The talk begins by outlining briefly the process that led to the discovery that optimal government policy generally is inconsistent over time. Essential to this property is that households and businesses behave in a forward-looking way within a dynamic environment. Investment decisions of various sorts are prime examples of such behavior. From there, the next step was to put models with people to use to evaluate the roles of candidate sources of business cycles. The talk then goes on to highlight some of the implications for research today. Among the issues is the importance, as a basis for any research paper, of a well-articulated question, which helps to dictate what the appropriate model economy is for the purpose of addressing that question. As the focus is on quantitative aggregate economics, model calibration plays an important role. Calibration can be challenging, but also at times surprisingly transparent once the decision problems of the model’s people are explicitly stated. In reporting one’s findings, one is generally aided by the important tool of the computational experiment, that is, the act of placing the model’s people in the desired environment and recording their behavior over time. In many cases, there are still puzzles or anomalies in the data relative to the theoretical framework. Often, these easily give rise to interesting questions for future research. Some of the ideas in the talk will be illustrated based on a recent working paper.