Operational forecasting in the mid 1950's differs somewhat from current methods, both in technical detail and in basic philosophy. The origin of many of the detailed differences in, as well as much of the general context of, current Numerical Weather Prediction can be found in the 50 years of contributions by Lorenz. This talk provides an overview both of predictability present and of the near term goals of current research (in programs like THORPEX); it aims to identify early insights into concepts which are now largely taken as obvious. The perspective is that of the generation which grew up knowing the advances in nonlinear dynamical systems made in the 80's, who focused on predictability of the real world in the 90's while retaining an interest in simple mathematical models. If an insight of the first kind consists of a good idea which falls along the trajectory of progress, then perhap insights of the second provides the background against which future ideas are judged: the climate which suggests which problems should be solved as well as pointing towards how one might approach them individually. The talk concludes with an attempt to discover insights in this contribution to the literature which will only come to be considered obvious later in this century.