These days I'm all scientific method: forming hypotheses, implementing them, measuring results, and the whole deal is a sort of amorphous "this should be better".
I came up with an idea that was sorta inspired by a paper and eventually convinced myself it wouldn't work, but then my boss (what do you call your "team lead"? I hate the term) suggested a modification that did, and it's all obvious in retrospect. I'm happy to have had the seed but I can see there's room for improvement.
There are other jobs here (y'know, hacking GMail javascript) that are more of the first sort of problem but part of the reason I came here is for these "harder" sorts of problems. I will persevere.