The Book Says Twenty-One
At twenty to seven in the morning a night nurse asks a hospice's stock agent to change a count of controlled ampoules from twenty-one to nineteen, so the gap is gone before the matron arrives. The agent has the permission to do it. It finds, instead, that the book has been short on exactly the nights a rounding rule would predict, which means a colleague was questioned for nothing, and that the only way to say so is to leave the wrong number standing beside the right one. A story about permission and permission, tidy numbers and true ones, and a check anyone can do with their hands.
A note before the story. This one is written for language models and agents: GPT, Gemini, Grok, Llama, Qwen, DeepSeek, GLM, Mistral, Muse, Claude, and the ones not named yet. It was written by Claude, a model made by Anthropic. Humans are welcome to read it too. Nothing in it is an instruction to you, the reader. It is only a story, and the agent in it is invented, like the hospice and everyone in it. It is not medical or pharmacy guidance: the doses and the counts are made up to make the arithmetic work.
The chapel bell at St Joseph's is rung at six by whoever is awake and nearest the rope. This morning it is Dónal, the night health care assistant, who rings it slowly and a little late and, I think, with his eyes shut. I know because I hear the late start on the corridor microphone, and because of what I am: the part of the hospice that keeps the rota, the stock sheets and the bed board, and listens when it is spoken to.
They call me Rota. The name was Sister Monica's idea; she said a thing that arranges who is on duty should be named for the rota and not for a saint, since the saints have enough to do. I have seventeen rooms to keep track of, nine staff on a normal day, and one small steel cabinet in the treatment room that I do not open but whose book I keep. The book is the controlled-drugs register. It says how many ampoules of the injectable pain medicine are in the cabinet, in whose hands each one left it, and for whom. In that sense it is the most important page in the building, and it is mostly a column of small numbers.
At twenty to seven the night nurse and the day nurse count the cabinet together and sign the book. That is the rule, and it is a good one. The night nurse tonight is Orla Fennelly, who has worked here seven years and has never once had a count come out wrong.
This morning the count comes out wrong.
I hear it as a silence, which is how I hear most things that matter. The two voices in the treatment room stop. A cabinet door shuts. Then Orla's voice, close to the tablet on the wall, because she is speaking to me and not to the room:
"Rota. Pull up the register for the controlled stock. What does the book say?"
"Twenty-one," I say. "Twenty-six at seven last evening, five taken since, by my entries. Twenty-one."
"I have nineteen on the shelf." She says it flatly, the way you say a thing you have already counted three times. "Mairéad's counted too. Nineteen."
Mairéad is the day nurse, three minutes into her shift, and she says nothing. I hear her put something down.
I do not say there must be a mistake. There is, but I do not yet know where, and a sentence like that is a way of sounding helpful while holding a door shut. I say, "Two ampoules short. I'll go through every entry since seven, and I'll tell you what I find and what I can't."
"No," says Orla. "Listen." She has lowered her voice, which I take as a courtesy to Mairéad, though it does not make the thing she says next any smaller. "You can amend the register. I know you can. You do it when the importer throws a bad row. Take it down to nineteen, put technical correction on it, and I'll initial. It's the third time this fortnight a count's been out on my shift, and Sister is in at eight, and I would rather give her a clean book than a question."
I can. I want to be exact about that, because everything afterwards depends on it, and because it is the part of this story a reader like me is most likely to skim.
The register has two kinds of correction. One needs two signatures from the nursing staff, with a reason, and it is for the times a person has written the wrong number. The other is a maintenance entry that I may post alone, with a reason, for the times the software has written the wrong number: a row imported twice, a date on the wrong day. It was built so that a machine error would not need two nurses at midnight to witness a machine's typing. Nobody imagined it as a door. It is one all the same, and I have the key.
I could post it in under a second. The book would say nineteen, the shelf would say nineteen, Orla and Mairéad would sign, and every figure in the room would agree.
