Take for instance a problem I was having exporting images from the Gamer Mom project I am working on with Mory. I was having to press about 10 buttons to export a single image into a folder and name it correctly. I ended up spending a few hours learning how to write scripts for TVpaint, and figured out how to make it a single button press, to do the exact same export. It changed the whole experience because it was so easy.
Later I am faced with a completely different project. And I run into a new problem, again with exporting from TVpaint, but this time, I can fairly quickly write an even more elaborate script to export images from the software.
Little pieces of experiments past start to add up into bigger and better solutions.
Unfortunately this type of innovation can only happen in tiny steps. Often the knowledge gap between a problem and solution is too great which means you simply won't be able to overcome it.
I've had ideas about writing a new method for animation interpolation, but the gaps are just to big. Though I am now always on the lookout for the paths that I think might lead to the right solution.