NotesNudge the notes you never go back to

Titles, Tags and Retrieval

Everyone who accumulates notes eventually builds a tagging system, and almost everyone abandons it. The usual diagnosis is inconsistency — you tagged things differently over time, the vocabulary drifted, you stopped bothering.

That is what happens, not why. The reason is that tagging solves a filing problem, and what you have is a retrieval problem, and they are not the same.

For a workplace-oriented comparison with this kind of workflow, read more provides additional context.

The asymmetry

At capture you know what the note is about. That is why tagging feels natural: you are labelling something whose content is in front of you.

At retrieval you do not know what the note is about. You have forgotten it exists. What you have is a problem — a question, a difficulty, a half-formed connection — and you are hoping something in the archive addresses it.

So you tagged by content and you are searching by need. The two vocabularies rarely meet, and the more comprehensive the tagging, the more confidently it fails, because a large tag vocabulary is harder to search than a small one.

For an independent external reference related to this topic, see Obsidian.

Why titles do better

A title is read at retrieval time, in a list, while scanning. It has to work for a reader who does not remember the note. That is a much more useful constraint than a label chosen while the content is fresh.

Title with a claim, not a topic.

Notes on pricing → Small budgets produce better constraints than large ones Zettelkasten thoughts → The Luhmann story is a sample size of one Meeting with A → A wants the report weekly, not the dashboard

A claim is scannable, evaluable, and — the part that matters here — it gives you something to attempt recall against. You cannot try to remember a topic label.

When tags do work

Not an argument against tags entirely. They earn their place in three narrow cases.

Status, not subject. To read. In progress. Superseded. Small closed vocabulary, unambiguous, and it changes over time, which is exactly what a tag is good at and a title is not.

Project membership. Which piece of work this belongs to. Also closed, also temporary.

The marked set. One tag for notes you want in review rotation. One tag, binary, no vocabulary to maintain.

The pattern: tags work when the vocabulary is small, closed, and about the note's state rather than its content. They fail when they try to describe subject matter, because subject vocabularies are open, drift, and are unknown at retrieval time.

What actually finds things

In rough order of how often each works.

Full-text search — increasingly semantic rather than literal — if the note contains the words you would use for the problem. Which is an argument for writing the why-sentence at capture — that sentence is written in problem language rather than in content language, so it is what search will match.

Scanning a title list, which is why claim-titles pay off.

Links from a note you did find, which is the case for connecting notes explicitly.

Something surfacing on its own, which is rediscovery rather than retrieval and is the only mechanism that works when you did not know to look.

Tags, last, and mostly for status.

The practical minimum

Every note gets a title that states something. Thirty seconds, and it is the highest-leverage capture decision after the why-sentence.

One tag for the marked set. One for status if you need it. No subject tags.

And write the why-sentence in the words you would use to describe the problem, not the words the source used. That single choice determines whether search finds it.

The thing to stop doing

Do not build a taxonomy. It is enjoyable, it feels like progress, and it is the comfortable half of the pair — organising is fluent, reviewing is not.

A taxonomy is maintained forever and consulted rarely. Search plus good titles is maintained never and consulted constantly.

The short version