With the new DOE computers entering ALCF and OLCF and with exascale computers arriving in the 2020s, we are poised to enter a new era of computing. The capabilities of these new, and subsequent, generations of computers stand to potentially transform wide areas of science. In the first part of this talk, I will first discuss science areas and – at least in my opinion – what are grand challenges that can be addressed with potentially huge benefits to science and to society in general. I will then try to put these challenges in the context of exascale computing, whatever that is, to see how suited they may be for such an approach. I will then attempt another way of looking at exascale computing by trying to think of what general structure of problems and computing approach may be suited for exascale computing, and what hurdles may need to be overcome to make exascale computing not only transformative to science, but also sustainable.

In the second part of the talk, I will describe a recent project. The impetus of the project was to study low-power resistive-switching non-volatile memories. Rather than giving the polished description based on perfect hindsight vision, I will describe the real, and often convoluted and serendipitous, path that the pursuit of science often takes. The outcomes of such enterprises are often not only surprising and (at least sometimes) scientifically valuable, but lead to new collaborations and connections, both in terms of science as well as between people, and new avenues that you had never thought of before. This, as much as some answer to the initial scientific question, is what makes the pursuit of science a whole lot of fun.