Main idea

Receiving an answer that matches your expectation demands nothing. Receiving a proposal you find odd — filing in the zone of stable things a translation guide written the day before and still in motion — forces you to say why it bothers you. And to say that, you have to pull out the rule you thought you knew, write it down, hold it up against the case.

That is the moment the operation turns around. Looking for the flaw in the other side's reasoning, you find the flaw in your own: the definition mobilised to object — the stable zone is the zone of what doesn't change — doesn't survive the object it was meant to keep out. The discomfort was sound, its stated grounds were not.

So what the proposal produces isn't the right answer, but the obligation to make things explicit. An answer that goes without saying leaves the tacit definition intact, and a tacit definition can neither be corrected nor passed on.

Why it matters

This gives value to proposals that are ordinarily dismissed in a second. The usefulness of an interlocutor — human or assistant — isn't measured by the soundness of what they propose, but by what their proposal forces the receiver to state.

It also suggests a simple move: faced with an output that looks wrong, write down the rule it violates before rejecting it. It costs little, and it is the only way to discover that the rule is fuzzy.

Nuances and limits

The discomfort may be nothing but habit, in which case making things explicit confirms what you already knew — no benefit, time spent.

The mechanism also wears out through repetition: an interlocutor who constantly proposes odd things stops being examined, and the rule they might have brought to light no longer surfaces.

Open questions

• How do you sort which proposals deserve examination when an assistant produces several per session and none of them carries a signal of seriousness?