So even though I'm probably 2-ish years from graduating (plus some very large error bars on that number), I've started writing bits and pieces of my thesis. Not just in the sense that I'm writing manuscripts, which I'm also working on, but in that I'm trying to plot out the larger story I'm trying to tell. I also want to right everything down while I still remember why I did it in a particular way.
My advisor seems confused whenever I mention this approach. He's very strongly in the camp of write the papers, and copy-paste to make a thesis. Given his reading habits, it does mean fewer things he has to read, which is good for getting anything out the door. However, at this point, many of my results are of the "and this approach didn't work" variety. Not necessarily paper material, but things I feel should be documented for future students in the group. A quick poll of Twitter seems to indicate that the thesis is a good place to include this sort of thing, and it can always be deleted when you go to do the manuscript. Of course, according to the comments at GenomicRepairman, it seems unlikely hypothetical future students are going to read it anyway.
I also like looking at the bigger picture, because it gives me a sense of things I've *done*, not just the giant list of things I still need to do. When I'm in the writing mood, I'm trying to write up methods sections and working on piecing together background sections as I read. It also helps me plan, and find the holes in my story I need to fill.
What approach did you take (or are you taking) to thesis writing?