Theories and numerical models of atmospheres and oceans are based on classical mechanics with added parameterizations to represent subgrid variability. Reformulated in terms of derivatives of information entropy with respect to large scale configurations, we find systematic forces very different from those usually assumed. Two examples are given. We see that entropic forcing by ocean eddies systematically drives, rather than retards, large scale circulation. Additionally we find that small scale turbulence systematically drives up gradient (“un-mixing”) fluxes. Such results confront usual understanding and modeling practice.