I began in a similar way to most old school Dungeon Masters when I started playing Labyrinth Lord over a year ago, in that I would use just a few notes and go from there, expanding them in play to bring them to life, but have since allowed my approach to drift and so have now developed my own style.
When I began, I ran some unconnected adventures, some of which were really bare bones and some of which were fully fledged boxed or printed products. Sometimes I would even run random dungeons. I connected them together very loosely so the campaign emerged as a result of play, rather than being structured up front. However, when my players hit 4th level, I felt it was time to shift things up a gear. So, I began thinking about a campaign, one that wasn't simply emergent, but had a beginning, a middle and possible endings.
I now run a campaign of very tightly connected adventures and quests. However, because I am still writing the campaign as we play it, week by week, I can make adjustments as necessary based on what the party does. I find this to be a very effective technique and would recommend it. I am 150 sides of A4 into it so far, without including the maps themselves, and there's yet more to come.
I'm finding a great approach that works for me is to begin with just a map. Sometimes I make one myself and other times I pick one from a published book. I then use only the names of places, how the bit looks on the map itself and the context within my own campaign as a starting point for what will actually be there. This frees up my imagination, helping to unhook it from game mechanics and standard roleplaying conventions, letting me think more clearly. I can then begin thinking properly about what should be there, why the location exists even exists in the first place, what (if anything) populates it... and so on. All the while I bear in mind the quest... but because the map wasn't actually designed to support the quest in the first place it ends up having little bits here and there that exist because they do and not because they were created to support the quest itself. I find that players respond to this quite well. Sometimes they will spend time trying to figure out something as though it is a puzzle when it isn't - I put this down mostly to the way computer games are designed (like a corridor with nothing to do but go onwards and every obstacle has to be overcome in order to keep going), something I do not want to replicate! However, the areas feel interesting, there are mysteries that never get solved and the places feel more real.
We're having a blast and I thought I would share this approach with you, because if you're anything like me, you might enjoy learning about other styles of Dungeon Mastering that work a treat.