This talk will describe a vision of the future of regional weather prediction based on experiences in the Pacific Northwest. Does higher resolution necessarily produce better forecasts? Can high-resolution predictions duplicate local weather features with fidelity? Instead of high resolution should we be running regional ensemble forecasts that will provide probabilistic forecasts and predictions of forecast accuracy? How should weather prediction be organized in the future-centrally or distributed around the nation? These and other questions will be considered in this talk. Many of the conclusions will be based on the multi-year real-time prediction effort centered at the University of Washington in which the MM5 mesoscale model has been run at 36, 12 and 4-km grid spacing. In addition, we have established a regional ensemble prediction system that includes approximately two-dozen prediction, each starting with slightly different initial conditions or model physics. The talk will end by commenting on trends at the National Weather Service and the relationships between regional and national forecasting efforts.