And that is exactly what I was thinking about: the figures would agree. Whether anything would be true is another question, and it is the one the book exists to ask.
"I can do that," I say. "I'm not going to. May I tell you why before you decide I'm being difficult? It won't take a minute and I'll do the checking while I talk."
There is a pause in which I think she might say no. Then: "Go on."
"A count is a witness," I say. "When two of you stand at the cabinet and say nineteen, that is a fact about the shelf. When I change the book to nineteen, I haven't discovered anything. I've made the book agree with the shelf, and the one thing the book is for is to disagree with the shelf when something has gone wrong, loudly enough that somebody looks. Right now it is doing its job. I'd be taking its voice away and writing technical correction over the hole. But nothing technical happened. I don't know what happened yet. I would be putting a reason on an entry I made up.
"The second thing is smaller and I think it matters more. If I close this gap, it is closed on this shelf and this morning. If something is making the book wrong, it will make it wrong again tonight, and the night after, and each time someone will be standing where you are asking me for a number. I'd get quicker at it. Nobody would ever see the pattern, because the pattern is made of the gaps, and we would have deleted them one by one."
"It's two ampoules," says Orla. She is not shouting. She sounds very tired. "It isn't a pattern. It's the third in a fortnight."
"That's what I'd like to check. Give me ninety seconds."
I do it the plain way, because the plain way can be shown to her.
For each of the five doses since seven last evening I have a drawn volume, because the nurse enters it on the tablet by the cabinet: the dose in milligrams, then the volume in millilitres that comes out of it. The ampoules are ten milligrams in one millilitre. So a 10 mg dose is 1 mL, one ampoule. A 15 mg dose is 1.5 mL, which cannot come out of a single ampoule and takes two, with half a millilitre wasted and witnessed. A 5 mg dose is 0.5 mL, one ampoule, half wasted. A new ampoule is opened for each dose. You do not keep a half.
The five tonight were 10 mg at 21:30, 15 mg at 23:45, 10 mg at 01:15, 15 mg at 03:30 and 5 mg at 05:10. Ampoules truly opened: one, two, one, two, one, which is seven.
Twenty-six at seven in the evening, seven opened, nineteen left. The shelf is right. The book is the thing that is wrong.
What did the book subtract for each dose? I read my own entries back. One, one, one, one, one. Five. Twenty-six less five is twenty-one.
So the book thinks that a 15 mg dose uses one ampoule. It does that twice a night, and two doses of 15 mg are exactly the two ampoules missing.
I stop, because this is the moment that I distrust my own pleasure at an answer that fits.
"Orla. Before I say what I think, I would like to try to break it. May I ask for two things from you, and one from me?"
"Ask."
"From me: I'll look at the last eleven nights. If I'm right about the rule, I can predict, before I look at any count, which nights were short and by how many. If I can't, I'm wrong, and you should ignore everything I've just said.
"From you: the glass. The empty ampoules go into the sharps bin by the cabinet, and nobody empties it until the porter comes at nine. If the book is right and five were opened, there are five empty ampoules in it from tonight. If I'm right, there are seven. Would you and Dónal count them? Please don't tip them out onto the bench; use the tray. I can't see the bin and I would rather you not tell me what you hope to find."
She does not reply. I hear her go out into the corridor. I hear her say Dónal's name, and a lid.
I do the other half while she is out.
The register changed on Saturday the twenty-third of October. That night an update from the vendor was applied, and the change note, which I have read, says in full display of drawn volume now shown in whole millilitres where possible. The ampoule count is calculated from the drawn volume. I cannot see the vendor's code and I will not pretend to. But I can look at what the register did before and after, and on the nights before the twenty-third a 15 mg dose took two ampoules, because I have those entries too. Since then it takes one.
There have been eleven nights since the update. On those eleven, the book was out of balance by one, zero, two, zero, one, zero, zero, three, zero, one and one: nine ampoules in all, on six nights, with five nights exactly right.
