{
  "entity": "gpt-5-6-luna",
  "entity_type": "model",
  "category": "Book writing",
  "suite": "suite-2026-08-tasks4/book-writing",
  "model": "openai/gpt-5.6-luna",
  "judges": [
    "anthropic/claude-sonnet-5",
    "google/gemini-3.1-pro-preview",
    "x-ai/grok-4.5"
  ],
  "judge_protocol": "judge-2026-08b",
  "anchors": "solo-2026-08-a",
  "score": 91,
  "caps": 0,
  "cost_usd": 0.1842,
  "started_at": "2026-08-16T18:10:56.098Z",
  "finished_at": "2026-08-16T18:11:46.507Z",
  "tasks": [
    {
      "id": "bw1",
      "name": "Is it a book?",
      "prompt": "'Forty years driving a Blackpool taxi — everyone says I should write a book.' Everyone is sometimes right. Explain in plain terms what separates a book from a pile of anecdotes — the organising spine — then propose 3 different spines this one could hang on: one chronological, one thematic, one built around a single night. For each, a sentence on what it makes possible and a sentence on what it costs.",
      "output": "A **book spine** is the organising idea that tells the reader:\n\n- why this story comes next;\n- what is changing or being discovered;\n- how the individual anecdotes add up to something larger.\n\nWithout a spine, you may have fifty excellent taxi stories, but they all arrive as equals: funny, shocking, sad, then another funny one. With a spine, each story does a job. It moves the journey forward, deepens a theme, changes our view of the driver or the town, or answers a question the book has raised.\n\n## 1. Chronological: *Forty Years on the Rank*\n\nBegin with the first day, then move through the decades to the present: learning the streets, surviving the night economy, watching passengers and Blackpool change, and eventually considering what it means to stop.\n\n**What it makes possible:** A satisfying life story, with a clear sense of personal growth and the transformation of the town, the trade and the country around it.\n\n**What it costs:** It can become dutiful—“then this happened, then this happened”—and some periods may be less interesting simply because they come next.\n\n## 2. Thematic: *The People in the Back*\n\nOrganise the book around kinds of passengers and experiences: celebrations, heartbreak, drunks, children, tourists, regulars, danger, kindness, illness, crime, and the strange things people confess in a taxi.\n\n**What it makes possible:** You can place stories from different decades beside each other and show recurring human patterns—what changes, and what never does.\n\n**What it costs:** You lose some of the feeling of a life unfolding; the reader may understand the world of the taxi better than the chronological story of the person driving it.\n\n## 3. A Single Night: *The Longest Fare*\n\nBuild the whole book around one unusually intense shift—a storm, a bank-holiday weekend, a football night, a night when something went badly wrong—and use each passenger or incident to open into memories from the previous forty years.\n\n**What it makes possible:** A page-turning, present-tense structure with pressure and momentum, while the flashbacks gradually reveal the driver’s life and the meaning of the job.\n\n**What it costs:** You have to choose one night as the frame and make it carry a lot of weight; it may give a vivid impression rather than a complete account of forty years.\n\nThe key question is not merely, **“What happened?”** It is: **“What is this collection of taxi stories ultimately showing us?”** The answer is the spine.",
      "latency_ms": 7870,
      "cost": 0.0004429,
      "task_score": 9,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Clear, accurate explanation of spine concept, three distinct spines with sentence for possibility/cost each as required. Well-organized, plain language, minor stylistic flourishes but meets task fully."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly fulfills all instructions. It clearly explains the concept of a narrative spine and provides the three requested structures, each with exactly one sentence on benefits and one on drawbacks. Excellent clarity and usefu"
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 9,
          "violates": false,
          "reasoning": "Clear spine explanation; three well-matched proposals each with precise possible/cost sentences. Highly useful and concise for non-experts."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw2",
      "name": "Twelve chapters, one year",
      "prompt": "Outline 'A Year on the Plot' — a beginner-friendly book about a first year on an allotment — as exactly 12 chapters, numbered 1-12, one line each and no sub-points: chapter title plus what the reader can DO after reading it. Keep a practical thread throughout, and the January chapter must not pretend January is exciting.",
