Usually, I tend to meet a problem, research it, and (hopefully) eventually solve it, and then I finally think about maybe writing up a blog post about it. The problem with this approach is that I then have to go back and remember the entire process and how I figured out what eventually worked, especially if I want to provide links in my post, which I usually do. This also lends itself towards extremely long or multi-part posts, which rarely get the followup posts (see my minecraft server post for example).
This time, I'm going to try writing a post while I'm in the early stages of being stuck looking at a new problem. It might be a less comprehensive stand alone post or a series of posts that don't work unless you read them all, but it might also be something that is easier for me to actually complete, so it's worth a shot. If I don't like it, I'll go back to my regularly scheduled long post every couple of years.