Computational science and engineering for many years was dominated by physical simulations based on mathematical modeling and numerical analysis. This is reflected in the articles CiSE published, on topics like computational electromagnetics, molecular dynamics for modeling macroscopic material properties, simulations of the earth's mantle convection-driven flow, quantum and molecular mechanics combined for chemistry applications, geographic environmental modeling (these, all in 1995). Scope broadened to areas like computation in medicine (2000, issue 5), data mining (2002, issue 4), and cloud computing (2009, issue 4). Notably, two special-issue themes were harbingers of topics trending today: Python for scientific computing (2007, issue 3), and reproducible research (2009, issue 1). In the first, some of CiSE's most highly cited articles appear, including the Matplotlib paper \cite{Hunter_2007}  with more than 11,400 citations indexed by Google Scholar as of this writing. The theme reappeared in 2011 (issue 2), with an article on the NumPy array structure that has amassed more than 6,000 citations, so far. In more recent years, it has been hard to resist the onslaught of interest in machine learning and related topics. CiSE first covered machine learning in 2013 (issue 5), highlighting applications like materials sciences and climate informatics. The pull of computer science perspectives, versus computational science, began to sway the content before long. Do we need to reclaim focus?