Extending the core

First, use it exactly as it is

The strong temptation on day one is to personalise the table before a single card is in the box — to delete the disciplines you will never write about, and to add the ones you know you care about. Resist both halves.

Deleting costs you nothing and gains you nothing. An unused discipline is a code you never type. It occupies no drawer space, because guide cards are cut for rows you actually file into, not for rows that exist in a table. Leaving 3030 in the list costs you exactly one line of a web page; deleting it costs you the day, two years from now, when you read something about Saturn's rings.

Adding early costs you a great deal. A tab you cut in advance is a prediction about which subjects will accumulate cards, and predictions about your own future interests are reliably wrong. What you get is a box of confident-looking tabs with two cards behind each, and the real filing decision — which of these near-identical rows — made harder rather than easier.

So: cut the core tabs, file for a few weeks, and let the table be wrong at you. The complaints it generates are the data you need, and you cannot get them any other way.

Then check that you want a discipline at all

Most of the time the feeling there is no row for what I care about is correct and the conclusion is wrong. A discipline is a field of study; what people actually care about is usually a practice, and a practice cuts across four or five fields at once.

Persuasion is not a discipline — it is rhetoric and psychology and advertising and negotiation, and no single row can hold it without becoming a second card box. The answer there is a hub card: one card, filed at the theme's centre of gravity, listing the addresses its material actually sits at. A hub costs one card and no taxonomy.

When a row is earned

A new tab should be paid for by cards that already exist, never by anticipation. Concretely:

• Three or four cards have made the same complaint. Not one card that fits badly — a repeated near-miss, in the same direction, over weeks. One awkward card is a card. Four are a pattern.
• You can name the discipline it keeps colliding with. If you cannot say which existing row nearly worked, you have not found a gap; you have found a card you have not thought about hard enough yet.
• The new row and the old row split on a real question. Ideally the mechanism-or-action question that settles most ties — that split holds up for decades, where a split on vocabulary does not.
• You would cut a physical tab for it today. If the answer is "not yet, there are only two cards", then it is not yet. Note it and carry on.

Until then, the near-miss goes on the card: file it in the closest row you can defend and write what did not fit in the margin. That margin note is the only record of the complaint, and it is what turns four vague frustrations into one obvious new row.

The room you have

A code is drawer + discipline + sub-discipline — one digit, then two, then one. Two digits for the discipline is the scheme's central trade, and this is what it bought: a drawer holds 99 disciplines, and the canonical cut spends 9% of them.

Every discipline gets nine sub-disciplines.

There are also four unused drawers, 6000 to 9000, which you will almost certainly never need, for reasons further down.

Three ways to add a row, in order of preference

1. A sub-discipline under an existing discipline

Take the next free units digit under the discipline it belongs to. 5050 Medicine and health becomes the parent of 5051, 5052, and so on up to nine. This is the move you want, nearly always:

• It cannot collide with anything canonical, because the core list leaves every units digit empty.
• It sharpens a tiebreak that already exists rather than inventing a new one — the parent's scope note already told you what competes with it.
• It survives a change to the core list, because the first three digits are not yours and you did not touch them.
• It reads correctly to anyone who knows the scheme: a non-zero fourth digit is the notation for "a sub-discipline of the row above me".

2. A new discipline in a free tens slot

Legal, and there are 449 slots free, so you will not run out. But understand what you are doing: a discipline is a claim about how the field is cut, and you are now making one inside the numbering space the canonical list uses. Two consequences worth accepting on purpose:

• Mark it as yours. A dot in the corner of the guide card is enough. When you compare your box against the published list a year from now, you want to know instantly which tabs you invented.
• Leave a gap. Number yours from the end of the drawer's used range with a few tens to spare, so a later core addition and your row do not want the same code.

And before you do it: check that what you have is not a sub-discipline of something. Most candidates for a new discipline turn out to be the fourth digit of an existing one, which is the same row with less risk.

3. A new drawer

Four first digits are unused, so this is possible. It is also almost always a mistake. The five drawers are not subject areas — they are five kinds of knowing: what people made, how people behave together, how the physical world works, what holds in any world, and what to do on Monday. A sixth kind is a genuine philosophical claim, and if you have one, you disagree with the core cut rather than extending it.

The one honest use is a drawer that is not knowledge at all — a personal archive, papers, correspondence. Those are not cards in the same box and giving them 9000 is reasonable. Keep them out of the taxonomy either way.

Write the scope note in two halves

A row without a scope note is just a word, and a word does not settle anything at two in the afternoon with a card in your hand. Every row on the core list has a note in the same shape, and yours should too:

The second half is the half that does the work, and it is the half people skip. Writing it forces you to answer the question you will actually face — not what is this row but why this row and not that one — and it is why the table can be read at a glance instead of studied.

Name the competing code explicitly. A note that says "not the theoretical side" ages badly; a note that says "the theory is 4070" is a pointer you can follow when you have forgotten your own reasoning.

What not to do

• Do not renumber a core code. Not to make room, not to close a gap, not for tidiness. Every card that points at a discipline points at the number, and the whole value of the scheme is that a pointer stays true forever. Gaps cost nothing; renumbering costs everything.
• Do not push meaning past the fourth digit. Everything after the slash is position, not classification. If you find yourself deciding that 5051/3 should mean fasting specifically, stop — you are rebuilding the taxonomy inside the branch, and the branch cannot bear it. Why →
• Do not divide by project, source, or date. Those are true of a card for a year and useless for the next thirty. A card's discipline should be the last thing about it that changes.
• Do not add a row for one card. The card is not homeless; it is inconvenient. File it in the nearest defensible row, note the near-miss, and let it wait for company.
• Do not fork the whole table. If you re-cut the drawers you lose the ability to compare notes with anyone using the published list, and you gain nothing that a fourth digit would not have given you.

Keep yours visibly separate

Two tiers, on purpose: the core list can be handed to a stranger, and additions can arrive for years without a canonical code moving. That property is worth maintaining in your own box too — mark your guide cards, and keep a list of what you added and what earned it.

This site does it by tagging every row in one file as core or personal, which is why one page can be canonical while another is openly one person's. The finder searches both, because when you are holding a card you want every row you have, not only the respectable ones.