Zettelkasten and evergreen notes

Niklas Luhmann was a German sociologist who spent most of his career at Bielefeld and died in 1998. For roughly four decades he kept a wooden card file, a Zettelkasten, literally a box for slips of paper. The figures repeated throughout the Zettelkasten literature put it at around 90,000 index cards, and his published output at more than 70 books and several hundred articles.

Those numbers are worth holding loosely. Retellings disagree on the article count, the book count depends on whether you include volumes assembled from his material after his death, and the cards are still being catalogued by the archive that holds them. Take them as an order of magnitude: one person, one box, an output difficult to explain by ordinary diligence.

The interesting question is the mechanics, not the volume. Ninety thousand cards sorted by topic would be unusable, because every card would sit under one heading chosen years before you needed it, and finding anything would mean remembering which heading you picked. Luhmann's box was not sorted by topic at all.

A note holds one idea, written in your own words, and stated completely enough that you can read it in five years without the source in front of you. That is the whole rule, and it is usually mistaken for a rule about tidiness. It is not. It is a rule about addressing.

A note containing six ideas has one address. Anything that wants to point at the third idea has to point at the whole note and hope the reader finds the relevant part again. Split the same material into six notes and you have six addresses, and the number of connections the system can express grows with the number of distinct things you can point at, not with the number of words you have stored. Atomicity is what makes the link the unit of value instead of the document.

The practical test is the title. If you can give a note a title that states its entire content as a claim, it is probably atomic. "Enterprise buyers evaluate single sign-on before they evaluate features" is a note. "SSO" is a folder pretending to be a note. Matuschak makes the same point as a writing habit: prefer note titles that are complete phrases, because a title you can state as a claim forces you to have decided what the claim is.

Atomicity has a second payoff that shows up years later. A note holding one idea can be rewritten when your understanding changes, without disturbing anything else. A note holding six can only be replaced, which in practice means it never is.

Luhmann gave every card a number, and once assigned it never changed. A card continuing the thought of card 21 became 22. A card branching off 21 became 21a, and a branch off that became 21a1, alternating numerals and letters. Because the scheme alternates, there is always room to insert a new card between any two existing ones: a thought that arrived in 1974 could be placed directly behind a card written in 1962 without renumbering anything.

This is the Folgezettel, and the point of it is that the address carries no meaning. It says where a card sits in a sequence of thoughts, not what the card is about. Topics were reached two other ways: through a keyword index that held only a few entry points per term, and through the links Luhmann wrote onto the cards themselves. You entered the box at one of a handful of doors and then followed connections. The structure was in the links, and the numbering existed only so that a link had something stable to point at.

In a digital system the Folgezettel is mostly obsolete. A filename is already a stable address, backlinks are computed for you, and full-text search reaches every card at once, so the machinery for inserting a card in the right physical drawer solves a problem you no longer have. Tools differ mainly in how they represent links and store files, which is the subject of Obsidian, Logseq and SiYuan.

What survives is the principle: do not let where a note lives be how you find it. If numbering appeals to you anyway, the idea has a modern descendant in Johnny Decimal, which uses fixed numeric addresses for the stability reason Luhmann did and works better for archives than for live thinking, covered in Johnny Decimal and digital gardening.

Sönke Ahrens's How to Take Smart Notes (2017) is the book that made the method legible to English-speaking readers, and its most immediately useful contribution is a distinction that most note systems collapse. There are three kinds of note, they behave differently, and confusing them is the usual reason a system stops working.

The flow runs in one direction: fleeting notes become literature notes, literature notes provoke permanent notes, and the permanent notes are the knowledge base. A system holding only the first two kinds looks productive and produces nothing, because nothing in it is yours. Most abandoned note collections are exactly that: a large, well-organised pile of other people's sentences.

Ahrens's larger argument is worth stating separately, because it reframes the effort. Writing is not the step that comes after thinking. It is the medium in which thinking is done and the only place where a vague understanding gets tested against a sentence that either works or does not. Notes are the residue of that, not the point of it.

Andy Matuschak, a researcher and programmer who publishes his working notes openly, uses the term evergreen notes for notes written to be revisited and developed over time rather than filed and left. It is the Zettelkasten with the apparatus removed: no card box, no Folgezettel, no fixed pipeline of note types. What is left is a set of properties a note should have.

• Atomic. One concept per note, for the addressing reason above.
• Concept-oriented, not source-oriented. The note is organised around an idea, not around "notes on chapter three" or "things from the conference". This is the principle that most visibly changes what a folder of notes looks like, because it removes the container named after a source and forces every claim to stand alone.
• Densely linked. Making links is the act that makes a note findable later. A note that links to nothing has been written into a void, and a system with few links is a filing cabinet with extra steps.
• Written for yourself. Not publishable prose. Comprehensible to you in five years, which is a much lower bar than publishing and a much higher one than a bullet point.
• Accreting. A note is never finished. Returning to one and rewriting it as your understanding changes is the work, not maintenance overhead on the work.

The line that gives the whole approach its direction is Matuschak's observation that better note-taking is the wrong goal, and better thinking is the right one. It is a useful thing to keep nearby, because almost every hour lost to a knowledge system is an hour spent improving the note-taking.

The two are not competing systems. Evergreen notes are what the method looks like once stable filenames, computed backlinks and instant search have removed the problems the cards were solving.

Very few people fail at this because they picked the wrong tool. Three failures account for most of it, and all three are habits rather than configuration.

• The collector's fallacy. The term was popularised on zettelkasten.de, and it names the feeling of having learned something that comes from saving it. Downloading the paper, clipping the article, highlighting the chapter. Collecting takes seconds and reading takes hours, so any system optimised for capture fills at a rate nobody can process, and the archive quietly becomes a record of intentions rather than a store of knowledge. The diagnostic is blunt: pick a note at random and say what is in it without opening it. If you cannot do that for most of your collection, you have a reading list.
• Notes that are quotes. A highlight records that a sentence impressed you. It does not record why, and the reason is the part that decays. Ahrens's rule, restate the source in your own words, is not a style preference: rewriting is the step where comprehension gets tested, and skipping it means you discover you did not understand the argument years later, if at all. Quote a line at most, and write the sentence around it as your own claim.
• Structure invented before there is anything to structure. The most common way a new system dies is a first week spent designing a folder hierarchy, a tag taxonomy and a template set for notes that do not exist yet. That structure is a prediction about material you have not gathered, and it will be wrong. Both traditions push structure to the end. Luhmann's index grew out of the cards. Nick Milo's Linking Your Thinking names the moment to act as the mental squeeze point, when a theme has accumulated more notes than you can hold in mind, and that is when a map of content gets written: see LYT and maps of content.

A fourth failure comes from taking the advice too literally. Dense linking does not mean linking everything to everything. A link that means "these are both about design" carries no information and dilutes the ones that do. Write the reason into the sentence containing the link, and if you cannot write a reason, there is no link.

It is also worth being clear about what the method is not for. A Zettelkasten organises ideas by their relationships; it is a poor place to keep the things you have to do this week. Tiago Forte's PARA organises the same material by how actionable it is, a different and complementary question, and the two coexist happily in one vault: see PARA and CODE.