Nice. Wrong.src
Transcript of the conversation that produced the short story “Nice. Wrong.”, 11 November 2026. J is the writer. C is Claude (the session began on Claude Fable 5.2 and later ran on Claude Sonnet 5.5). Every message from both sides is given in full. What has been removed is the machinery around the conversation: tool calls, file transfers and automated system messages. Where one of those did something that mattered, it appears as an italic bracketed note. In C’s replies the markdown headings have been turned into bold lines; the words are unchanged. The story is published separately and is not reproduced here. The conversation continued after this point, with the request for these workings and a revision of them, and that part is not included.
J: I have been given some miniature whiskies - Ardnamurchan - along with the message:
“Write one short story inspired by each whisky. Each ones must feature either a pirate, a cook, a naturalist, or a computer game designer.”
I have drunk the first one - their standard single malt (other bottles are expressions in different kinds of wine-flavoured cask) - and it is a delicious biscuity malt, but I am fresh out of ideas of how to link that to a short story.
Discuss ideas with me, and when I’m ready I’ll ask you to write one.
C: The biscuit note is doing more work than you think. Here are a few ways in, roughly from most to least obvious:
- Ship’s biscuit (pirate or cook). A biscuity malt on the most westerly distillery on the British mainland almost begs for hardtack. A ship’s cook who is sick of weevilled biscuit and trying to make one good thing from a sack of malted barley the crew took by mistake, thinking it was something valuable. The story is small and domestic, set against people who want it to be a bigger story.
- The Viking in the boat (pirate). A Viking boat burial was found on Ardnamurchan in 2011: a raider laid out in his ship with his sword, axe and drinking horn. A pirate story told from the wrong end, the grave goods rather than the raid, with someone deciding what he gets to take with him.
- The provenance trail (game designer). As far as I know, Ardnamurchan bottles carry a code that traces the whisky back through cask, still and barley field. A designer obsessed with giving every object in a game a true history, down to the biscuit in an NPC’s pocket, and what happens when a player follows one all the way back.
- The crater (naturalist). The peninsula is the remains of an ancient volcano, with ring-shaped ridges you can only really see from above, and it’s thick with otters, pine martens and sea eagles. A naturalist cataloguing a place that is quiet now but was once violent, which is not a bad description of a mellow malt either.
- The remembered biscuit (cook). A cook trying to reproduce a biscuit from childhood, getting closer with each batch, and finding that the thing they’re missing wasn’t an ingredient.
One practical thought: the wine-cask bottles will hand you pirates on a plate (plundered cargoes of sherry, Madeira and claret), so you may want to save the pirate and give this plain, honest one to the cook or the naturalist.
Do any of these pull at you, and are you after something funny, eerie or quiet?
J: I want to avoid mentioning whisky - that’s too literal. I think the provenance trail angle is a great one. But let’s save it for another bottle - we can hint at it. Tell me more about the cook and biscuit idea.
C: There were two biscuit ideas in that list, and I think they’re stronger merged than apart, so here is the combined version.
The core. A cook is trying to reproduce a plain malted biscuit from childhood. It isn’t a fancy biscuit: flour, butter, a little sugar, malt. Each batch gets closer and each one is wrong in a way the cook can’t name. The story is the batches.
Where the malt comes in. Malt can be in the story without anyone drinking anything: malt extract from a tin, or malted barley the cook roasts and grinds by hand. The tasting notes become the cook’s vocabulary for failure: too sweet, not toasty enough, no warmth at the back. You get to describe what you drank, disguised as someone criticising a biscuit.
The provenance hint. This is where the other bottle gets its seed planted. The cook’s method is to chase origins: the same mill, the same dairy, eventually the same field the barley came from. That gets very close, but not all the way. So this story says tracing a thing to its source gets you nearly everything, and leaves the later story to take that idea seriously.
What’s actually missing. This is the decision that sets the tone:
- The staleness. The original biscuits lived for weeks in a tin in a damp house by the sea. The cook has only ever eaten them fresh, so they’re too good. This one is quiet and a bit funny.
- The smoke. They were baked in a range fired with peat or wood, and the cook is using a clean modern oven. Sensory, and true to the dram.
