM. Simulations using GCM code for some components but replacing others with simple idealizations are a useful framework for identifying and fixing coupling problems. By simplifying the model, issues are more easily understood. Because code from the full model is still being used, however, improvements made in the simplified context can be immediately incorporated into the full model. Simplified configurations are also more computationally efficient, which makes them particularly attractive for testing model behavior at high-resolution. These benefits motivated us to use the single-column and kinematic-driver frameworks to examine process coupling and timestep sensitivity in CAM5 cloud parameterizations. Based on this work, we have begun using these simplified frameworks as a testbed for improved coupling and timestepping methodologies. This presentation will summarize our current progress.