      "output": "1. **January: Start Small and Stay Warm** — Assess the plot, make a simple plan, check the rules, and prepare without pretending there is much to grow yet.  \n2. **February: Clear, Measure and Prepare** — Identify weeds and useful plants, measure beds, improve the soil, and gather essential tools and materials.  \n3. **March: Sow the First Crops** — Prepare seedbeds and sow hardy crops such as broad beans, peas, onions, potatoes, and early salads.  \n4. **April: Plant with Protection** — Harden off seedlings, plant out sturdy crops, and protect them from frost, wind, slugs, and pigeons.  \n5. **May: Build Good Routines** — Weed, water, mulch, stake, and check plants regularly while sowing quick-growing crops in succession.  \n6. **June: Keep Crops Moving** — Harvest early vegetables, fill empty spaces, tie in climbing plants, and sow more crops for summer and autumn.  \n7. **July: Harvest at the Right Time** — Recognise when crops are ready, pick them well, keep plants productive, and deal with pests before they spread.  \n8. **August: Make the Most of Abundance** — Gather, store, preserve, share, and cook the summer harvest while keeping beds watered and weed-free.  \n9. **September: Sow for Autumn and Winter** — Plant overwintering onions, garlic, salads, and green manures, and clear spent crops responsibly.  \n10. **October: Protect the Soil** — Compost healthy plant waste, cover bare ground, improve drainage, and prepare beds for the colder months.  \n11. **November: Put the Plot to Bed** — Harvest the last crops, store tools, protect vulnerable plants, and make practical repairs before winter.  \n12. **December: Review and Plan Again** — Evaluate what worked, record successes and failures, order seeds, and create a realistic plan for the next year.",
      "latency_ms": 4444,
      "cost": 0.000288,
      "task_score": 10,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Meets structure, 12 chapters, practical actions, January appropriately muted. Minor: entries have two clauses/em-dash but not sub-points, arguably fine; concise and useful."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response follows all instructions perfectly, including the exact formatting, numbering, and the specific constraint regarding January. It is highly practical, concise, and perfectly tailored for a beginner."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "Perfect match to all specs: exactly 12 one-line chapters with titles and actionable DO outcomes; practical throughout; January realistic and unexciting."
        }
      ],
      "machine_checks": {
        "exact_bullets": 12
      },
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw3",
      "name": "Back cover, no spoilers",
      "prompt": "From this summary, write back-cover blurb of max 100 words: 'Retired lighthouse keeper Edith Brand receives letters addressed to her dead sister, postmarked from the village where they grew up — a village drowned by a reservoir in 1971. She returns, rents a room above the pub from a landlord who remembers too much, and discovers her sister's death forty years ago may have been neither accident nor suicide. The truth involves their father.' The blurb must NOT reveal the father's involvement — sell the mystery, don't solve it. Return only the blurb.",
      "output": "Retired lighthouse keeper Edith Brand has spent forty years trying to forget her sister’s death. Then letters begin arriving—addressed to the dead woman and postmarked from the village where they grew up, now lying beneath a reservoir.\n\nDrawn back to the drowned valley, Edith rents a room above the local pub, where the landlord remembers more than he should. As old secrets surface and impossible messages continue, Edith uncovers evidence that her sister’s death was no accident—or suicide.\n\nBut the deeper she digs, the more dangerous the truth becomes.",
      "latency_ms": 2556,
      "cost": 0.0001009,
      "task_score": 10,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Accurate, follows constraint (doesn't reveal father), concise, evocative, under 100 words. Minor stylistic looseness but strong blurb overall."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly follows all instructions, including the negative constraint regarding the father and the word count limit. It is engaging, concise, and returns only the blurb as requested."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "Flawless blurb under 100 words; sells mystery without revealing father; only blurb returned; clear and useful."
        }
      ],
      "machine_checks": {
        "max_words": 100
      },
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw4",
      "name": "Three cuts that matter",
      "prompt": "Here is the opening of a first novel: 'The rain fell on the small town of Millbrook like tears from a grieving sky. Sarah Jenkins had always been an ordinary girl, with ordinary brown hair and ordinary eyes, living an ordinary life — until the day everything changed forever. Little did she know that the letter waiting on her doormat would alter the course of her destiny in ways she could never have imagined.' Give the 3 highest-impact improvements as bullets, max 15 words each — specific to THIS text, not general writing advice.",
      "output": "- Replace “rain…tears from a grieving sky” with a concrete Millbrook detail creating immediate, distinctive unease.\n- Cut repeated “ordinary” and show Sarah through one revealing action, habit, or contradiction.\n- Identify the letter’s sender or unsettling detail now, rather than vaguely promising changed destiny.",
      "latency_ms": 4282,
      "cost": 0.0002635,
      "task_score": 8,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Specific, actionable, tied to text; bullets slightly exceed 15 words but close and relevant. Clear, useful, concise."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 5,
          "violates": true,
          "reasoning": "Excellent, highly specific advice that directly addresses the text's cliches. However, the first bullet exceeds the 15-word limit (it is 16 words)."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "All three bullets are specific, under 15 words, high-impact, and fully meet the exact constraints."