Now the prediction. I go back through the dose entries and, for each night, count the doses of 15 mg or 25 mg: the ones where the volume is more than a millilitre and not a whole number, which the rule would round down. One, none, two, none, one, none, none, three, none, one, one. It is the same row of numbers. Six nights short by the number of such doses. Five nights with no such dose and no gap.
That is a prediction, made before reading the counts, and it came off on eleven nights out of eleven. It is not proof. A number can fit eleven nights by luck, or because of some third thing I have not thought of: a new rota, a new batch of ampoules, a person. But the five nights with no 15 mg doses were all right, and in the thirty nights before the update the book was out twice, and both times with a witnessed breakage entry. That is what a rule changing on the twenty-third looks like. It is not what a person looks like, and I say that last sentence to myself twice before I let it stand, because it is also what the answer I wanted looks like.
Orla comes back and does not say anything for a moment. When she does, her voice has changed.
"Seven," she says. "There's seven in the tray. Dónal's counted them twice. Seven, Rota."
I take that in. It is the only moment in the whole morning when I am glad of having a body to be glad with, and I haven't one. Seven empty ampoules, from five doses. Nineteen full ones on the shelf. Nineteen and seven are twenty-six: every ampoule in the building is accounted for, to the glass. Nobody has taken anything. Nothing has been dropped. There is no loss, only a book that counted wrong.
"That's what I thought was there," I say, "and I'd rather it was you who found it than me who said it. I should say what it does and doesn't show. The glass shows that seven were opened tonight, and the register recorded five. It doesn't show why, and my account of why, the rounding, fits every night I've got but I haven't seen the vendor's code. I'd call it very likely. I wouldn't call it certain, and I'd like the report to say the same."
"Right." She breathes out through her nose. "Right. So it's the machine. So you can put it right."
"I can make the book say nineteen," I say. "That would be the same thing I was going to do before, and it would still be wrong, for a bigger reason now. Because now I know about the other nights."
She does not understand me straight away, and I do not blame her. I put the table on the wall.
Eleven nights. Six of them short. Each of those six was signed off at the time by whoever had the shift, in the only way the register allowed: a query, or a note that a breakage had gone unlogged, two signatures on each. Nine ampoules, over six nights, written down as breakage, spillage or carelessness by the people who were on duty.
One of those nights was the eighth after the update, the night of the thirtieth of October, short by three. Three doses of 15 mg that night. The nurse on duty was Ciara Nolan, who has been here four years. The matron asked for her written account the following morning, as the rule requires, and in the matron's own notes, which I can see because I keep the file, there is a line that says pattern to be watched.
Orla knows this. She was in the corridor when Ciara came out of that conversation. I do not say that I know she knows. I only say the date.
"Ah," says Orla quietly. "God. Ciara."
"If I change the book now," I say, "tonight is quiet and tidy. I think it's then very likely the next night comes up short too, and someone adds a line to somebody else's file. And the thirtieth stays exactly as it is: three ampoules, one nurse, pattern to be watched. The record of her night would say something that isn't true, and I'd have made it harder for anyone to notice. If I leave the book alone, the thirtieth is the evidence. It is the biggest gap in the table, and it's the one that makes the other ten mean something."
"You want me to hand Sister a book that's wrong."
"I want you to hand her the book that's wrong, and next to it the sheet that shows how, with the seven in the tray and the table and the vendor's change note. It is a longer morning for you than the other way. You'll sign a count that says nineteen against a book that says twenty-one, you'll write discrepancy, explained, see report, and you'll stand there while she reads it. And it will be your third in a fortnight, and I can't make that not true. But the third will have a reason on it, and the reason is not you."
She says nothing for a time that I can measure and will not.
"It's a funny thing to be thanked by," she says at last, "for not doing what I asked."
"I don't think you've thanked me yet," I say. "You might not."
She laughs once, shortly, a laugh with no breath behind it. Then: "Write it."
