My idea is to periodically patrol the technical forums (and maybe some other ones too, like Business of Game Development) and look for questions that I can't think of an answer to - or, worse, can't think of how I'd go about researching an answer. Once a few such questions are identified, the idea is to find out how to answer them. Maybe that would involve waiting for other answers; but usually it involves educating oneself about the problem domain and learning a few things.
This idea came to mind again tonight. I've learned something interesting about my ideas: if I deliberately ignore them and they persist in coming back at least once, it's extremely important that I pay attention to them. This is at least the sixth or seventh time I've had this idea, so it's time to start doing it.
But not tonight. Tonight I'm going to go to bed.
As soon as I finish this SICP video. It's interestingly pertinent to my current work (defining and manipulating complex data structures by creating them recursively out of trivial elements). And the "OMFG HEAD ASPLODE" expressions from the audience are just priceless.