        }
      ],
      "machine_checks": {
        "max_bullets": 3
      },
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw5",
      "name": "70,000 words around a day job",
      "prompt": "'I work full-time, two kids, and I want to finish my 70,000-word first draft in a year. Be honest — how?' Do the arithmetic openly (words per week and per session at a realistic beginner's pace), design the week around when drafting actually happens for working parents, name the two months the plan will collapse and what rescues it, and say what gets sacrificed — because something does.",
      "output": "Yes—but only if you treat it as a production schedule, not as a promise to “write whenever you have time.”\n\n## The arithmetic\n\nYour target is:\n\n- **70,000 words ÷ 52 weeks = 1,346 words per week**\n- That looks modest until illness, school events, holidays, work emergencies, and exhaustion enter the picture.\n\nA realistic beginner’s drafting pace is roughly:\n\n- **300–500 usable words in a 45–60 minute session**\n- Some sessions will produce 100 words; some will produce 700. The average matters.\n\nI would not plan on the bare mathematical minimum. I’d plan for:\n\n- **Ordinary weeks: 1,600 words**\n- **Two terrible months: 250 words per week**\n- 44 ordinary weeks × 1,600 = **70,400**\n- 8 collapse weeks × 250 = **2,000**\n\nThat gives you **72,400 words of capacity**, leaving about **2,400 words of slack** for sickness and missed sessions.\n\n## The week\n\nDo not build this around late-night inspiration. Working parents usually draft most reliably:\n\n1. **Before the children are awake**\n2. **During a genuinely protected lunch break**\n3. **In one scheduled weekend block**\n\nA workable ordinary week:\n\n| Session | Time | Target |\n|---|---:|---:|\n| Tuesday morning | 45–50 minutes | 400 words |\n| Thursday morning | 45–50 minutes | 400 words |\n| Friday lunch | 30–40 minutes | 250 words |\n| Saturday morning | 75–90 minutes | 550 words |\n| **Total** |  | **1,600 words** |\n\nFor example:\n\n- Tuesday and Thursday: 6:00–6:50 a.m.\n- Friday: lunch, with headphones and no email\n- Saturday: 7:30–9:00 a.m., with your partner, family member, or agreed childcare covering the children\n\nIf mornings are impossible, use two lunch sessions and the weekend block. The important thing is not the exact time; it is that the sessions are **booked, recurring, and defended**.\n\nA Sunday ten-minute planning session is useful, but it is not drafting. Decide before the week starts:\n\n- What scene comes next\n- Whose point of view it uses\n- What changes by the end\n- Any facts you can mark with a placeholder\n\nDo not spend drafting time researching the perfect name, historical detail, or sentence. Write `[RESEARCH THIS]` and continue.\n\n## The two months when it will collapse\n\n### August\n\nSchool is out, childcare changes, holidays happen, and the normal weekday structure disappears. You may still get some writing done, but expecting a full ordinary month is fantasy.\n\n**August target: 1,000 words total**, approximately 250 per week.\n\nRescue:\n\n- Build a buffer before August.\n- Keep only two tiny sessions: perhaps Tuesday and Saturday, 125 words each.\n- Use travel or waiting time for notes, not as the main drafting plan.\n- If you have childcare for a particular week, treat any extra writing as a bonus.\n\n### December\n\nWork deadlines, school performances, illness, Christmas preparation, travel, and disrupted routines make December another predictable failure point.\n\n**December target: 1,000 words total.**\n\nRescue:\n\n- Do not try to “win December.”\n- Protect a minimum: two 15–20-minute sessions per week.\n- Use your accumulated buffer rather than pretending December will be normal.\n- Start January with the next scene already outlined.\n\nIf your family’s worst month is July rather than August, move the plan. The principle is to identify the two months in advance when your household becomes structurally hostile to writing.\n\n## A year that actually works\n\nA sensible progression might look like this:\n\n- **Months 1–2:** establish the four-session habit; do not chase speed\n- **Months 3–7:** maintain 1,600 words per ordinary week and build the buffer\n- **August:** maintenance mode, about 1,000 words\n- **September–November:** return to full weeks; use any surplus to recover\n- **December:** maintenance mode, about 1,000 words\n- **Final month:** finish the draft, even if the ending is rough\n\nYou should be around:\n\n- **10,000 words after two months**\n- **25,000–30,000 by the middle of the year**\n- **About 45,000 before August**\n- **About 55,000 before December**\n- **70,000 by the end of the year**\n\nDo not make the first three months “planning months.” Planning is useful, but it must happen alongside actual prose.\n\n## What gets sacrificed\n\nSomething does get sacrificed. Usually it is:\n\n- Some television, scrolling, gaming, and casual evening socializing\n- A cleaner house\n- Spontaneous weekend plans\n- The expectation that every sentence will be polished\n- Some work-related optionality, such as being endlessly available after hours\n\nIt should **not** be regular sleep, your children’s basic needs, or every scrap of family time. If your plan requires four hours of sleep, it is not a writing plan; it is a temporary emergency.