Finding better ways to write code that let newcomers make sense of them faster. More details: http://akkartik.name/about
The teaching fits into this in two ways:
a) It seems like a more ambitious test. If I can make codebases easier for non-programmers or inexperienced programmers to understand, then experienced ones should hopefully be easy.
b) It's a way to get feedback. It's hard to find experienced programmers willing to try out a strange new way of writing code that isn't going to be useful in real products for a very long time. Without this feedback I'd be likely to burn out long before I can fully validate or invalidate my hypothesis. But at least for me, teaching is extremely rewarding/addictive.
Oh, there's a third way: since I get paid for my teaching, there's the distant possibility that I might be able to scale up the teaching to fund my research so that I can work on it full-time.
Thanks! I remember seeing it yesterday at 1 point and assuming it was done. Didn't notice when people started upvoting it. Oh well, probably too late now.