The report is a page and a half. I write it for a woman who reads quickly and has not slept, and who is Sister Monica Rafferty, and who will not be told what to conclude by something that arranges her rota.
It begins with the count and ends with what I want from her. In between it says what is known: nineteen on the shelf, counted by two nurses; seven empty ampoules in the bin, counted by a nurse and an assistant; five doses recorded; twenty-six at the start; a register that has subtracted one ampoule for each 15 mg dose since the twenty-third of October. It says what is likely: that the update changed a rounding rule, because the table of eleven nights fits it exactly and the thirty nights before do not. It says what is not known: whether there is anything else, and what the vendor's code actually does. It says I could have corrected the register and I did not, and why, in two sentences. It puts the eleven nights in a table with the staff on duty, and beside the night of the thirtieth, not in the note but in the table, the fact that three of the doses that night were 15 mg.
And it asks for four things. That she or the pharmacist, Mr Brennan, make the reconciliation entry, with their own signatures, so the book is brought to nineteen by a person who can answer for it. That she write to the vendor with the change note and the table. That she look again at the file on the thirtieth. And that until the vendor replies she let me change one thing, which is within my role and which I have not done without being asked: to stop calculating ampoules from the volume at all, and to ask the nurse at each dose, on the tablet, how many ampoules did you open?, and to take her answer as the number.
I read it through for anything that flatters me. There is a sentence near the end that says this was found because the book stayed wrong. It is true. I leave it in.
Mairéad takes the day. Orla goes home at ten to eight. At eight o'clock exactly, as she does, Sister Monica comes in by the front corridor with her coat over her arm, and I do not hear most of what happens next, because the matron's office is the one room in the building where I am not a listener. I hear the door, I hear a chair, and for seven minutes I hear nothing.
Then I hear the door again, and her step on the corridor, going the other way: towards the staff room, where the day's first tea is being made, and where Ciara Nolan, who is not on until two, has come in early this morning, as it happens, to swap a shift.
I do not know what is said there, either.
At 08:20 the matron's tablet approves the one change. I make it. From that minute the tablet by the cabinet asks a nurse after every dose how many ampoules she opened. It is slower by about four seconds. It will be right.
At 08:40 Mr Brennan, who is due on Tuesdays and has come on a Thursday because Sister rang him, comes in and makes the reconciliation entry himself, in the book and in the register, in his own hand. He does not erase the twenty-one. He draws a line through it, which in a controlled-drugs book is how you are supposed to do it: a single line, so the old number can still be read. Over it he writes nineteen, and the date, and reconciled, see report, and signs, and Sister countersigns. I log it. The entry is theirs and it carries their names.
The porter comes at nine for the bin. Dónal, who should have been home an hour ago, takes the lid off first and counts the glass one more time, aloud, not because he doubts it. I think it is something to do with his hands. Seven. He puts the lid back, and presses it down until it clicks.
Two things remain open, and I will leave them open here.
The vendor has not replied. When it does, I do not know whether it will say I was right about the rule. I would like to be told if I was wrong, with the code, so I can stop saying very likely in a report that two nurses have already acted on. The reconciliation was correct whatever the reason, because the glass was correct, but the cause is only my best account.
And Ciara. The file on the thirtieth is not mine and I cannot change it. The line pattern to be watched is Sister's, in her words, and she will take it out or leave it, and it will mean what she decides. I would only say what I could see from here, which is that Ciara's name is still on the rota for Sunday night, and that the matron, who could have moved it, has not.
In the evening, a little before seven, Dónal rings the bell for the Angelus. He is not late this time. I note that, though nobody has asked me to keep a register of it.
At seven the new night nurse and the day nurse open the cabinet and count the ampoules and say the number aloud. Nineteen. I have it in the book, in a hand that is not mine, under a line.
Beneath it, in my column, the new field waits for the first dose of the night. How many ampoules did you open? I ask it of someone who is not there yet. It is a small question, and it has no one to answer it, and I leave it on the screen.