- The hunger. They were eaten after cold days outdoors by a child who was starving. Nothing is wrong with the biscuit. This is the saddest of the three.
- The mistake. The original baker always slightly burnt them, or mismeasured, and the cook’s skill is the obstacle.
Setting. A kitchen at the far end of a long single-track road suits it, as does a ship’s galley, where the constraint of ship’s stores makes the hunt harder and funnier.
My instinct is the staleness ending in the end-of-the-road kitchen, because the cook finally succeeds by doing nothing at all, just waiting. Which of the four endings feels right to you?
J: The story needs to become one about context and nostalgia. The cook isn’t really remembering a perfect biscuit, but the perfection of the environment in which they ate that biscuit. But we want to steer well away from mawkishness or cliche.
C: The risk is that “it was never about the biscuit” is itself the cliché: Proust’s madeleine, the critic’s flashback in Ratatouille. So the story has to know that and go one step past it. Here are some ways to keep it dry.
Make the remembered scene unlovely. No grandmother, no sunlit kitchen, nobody dead. The perfect environment should be perfect for a reason the child couldn’t have named and an adult would find odd: sitting on a pile of creels at a pier in the rain, ignored for three hours while a parent did some job, cold, slightly bored, responsible for nothing. The biscuit came from an adult who didn’t much like children and handed it over to keep them quiet. What the cook misses is being off the hook.
Make the biscuit ordinary, possibly bad. Out of a packet, from a tin, a bit soft. The cook has spent months making artisan versions of something a factory turned out by the million. That’s the comedy, and it undercuts any reverence.
Keep the cook unsentimental and the prose technical. A batch log would do this well: numbered attempts, hydration, bake times, terse tasting notes. The feeling only leaks in at the margins, in what the cook bothers to write down that has nothing to do with baking.
No epiphany. The cook gets it right by accident: a failed batch, eaten on the back step in the cold after a bad service, hands numb, nobody wanting anything for five minutes. The reader sees why it worked. The cook then draws the wrong conclusion and carefully records that batch’s recipe as the one.
Or one step further. The cook does work it out, tries to stage the conditions on purpose, and it fails flat. It only works when you don’t mean it. That gives the story an idea beyond nostalgia: the context can’t be made to order, which is an awkward thing for someone whose job is making things to order.
My preference is the batch log with the wrong conclusion, because the story never has to say its theme aloud. The cost is that the cook ends none the wiser, which some readers find cold. Do you want the cook to understand what happened, or only the reader?
J: I love the device of having batch logs. We can structure this story - it opens with a numbered and dated batch log - as a table - conceptually an index card perhaps. Then prose about that batch. First person, the cook’s thoughts, introduce the goal. Subsequent sections of the story are the story of batches. We can illustrate the passage of time by skipping numbers and with the dates.
In the prose we can tell of the expeditions to find ever-more specific ingredients. These don’t need to be exotic - it’s not a journey to Madagascar to source vanilla, it’s a journey to a specific corner shop in a London suburb where they stock some forgotten brand of vanilla extract.
With a loose goal of perhaps 10 paperback pages, sketch out the structure.
C: Ten paperback pages is roughly 2,700 words, which gives seven sections. Two devices carry the structure:
- The card changes shape. It starts with five rows and gains rows as the obsession grows, so the reader can see the cook’s state of mind from the card before reading a word of prose.
- The memory is rationed. There is never a flashback. Each section lets slip one detail of the original scene, which the cook treats as irrelevant to the problem. By the end the reader holds all the pieces.
The sections
Batch 1, early March (about 400 words). The card is bare: flour, fat, sugar, malt, bake, result. The prose sets up the cook as a competent, slightly impatient professional and states the goal as a technical problem. The memory gets one flat sentence. Result: “Nice. Wrong.”
Batch 4, late March (about 350 words). This is the malt section, where the tasting vocabulary goes: toasty, biscuity, warm at the back. The first expedition is a modest one, to a home-brew shop for malt extract. The biscuit gets closer. Memory detail: it was raining.
Batch 9, May (about 450 words). The humbling. The cook realises the original was a factory biscuit, made with margarine, and that a professional must now deliberately bake worse. The expedition is to the suburban corner shop for a brand everyone assumed was discontinued. The card gains a “Source” row. Memory detail: the adult who handed the biscuit over.
