Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Main idea
A rereading produces a lot of material that looks like rules without being rules. Translating a book into English, you spot a list of difficult words — matière, acquis, dette, arbitrage, reprise, cadre, état. Each one flags a place where the translation can go wrong. None of them yet says what to do.
Three statuses follow one another and mustn't be conflated. Hesitating over the best translation of matière belongs to the work in progress. A proposal that works in one chapter remains a local hypothesis: it worked there, it commits nothing elsewhere. And when you settle that the concept Contexte will always become Context, with the same capital letter and the same formatting, the statement changes in kind — it no longer describes what was observed, it states what every translation will have to apply.
What moves something from one status to the next is neither repetition nor obviousness: it is an examination that settles the matter. Without it, a happy phrasing becomes a word's official translation throughout the book by sheer inertia.
Why it matters
This is the limit without which a working frame that learns turns into a diary of everything the mission learns. A rule file is worth what its selectivity is worth: if everything gets in, nothing in it is binding any more, and whoever reads it before acting no longer knows what they must comply with.
It also gives a concrete move: keep points of attention elsewhere — a lexicon, a note, the state of the work — and promote only some of them, saying so.
Nuances and limits
The examination isn't always a formal decision. Many choices stabilise through use, and nobody could say when they became mandatory. The distinction then stays useful as an intention rather than as a procedure.
And the opposite excess exists: demanding explicit validation for every term paralyses a translation job that counts dozens of them per chapter.
Open questions
- Who validates, when the work is carried out alone with an assistant, and no third party comes along to contest the promotion of a hypothesis into a rule?