For me, it's always about focusing on proving my idea correct as fast as possible.

I think X would be cool. - How quickly can I test if X is cool with quick and dirty tactics. I tend to use notepad and make notes along the way, so I can flesh it out into something reproducible later. If the idea doesn't work, at least I have a text file of my process and thinking during that time.

If it does work, I find being able to scaffold your projects quickly is key. Build a script that creates a common set of folders and a new git repository. Things like (notes, source, libraries, docs, archives etc). From here, you can just add things to your git repository as you develop/test/build your idea. It doesn't have to be a project that involves code, as you can manage any files from a git repository. They key is having that scaffolding.