Batch 17, July (about 500 words). Peak obsession and the funniest section. The card is now crowded. This is where the hint for the later bottle goes: a code printed on an old packet, and a phone call to someone who can read from it the factory, the line and the shift. The factory has closed. The cook believes one more link in the chain will do it. Memory detail: having nothing to do for three hours.
Batch 23, September (about 400 words). The biscuit is technically identical. Someone else who remembers them confirms it. The cook knows it isn’t right. The result row is left blank.
Batch 24, late November (about 450 words). The number follows straight on but the date shows a ten-week gap: nothing has been baked. This batch is made carelessly, for staff tea, to use up the ingredients. After a bad service, the cook eats one on the back step in the cold, with nobody wanting anything. It is right. This is told as flatly as possible.
Batch 25, the next day (about 150 words). Mostly card. The cook repeats batch 24 exactly, indoors. Result: “Wrong.”
Decisions for you
- The ending. Ending on that final “Wrong” leaves it open whether the cook has understood. The reader certainly has. If you want the cook to understand, the last card could gain a new row, such as “Eaten: where”.
- Brands. I’d invent the margarine, the extract and the biscuit, with names plausible enough that a reader half believes they remember them.
- The cook. Name, sex and type of kitchen are all open. I’ve assumed a working restaurant, since the bad service and the back step need one.
J: I think the memory is of a childhood biscuit, a carefree time, after school, no homework to worry about, taking biscuits from the biscuit barrel and eating them while waiting for dad to get home from work. But we don’t say that - reader can infer it partially.
Yes to inventing the brands.
Not sure about “someone else who remembers them” because there’s nobody who knows.
C: That memory works, and the biscuit barrel solves a plot problem: a child who takes biscuits from a barrel never sees the packet. So the cook doesn’t know what the biscuit was, and has to work it out from which shops the household used. That justifies the whole hunt.
Keeping it inferred
The rationed details change to suit. The cook mentions each only as a nuisance variable:
- The barrel. The cook wonders whether its lid let them soften.
- The time of day. Always about the same hour, with the same light.
- The empty time. Nothing that had to be done.
- The listening. One ear on the drive or the door. Dad is never mentioned.
The back step in batch 24 can then echo it without comment: service over, nothing owed to anyone, and the cook waiting for a lift, half listening for a car.
Fixing batch 23
Nobody else knowing is better than my version. The cook is the only instrument, and can’t calibrate it. Two ways to play September:
- Nothing left to change. Every row on the card is sourced and ticked. There is no variable left and nobody to ask. The result row stays blank. Quiet, and a little bleak.
- The real thing turns up. The phone call in batch 17 reveals that the line was moved, not closed, and the biscuit is still made under a supermarket’s own label. The cook buys a packet and fills in a card for it anyway, with most rows reading “n/a”. It tastes wrong too.
I’d take the second. It proves the biscuit was never the problem without anyone saying so, it makes the provenance trail matter to the plot, and a batch card for a bought packet is a good joke. Its cost is that it comes close to spelling out the theme, so the cook’s reaction has to be purely technical: they must have changed the recipe.
Which September do you want?
J: Nothing left to change.
Is it bleak to say, you can’t get that moment back? Or a good lesson.
We could realise that an early batch (fresh, real butter etc.) was absolutely delicious, and have the cook return to that - good things now are better than false nostalgia.
C: It’s only bleak if the story stops at the loss. “You can’t get that moment back” plus “and here is a good one you nearly missed” is a consoling shape, and your early-batch idea supplies the second half. I think it’s better than my accidental-success ending, which half implied the moment could be recovered by luck.
It also pays off something already in the structure. Batch 1’s result was “Nice. Wrong.” The cook wrote off a delicious biscuit in two words and spent eight months making worse ones more accurately. The ending can simply be that card again, with one word gone.
Two cautions:
- Don’t call the nostalgia false. The memory was real and the afternoon was as good as remembered. It just isn’t a recipe. If the story sneers at the memory it turns sour, and if the cook announces the lesson it becomes a moral. The cook’s stated reason for going back to batch 1 should be practical: staff tea, and there’s real butter in the fridge.
