This chapter provides an overview of the most powerful practical tools developed so far, and under development, in the areas which the Engineering Directorate of National Science Foundation has called “cognitive optimization and prediction”. It deals with a condensed overview of key tools and discusses the historical background and the larger directions of the field in more narrative terms. The chapter discusses how these tools compare with older tools for neurocontrol, which have also been refined and used in many applications “Cognitive prediction” refers to prediction, classification, filtering, or state estimation under similar conditions. The most powerful and general new methods are adaptive, approximate dynamic programming and neural model-predictive control. Time-lagged recurrent networks have also been very successful, under a variety of names, in many other time-series prediction applications. Cellular Neural Network Chips (CNN) provides the best practical access to these kinds of capabilities. CNNs have been produced with thousands of processors per chip.