On Keeping Notes

I have tried most of the note-taking systems and abandoned all of them. What survived is simple enough to be slightly embarrassing, which I now think is the point.

Systems fail at the capture step

Every elaborate system I tried failed in the same place. Not at retrieval, not at organisation, but at the moment of writing something down.

If capturing a thought requires choosing a category, applying tags, and deciding where it belongs, then capturing a thought has a cost. When the cost exceeds the immediate value of the thought, which it usually does, the thought does not get captured. The system then contains only the ideas that felt important at the time, which is a biased and not especially useful sample.

The best ideas I have had did not feel important when I had them. They felt like minor observations. A system that filters for perceived importance at capture time filters those out.

What I do instead

One file per subject. Plain text. New thoughts go at the bottom with a date.

That is nearly the whole system. There is no tagging, no linking, no hierarchy beyond the filename. When a file gets long enough to be unwieldy, I read it through and split it, which is also how I reread things.

The lack of structure is deliberate. Structure imposed at capture time is structure imposed before you know what the material is. Structure that emerges when you split a file is structure derived from the material itself, and it is generally better.

Rereading is the actual work

The mistake I made for years was treating notes as storage. Write it down, it is safe, move on.

But notes that are never reread are functionally identical to notes that were never written. The value is not in having recorded something. It is in encountering it again later, in a different context, when it connects to something it could not have connected to at the time.

I now reread one file a week, chosen more or less at random. Most rereads produce nothing. Occasionally one produces the connection that turns two half-formed observations into something worth writing about properly.

The unglamorous conclusion

The systems I abandoned were all better designed than what I ended up with. They had genuine advantages, and if I had used them consistently they would have outperformed a folder of text files.

I did not use them consistently. That turned out to be the only variable that mattered, and it is the one that system design cannot fix.