Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Angle
None of the criteria of atomicity can be checked on the note that carries them. Cohesion is observed when you try to link the note and cannot say what the link designates; reformulation is observed when a reader can argue with the statement without reopening the original text; granularity is observed when the network stays readable; filing by concept is observed when a second reading settles into the existing note instead of creating a twin; and the value of the whole is observed the day an article gets written by assembly instead of one long synthesis session. All these checks are deferred and external. That is why demanding atomicity during capture is an error of timing rather than an excessive requirement: you are applying a criterion whose object of measurement — the neighbourhood, the next source, the text to come — does not exist yet. The book summary, which fails all these tests, is not badly written: it is written for one book and never had an elsewhere in which to be judged.
Synthesis
The most widespread misreading of atomic notes is a matter of a misplaced criterion. "One idea per note" gets read as an instruction about length, you cut your thinking into slogan-sized fragments, and you end up with exact notes that nothing connects. The remedy isn't another size threshold: it is to move the point of measurement outside the note.
That move reorganizes rules that looked independent. The title must assert something not out of elegance, but because a contestable statement is the only object another note can oppose; without it, the relation stays unnameable. Size is bounded on both sides because a note that is too broad makes the link ambiguous and a note that is too fine multiplies links to the point of illegibility — the two flaws are a single flaw, measured on the network. Reformulation detaches the statement from its author because reported speech cannot enter a guide, a critique of method or another line of reasoning without your first going back to the source. Three rules, one reason.
Filing by concept is the long version of the same move. A base filed by origin grows with your readings without any of its entries getting stronger; a base filed by concept grows denser, because the second source comes to confirm, the third to qualify, the fourth to limit a note already there. The note is therefore judged not by what it contains at the moment it is written, but by its capacity to receive what it does not yet contain.
There remains the practical consequence, and it is counter-intuitive: the only check that can be exercised while writing does not bear on the note. Getting stuck on its title is the only signal available in the moment, and it doesn't say the note is badly made — it says the idea isn't clear yet, or that there are two of them. Everything else waits to be tested by a link, by one more source, or by a text that leans on it.
Tensions / contradictions
The criterion is right and it is of little use at the moment you would need it most. If everything is verified later, nothing guides the hand of the person writing now, except a negative signal — the title that won't come. The setup therefore requires accepting that you write notes you will only later know held up.
A second tension, with the way these notes actually get reread. Verification by the neighbourhood requires the network to actually be traversed; a base you feed without ever crossing it returns none of these verdicts, and its notes stay indefinitely unjudged while looking correct.
Questions
- Can a deferred criterion be instrumented, that is, measured otherwise than by waiting for the occasion on which it shows itself?
- For how long must a note stay without a link and without an additional source before you conclude it carried no reusable idea?