In order to meet the increasing challenges of extreme and variable weather and its impacts on water resources planning, we need more tools with better accuracy to predict changing hydrologic patterns. In a series of recent publications [1-3], we and others demonstrate that time series neural network models are capable of predicting streamflow with significant skill. In this work we discuss our experiences in applying these neural network techniques in operational forecasting environments, and for historic streamflow simulations in ungauged basins. We couch this discussion with examples from the year long CEATI Streamflow Forecasting Rodeo in which we competed with other public and private forecasters at 19 difficult sites across the US and Canada. We present our work on creating historic streamflow records in ungauged basins in California, informing state-wide planning and environmental flows efforts. Finally, we discuss collective challenges and exciting new research related to seasonal forecasting and distributed modeling using neural networks. [1] Kratzert et al., (2018). Rainfall–runoff modelling using Long Short-Term Memory (LSTM) networks. Hydrology and Earth System Sciences: 22.11. 6005-6022. [2] Kratzert et al., (2019). Benchmarking a Catchment-Aware Long Short-Term Memory Network (LSTM) for Large-Scale Hydrological Modeling. Hydrology and Earth System Sciences Discussions: 1-32. [3] Kratzert et al., 2019). Prediction in Ungauged Basins with Long Short-Term Memory Networks. Water Resources Research, 55:12. 11344-11354