A labeling system you will actually keep

Every labeling system dies the same way. You start with a clean taxonomy, you add a label for the edge case, then another, and six weeks later you are staring at 34 labels trying to remember whether this post is "growth" or "marketing" or "acquisition". So you file it under none of them, and the system is over.

The fix is not discipline. It is fewer labels, chosen along the right axis.

Label by use, not by topic

This is the whole idea. Topic labels describe what a post is about. Use labels describe what you will do with it. Only one of those helps you on a Tuesday morning with an empty composer open.

Topics also multiply without limit, because subjects do. "Do with it" does not: there are only so many reasons you keep something.

• Hooks — the first line worked on me. I want the shape.
• Frameworks — a repeatable way of thinking about something. Steps, models, checklists.
• Stories — a personal account I might reference, disagree with, or answer with my own.
• Data — a number or study I would cite.
• Contrarian — a take I disagree with, kept on purpose. These become the best posts you will write.

Five. That is the whole taxonomy, and it has survived on my account longer than any topic tree I ever built.

Why five is the number

Labeling has to cost less than the value it returns, or you stop doing it. Five options is a glance. Twelve is a decision. A decision at save time is where the habit breaks, because saving happens in the two seconds before you keep scrolling.

A taxonomy you abandon in week three is worse than no taxonomy, because now you also have to clean it up.

If you genuinely need a sixth, take one out first. Forcing the trade keeps the list honest, and you will find the sixth was usually a topic sneaking back in.

What about topics?

Search covers them. That is the part people miss when they build the big tree: you do not need a "hiring" label if you can type "hiring" and get every save that mentions it, including the ones you would never have thought to file there.

Labels are for the axis search cannot see. No search query finds "posts whose opening line I want to steal", because that is a judgement you made, not a word in the text. That judgement is exactly what is worth storing.

So: labels for judgement, search for subject. Once you split it that way, the label list stops growing on its own.

Starter set, and how to adapt it

Copy the five above and use them unchanged for a month. Do not customise on day one, because you do not yet know which ones you reach for. Then check the counts.

1. A label with almost nothing in it is not your axis. Cut it.
2. A label holding a third of your library is doing two jobs. Split it once, and only once.
3. A label you have never searched is decoration. Cut it.

Run that check quarterly, not weekly. The point of the system is that it mostly runs itself.

The backlog question

Do not retro-label everything. It is hours of work for a library you have not used yet, and it is the single most common reason people bounce off organising their saves at all.

Label going forward. Label old posts only when a search surfaces one and you think "I will want this again". That way the effort lands exactly on the posts that turned out to matter, which you cannot know in advance.