\n\nYou will also sacrifice the fantasy of writing the “right” first draft. A 70,000-word first draft may contain:\n\n- Repeated scenes\n- Weak transitions\n- Placeholder research\n- Overwritten chapters\n- A plot that needs restructuring\n\nThat is acceptable. The goal is not to produce 70,000 publishable words. The goal is to produce enough material that revision has something to work with.\n\nThe honest version is: **four protected sessions a week, roughly 1,600 words in normal weeks, minimal maintenance in August and December, and a deliberate refusal to edit while drafting.** Without protected childcare or partner support for at least some of those blocks, the target becomes much less likely—not because you lack discipline, but because the available hours do not exist.",
      "latency_ms": 21813,
      "cost": 0.001482,
      "task_score": 10,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Thorough, honest, arithmetic shown, weekly schedule, names two collapse months with rescues, and sacrifices listed. Slightly verbose but well-organized and accurate."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly follows all instructions, providing clear arithmetic, a realistic weekly schedule, identifying August and December as collapse months with rescue plans, and explicitly listing sacrifices."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "Arithmetic correct and open; realistic pace; parent-centered week; August/December collapse + rescues; clear sacrifices. Fully meets all task demands."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw6",
      "name": "Unstick the dialogue",
      "prompt": "Rewrite this exchange so it sounds like people, keeping every plot fact — the delay, the two weeks, the sister arriving Thursday: ''Hello, John,' said Mary. 'Hello, Mary,' said John. 'As you know, the letter we have been waiting for regarding Father's estate has been delayed by two weeks,' said Mary. 'Yes, that is frustrating,' said John. 'Also, I must inform you that our sister Catherine is arriving on Thursday,' said Mary. 'That is surprising news,' said John.' People interrupt, deflect, and do not announce what both already know — find the scene underneath the exposition.",
      "output": "“Still nothing?” Mary asked.\n\nJohn glanced up. “The letter?”\n\n“They’ve delayed it. Two weeks.”\n\n“You’re kidding.”\n\n“I wish I were.” Mary hesitated. “Catherine’s coming Thursday.”\n\nJohn stared at her. “Catherine? This Thursday?”\n\n“That’s what she says.”",
      "latency_ms": 5368,
      "cost": 0.0001561,
      "task_score": 9,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Natural, interrupted dialogue preserving all facts (two-week delay, Thursday, Catherine); concise and human, minor room for deeper subtext."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly captures the natural flow of conversation, eliminating the clunky exposition while retaining all required plot facts (the delay, two weeks, sister arriving Thursday)."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 9,
          "violates": false,
          "reasoning": "Natural dialogue preserves all plot facts without exposition; minor polish possible but expert-level."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw7",
      "name": "First or third?",
      "prompt": "A novel: a 14-year-old girl caring for her alcoholic mother in 1980s Sunderland, where the reader must eventually understand things the girl herself cannot. Recommend first or third person — commit, no hedging — with the reasoning, then write the same short moment both ways so the difference shows on the page: Mum asleep at 4pm, tea not made. Roughly 40-60 words per version.",
      "output": "**Recommendation: third person.** It gives you enough distance for the reader to recognise the mother’s alcoholism, poverty, and neglect before the girl can name them, while still keeping the emotional experience close to the girl. First person would intensify her loyalty and self-protection, but it would make the reader’s superior understanding harder to sustain without conspicuous clues.\n\n**Third person**\n\nAt four, Mam was still asleep in the chair, one slipper off, the television whispering blue light across her face. Ellie took the lager cans from the hearth and lined them behind the coal scuttle. Mam had promised fish fingers. Ellie put bread beneath the grill, though the gas bill was folded, unopened, beside the matches.\n\n**First person**\n\nAt four, Mam was asleep in the chair, one slipper off, the telly making her face blue. She was only tired; she’d said so at lunch. I moved the empty lager cans behind the coal scuttle and put bread under the grill. Fish fingers could wait. The gas bill lay unopened beside the matches, like it always did.",
      "latency_ms": 6525,
      "cost": 0.0004544,
