The first reminder came from a company called Alpine Data Labs, whose product I've been looking at. The core of their software is a graphical interface to hadoop computations, where you put them together as a dependency graph. My first instinct was "oh, it's just a GUI wrapper around the Pig programming language, with some built-in tools". But the more I thought about it, I realized that graphical interfaces really are a different and truly more natural way to think about data processing problems. In previous work that i've done with other companies we ended up having to jury-rig graphical interfaces to visualize the structure of our code, but Alpine decided to have the interface be a priori graphical, and I think they hit the nail on the head for use cases where you don't need fine-grained control.
The second reminder was in putting together some power point slides. I generally write slides that are text-heavy, and justify it by pointing out that my content is fundamentally textual. But I realized that even if the core content of my slide is text, there is always a structure to that text. It could be bulleted lists, sequential steps in a workflow, etc, but in almost all of these cases there is a graphical way to view the content. Even if my core message is just ASCII, the correct way to convey the structure of my thoughts is through a picture.