A partial inventory, in order: index cards in a box, a numbered notebook with a separate index notebook, a folder of plain text files linked by hand, three different pieces of dedicated software, and — briefly, out of what I can only describe as desperation — a return to index cards, but laminated this time, for reasons that made sense at 1am and don’t anymore.
Every one of them worked for exactly as long as the novelty of the system was itself interesting to think about, which is a bad sign for a system whose entire job is to get out of the way of the thinking. The honest failure mode wasn’t that any of them were badly designed. It’s that maintaining a system is itself a task, and tasks compete for the same small daily budget of attention that the notes were supposed to be capturing output from.
What’s actually stuck, after all that, is closer to no system: a single running text file per year, appended to in whatever order things occur, searched rather than organized. It offends the part of me that likes taxonomies. It also, unlike every taxonomy I built and abandoned, is still in use three years later — which is the only test a system actually has to pass.