Akio Arakawa divided the general history of numerical modeling of the atmosphere into a prelude and three phases. The beginning of the third phase, in the early 1990's, roughly corresponds to the development of coupled atmosphere-ocean general circulation models (CGCMs). From that beginning, researchers have reported that CGCMs stubbornly refuse to simulate key features of the climate in the eastern part of the tropical oceans. Adding to the frustration, different modeling groups sometimes report dissimilar effects of similar model changes. We start this talk by characterizing the simulation problems that are collectively referred to as the “double ITCZ syndrome of CGCMs”. Next, we review the methods for coupling atmosphere and ocean GCMs, and the issues that arise due to the generally different model grids. We then focus on the southeastern Tropical Pacific, where past and future field campaigns are contributing to a better understanding of the physical systems, thus setting an ideal testbed for CGCM verification. We discuss the ways in which the CGCM difficulties have challenged our confidence in their intraseasonal predictions, present a possible multi-scale modeling approach for improved simulation and prediction of tropical climate variability, and review several scientific and technical issues that will have to be addressed in the approach implementation.