Recent forecast experiments we have been conducting with the WRF model using high-resolution (dx < 5 km), no deep convective parameterization, and very smooth initial conditions, indicate that the model is producing realistic mesoscale and near-cloudscale structures. Examination of kinetic energy spectra and structure functions calculated from the forecasts strongly suggest that the model is producing the correct variance for the fine scales (mesoscale and near-cloud scales). In most cases, however, the forecasts appear very similar to the coarser-resolution forecasts used to provide initial and boundary conditions to the WRF model. A number of problems have also become apparent, including the over-prediction of rainfall and the tendency to predict convection that is too intense and too long-lived. Additionally, higher-resolution tests using idealized squall-line simulations indicate that we are far from statistical convergence using grid-spacings greater than a few hundred meters. These results, taken together, suggest that the high-resolution forecasts can be considered as successful, though expensive, downscaling experiments. Further development of sub-gridscale paramerizations, physics, and assimilation systems are needed to improve the downscaling and to improve the forecasts of the newly resolved structures - to improve on downscaling. The forecasts provide clues to where these improvements need to take place. We will show results from the forecast experiments and our other convective tests to illustrate these points.