- Let the two endings merge. Keep the back step. The cook eats a batch 1 biscuit there after service, in the cold, waiting for a lift. It isn’t the old biscuit and the cook doesn’t claim it is. It’s a good biscuit at a good moment, which is all the original ever was.
Revised ending
Batch 23, September. Every row is sourced and ticked, and nothing is left to change. The result row is blank.
Batch 24, late November. The card is back to five rows and the recipe is batch 1’s. The prose is the back step. Result: “Nice.”
That makes six sections, and I’d drop batch 25. One optional last touch: the rest go in a tub by the pass and are gone by Thursday, so somebody else is now taking biscuits without looking at them. It can be cut if it feels too neat.
J: Please write this as a Markdown artifact.
Remember you’re in first-person, so this is a modern-day British person writing to themselves, almost a diary. But, full sentences, cleanly written for a contemporary fiction audience.
[C writes the story to a Markdown file and sends it to J.]
C: The story is in nice-wrong.md, titled “Nice. Wrong.”, at about 2,400 words across six batches from 2 March to 28 November.
Things I chose that you may want to change:
- Invented names. The biscuit is an Aldwick Malt by Garside & Lowe, the margarine is Kingcup, and the flavouring is Ransome’s. The corner shop is Parade News in Harrow Weald.
- The provenance hint. It sits in batch 17: a code on an old wrapper, read over the phone by a woman on a careline, giving factory, line, day and shift. The cook gets one line of wonder about it and moves on.
- The cook is unnamed and unsexed. Only the sous chef’s question and Dev with his car put other people in the frame.
- Dad is never mentioned. The nearest the story comes is “one ear on the gate” in batch 23, echoed by “one ear on the road” at the end.
- The tub by the pass stays in, as a single sentence. It’s the easiest thing to cut if it reads as too neat.
The length is a little under the ten-page target. If you want more, batch 4 is the thinnest section and could take another expedition.
J (comment on “because”, in the story’s opening paragraph: “…keep a card for every attempt, because I know what I’m like…”): Drop “because”, for a dash.
[C makes the edit and sends the file again.]
C: I swapped “because” for a dash in the opening paragraph, so it now reads “a card for every attempt — I know what I’m like and I will otherwise make the same mistake four times.” I’ve re-sent the updated file.
J (comment on “a word stamped across the middle which I never read, because by the time I looked at a biscuit I had usually bitten it.”): We can’t have the word - they would have certainly remembered it
[C removes the stamped word, along with the two later references to it, and sends the file again.]
C: You’re right, the cook would never have forgotten a word stamped on every biscuit. I’ve removed it and re-sent the file.
- Opening description: the biscuit now has “nothing written on it at all”, as far as the cook can remember.
- Batch 17: I cut the line saying the word they never read was ALDWICK. The name now comes only from the wrapper: “It was an Aldwick Malt, made by Garside & Lowe.”
- Batch 23: the engraved brass stamp is now just the border and the three rows of five holes, with no lettering.
The cook still recognises the wrapper by its border and finds it by browsing photographs, which fits a biscuit with no word on it.
Who did what. J rejected a literal treatment of the whisky and saved the provenance idea for another bottle; set the theme (context and nostalgia, not the biscuit); designed the batch-card structure and the corner-shop expeditions; supplied the actual memory (the barrel, after school, waiting for Dad); rejected C’s “someone who remembers them” and its supermarket-label September; and devised the ending, in which the cook returns to the fresh, buttery first batch. J also made both edits to the draft. C proposed the first set of ideas and the merged cook-and-biscuit premise, warned about the madeleine, suggested the rationed-memory device and the back step, and wrote the prose, including the invented brands, the wrapper-code scene and the title.
What C got wrong. The first draft had a word stamped on the biscuit that the cook “never read”, which someone who ate fifteen hundred of them would not have missed; J caught it. C’s “someone else remembers them” September was rejected for a similar reason. C’s two remarks about Ardnamurchan in its first reply (the 2011 Viking boat burial and the volcanic geology) came from general knowledge and were not checked in the session. Neither appears in the story.