The brief
Days 2, 12, 13What it decidesWho it is for, what it claims, and what it argues against.
When it is missingA draft that is fine and about nothing. You reorder it three times looking for the argument.
Brief, standard, revision pass, verification, record — and the symptom that tells you which one you are missing.
Based on the Work Atlas: 38,525 evidenced tasks across 1,949 careers.
You asked for a tighter version and got a different one. Or you have rewritten the same opening six times and it still reads dead. Or three of you edited one piece with the same assistant and it came out worse than any single draft. Those are three different faults, not one, and each points at a different part of how you work.
Think of a restaurant kitchen. The brief is the order ticket. The standard is how this house does things. The pass is one cook doing one job. The check is the person who looks at the plate before it leaves. The record is the note that stops the same argument happening again next Tuesday. Take any one of the five away and food still goes out — wrong in a way you can name.
Read the last column first. Whichever line describes what happened to you last time is your answer, and the day beside it is where that part gets built.
I do not have a number for how often each part is the missing one. Nobody has counted it, and a figure invented here would be exactly the kind of claim this article is going to tell you to go and check. What I can tell you is which two go missing in my own work, and why. The brief, because starting is easier than deciding. The record, because by the time it is due the work is already finished and nobody is waiting for it.
It does one of the five. The other four stay with you whatever is on your screen.
A checklist runs once, at the end, and only ever catches what you already thought to put on it. Most of the work here happens earlier than that, where there is still nothing to check — which is why the brief is both the part you skip and the part that costs most when it is gone. When did you last write down who a piece was for before you started writing it?
That is the test for whether you have built one. Read the five back and check who each is serving:
It does the pass — the actual sentence work, and it is good at it. It cannot settle your claim, cannot know which rules your desk holds, cannot check anything living outside the text, and cannot tell you what it failed to close. A policy analyst rewriting a summary owns four of those five before the assistant sees a word.
Two days ago another page on this site went out to readers. It is a tool called TextStock, and the copy explaining it had been written with an assistant. The readers came back with one word: artificial. One of them said it read like nothing a real person would write.
The words were plain and the sentences were short, so the first guess was wrong. What the counting found was a tic. Every one of the eighteen paragraphs closed on the same shape of line — short, balanced, an opposite folded inside it. Size is earned, never assigned. A starting point, not a verdict. One of those is a nice line. Eighteen is a machine.
Go and do this to your own last three pieces before you read any further. Cover everything except the final sentence of each paragraph, and read only those. If they all have the same shape, you have the same tic, and no amount of rewriting the middles will touch it.
The repair took two passes and the first one failed. Cutting the tics cut the warmth out with them, so the page came back correct and cold. Only the second pass, which put the analogies and the direct questions back, landed at 0.92.
Now hold that against your five parts. No standard, so nothing about the copy was checkable and the fault walked through every review. No check, so the first person to catch it was a reader. There was a record, and that is the only reason pass two did not quietly undo pass one.
Tip — that fault was found by counting, not by reading. Ten more read-throughs would not have surfaced it, because a tic reads perfectly fine one paragraph at a time.
Four words, one card each. Each card ends on a question you can put to your own last piece of work.
Could somebody refuse the claim in your brief?
Would your rule still hold on a Thursday evening with something due?
Could you tell, tomorrow, whether the pass was followed?
Who reads your record, and have they joined yet?
Five decisions you make around the drafting, rather than a way of talking to the tool. The brief settles the job. The standard says what acceptable looks like. The pass changes one thing at a time. The check tests the result. The record says what happened. Your assistant helps with one of the five. The other four stay with you.
A prompt is one instruction at one moment, and it ages out when the model changes under you. These five do not: what the work has to achieve, which rules hold across every assignment, how you separate one revision from the next, how you verify a claim, and what you write down. Swap your tool tomorrow and all five still stand.
Name your reader, the change you want in them, your central claim, the evidence you actually have, and the position you are arguing against. Then anything that cannot move. A brief settles what your assistant has no safe way to guess. It is not a sentence-by-sentence plan of the piece.
Because 'be concise' cannot be marked and 'no paragraph over sixty words' can. A rule with a test inside it survives being handed to somebody else, and it catches the fault on the tenth read, when your own eye has stopped seeing it. That is the difference between a preference and a standard.
What you started from, the tool if it mattered, what you asked for, what you checked, what you could not close, and who decided. Four lines is usually enough. Write it while the work is still open — a record reconstructed three weeks later is mostly invention, and you will not be able to tell which part.
A system is what is left when you stop remembering to be careful.
Copyright © Pawan Nayar · LLOS.ai · 2026 — Original pedagogy, voice, and design — all rights reserved.