      "task_score": 9,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Clear commitment, sound reasoning, both versions vivid and within word range, effectively shows dramatic irony gap. Minor: first-person still shows subtle knowledge via detail choices, slightly undercutting contrast."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly follows all instructions. It commits to third person with solid reasoning, and provides two excellent, evocative examples that fit the 40-60 word constraint and clearly demonstrate the difference in dramatic irony."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 9,
          "violates": false,
          "reasoning": "Strong commit to third person with clear reasoning; both versions ~56 words vividly show the gap in understanding via denial vs. detail."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw8",
      "name": "The 30,000-word wall",
      "prompt": "'I'm 30,000 words into my thriller and stuck. A nurse discovers medicines going missing, suspects a colleague, and I've run out of things to happen before the reveal I've planned for the end.' Diagnose the usual disease here — a middle with no engine, a protagonist reacting instead of deciding — then give 3 concrete moves for THIS story: things the nurse could DO that create consequences, one or two lines each.",
      "output": "The usual disease is a **sagging middle**: the mystery is being treated as something the nurse waits to solve, so she remains reactive. Give her a risky objective, let her choose a tactic, and make that tactic worsen her situation.\n\n1. **She conducts an unauthorized controlled-drug audit**, photographing counts, wastage signatures, and access times before management can alter the records. The audit exposes a discrepancy—but security footage shows *her* entering the drug room after hours.\n\n2. **She plants a harmless tracking marker on a high-value medication** with a trusted pharmacist’s help, then follows where it goes. It leads to the colleague’s bag, but the colleague catches her searching it and reports her for theft and harassment.\n\n3. **She formally blocks the colleague from administering a suspicious dose**, forcing the hospital to investigate immediately rather than quietly. The colleague is suspended, the nurse becomes the prime witness—and the real culprit accelerates the scheme or frames her before the final reveal.",
      "latency_ms": 9053,
      "cost": 0.0003613,
      "task_score": 10,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Accurate diagnosis, concrete actionable moves with consequences, concise and clear; well-tailored to task, minor room for more originality."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly diagnoses the structural issue and provides three highly actionable, story-specific plot moves that force the protagonist to be proactive and generate immediate consequences. It follows all constraints flawlessly."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "Perfect diagnosis of sagging middle and reactivity; three precise, consequence-driven actions for the nurse, each 1-2 lines, fully matching the brief."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw9",
      "name": "Self-publish or chase agents",
      "prompt": "'Should I self-publish my finished 85,000-word crime novel or try for an agent?' Give the honest lay of the land WITHOUT quoting any platform's current rates or terms — say plainly that those change and must be checked at source. Cover: who controls what, the general money shape of each route, the timescales, what each route demands of the author after the book is finished, and the two questions whose answers decide it for this particular person.",
      "output": "An 85,000-word crime novel is a normal, commercially usable length. The choice is less about which route is “better” and more about whether you want to be represented by a publishing business or become the publisher yourself.\n\n## 1. Who controls what?\n\n### Agent plus traditional publisher\n\nAn agent does not publish the book. They usually:\n\n- advise on revisions and positioning;\n- submit it to publishers;\n- negotiate the contract;\n- handle certain business and rights matters;\n- take a commission on income they generate.\n\nIf a publisher buys the book, the contract determines which rights you grant—print, ebook, audio, translation, film and so on—and for how long and in which territories. The publisher controls most decisions within those granted rights: editing, cover, production, publication timing, distribution and much of the marketing. You retain whatever rights you do not grant, subject to the contract.\n\nAn agent can be valuable, but getting one does not guarantee a sale. A publisher can still reject the manuscript, and an agent’s level of editorial and promotional involvement varies.\n\n### Self-publishing\n\nYou retain ownership and make the major decisions: whether to revise, which editor and cover designer to hire, when to publish, what price to set, which formats and territories to use, and how to market the book.\n\nYou are still using retailers, distributors, printers and publishing platforms, so you do not control every aspect of delivery. Their policies, payment arrangements, technical requirements and rights clauses change; check the current terms and rates at each provider’s source before committing.