Main idea

A state of work written at the end of a session is almost always written by someone who already knows everything. They note what strikes them as notable after six hours inside the subject — so what is new, surprising, still warm — and they leave out what has become obvious, which is precisely what you need to know in order to resume.

The instruction that corrects this bias fits in one question, and it has to be asked from someone else's position: if an assistant arrived right now having read nothing, what would it need to know to pick up without error and without starting over? It isn't a summary of the session. It is an inventory of what a newcomer would lack — where the work stands, what has been settled and is no longer up for discussion, what remains open, what the first action is.

The shift in point of view does all the work. Writing for yourself produces a memory aid that only functions as long as you still remember; writing for a stranger produces a document that still functions three weeks later, when you have become that stranger yourself. And the question has the advantage of being testable: you can hand the state to an assistant that has nothing else and watch whether it sets off correctly.

Why it matters

This gives a writing instruction where usually there is only an intention. "Note where we're at" doesn't say what to write and produces unusable states; the newcomer's question settles, line by line, what has to appear.

It also repairs the flaw through which resumption states fail silently. Their author can't judge their quality — they hold the whole context in their head, so everything seems sufficient. Writing from a third party's ignorance is the only position from which the gap becomes visible.

Nuances and limits

The exercise isn't free: it asks you, at the end of a session and at the moment you are least available, to deliberately forget what you know. Dashed off, it falls straight back into being a memory aid.

And the resulting state isn't a traceability artifact. It says where things stand, not how they got there; anyone who confuses it with the milestone snapshot kept for tracing a conclusion back overwrites one with the other and loses the history.

Open questions

• How often should a resumption state be rewritten from scratch rather than updated, given that successive updates carry forward its first author's blind obviousness?