Main idea

Three documents take shape naturally around a project: the card template to fill in, the method document, the working instructions given to the agent that executes. One and the same rule — "a gap never modifies the body of the card" — ends up written in all three, each time with slightly different words, and each copy was made out of a desire to do things properly.

As long as nothing changes, nobody suffers: all three say the same thing. The day the rule evolves, you fix one, forget the other two, and three documents contradict each other with nobody knowing which one is authoritative. Duplication costs nothing when you write it and everything at the first modification.

An AI resists it less well than a human: it reads all three without seeing that they diverge, and bases its analysis on whichever it opened first. Hence the opposite discipline — each rule lives in a single place; the card template describes only the form, which fields and which allowed values, and refers to the method for everything else.

Why it matters

This gives you a writing test you can apply at the very moment you are tempted to copy: the question isn't "is it useful here?" — it almost always is — but "where will this rule live when it changes?".

The irony often turns back on this kind of project: you build a reference model meant to eliminate duplicated definitions while duplicating its own rules.

Nuances and limits

Duplication that is deliberate and flagged is a different case: in an object card, the links are written twice — once in the header for tools, once in the body for the network view — and that exception is written down as such.

And the cross-reference has a reading cost: a reader who has to open three documents to reconstitute a rule ends up copying it into their own notes.

Open questions

• Where do you house a rule that concerns both the form of an artefact and the method that produces it, without cutting it into two unintelligible halves?