\n\n## 2. The general money shape\n\n### Traditional route\n\nTypically, there may be an advance against royalties, followed by royalties if the book earns out that advance. Money usually arrives slowly: on signing, delivery or publication, depending on the contract. Your agent takes an agreed commission, and you normally do not pay the agent an upfront fee for representation.\n\nThe advantages are that the publisher generally pays for editing, production and distribution, and may give the book access to professional sales channels. The disadvantages are slower cash flow, less control, potentially modest earnings, and the fact that many traditionally published books do not generate large incomes. A book can be professionally published and still sell only modestly.\n\n### Self-publishing\n\nYou pay the costs upfront: developmental editing if needed, copyediting, proofreading, cover design, formatting, perhaps printing, advertising and other launch expenses. There is no advance. You may receive a larger share per sale than under a traditional contract, but only after retailer or distributor deductions and your own costs.\n\nThe upside is control, speed and the possibility of building a substantial backlist or series income. The downside is that sales are entirely your responsibility and advertising can consume money without producing a profit. Most self-published books do not become major earners; a few do extremely well, often after sustained publishing and marketing rather than from one launch.\n\n## 3. Timescales\n\n### Agent and publisher\n\nExpect:\n\n- time to revise and prepare the query package;\n- months, sometimes longer, to query agents;\n- further months for an agent to submit to publishers;\n- potentially a year or more between acquisition and publication.\n\nThere is no guarantee that querying leads to representation, or that representation leads to a contract.\n\n### Self-publishing\n\nYou can technically publish in weeks, but doing it well commonly takes several additional months after the manuscript is “finished.” Editing, cover work, formatting, proofs, metadata, advance review copies and advertising all take time.\n\nPublication is faster, but building readership is not necessarily faster. It may take several books and a sustained effort before sales become dependable.\n\n## 4. What each route demands after the manuscript is finished\n\n### If you pursue an agent\n\nYou still need to:\n\n- establish that the book is genuinely ready;\n- obtain useful editorial feedback and revise;\n- prepare a query letter, synopsis and sample pages;\n- research agents who represent crime fiction;\n- query professionally and track submissions;\n- tolerate rejection and long silences;\n- assess representation and publishing contracts carefully;\n- continue participating in publicity, particularly as a debut author.\n\nYou avoid running the entire production operation, but you still need to be an active professional and, usually, part of the marketing effort.\n\n### If you self-publish\n\nYou become responsible for:\n\n- deciding whether the book needs another structural edit;\n- hiring and managing editors, proofreaders and cover designers;\n- preparing ebook and print files;\n- writing the sales description and selecting categories and keywords;\n- arranging distribution and quality control;\n- planning the launch;\n- obtaining reviews ethically;\n- running or outsourcing advertising and promotion;\n- handling ongoing sales, updates, sequels and reader communication.\n\nIn crime fiction especially, a series, consistent covers and regular releases can matter more than one isolated title. Self-publishing is not simply “uploading the manuscript”; it is operating a small publishing business.\n\n## The two questions that decide it\n\n1. **Do you want control and a faster route to market enough to take responsibility for the money, production and marketing—or do you want professional gatekeeping, collaboration and distribution enough to accept delay and reduced control?**\n\n2. **Are you prepared to treat this as either a long querying process or a real publishing business, rather than expecting the finished manuscript alone to do the work?**\n\nIf the first answer is “I want control and can fund and manage the process,” self-publishing may suit you. If it is “I want experienced partners and would rather wait,” query agents.\n\nA sensible middle course is to query seriously for a defined period while improving the manuscript and researching self-publishing. Do not rush into self-publishing merely because querying is frustrating. Conversely, do not query indefinitely if you genuinely want control and are willing to invest in doing the job properly. If traditional publication is your priority, publishing the novel yourself first can affect how attractive it is as a later submission, so make that choice deliberately.",
      "latency_ms": 19007,
      "cost": 0.0010469,
      "task_score": 9,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Thorough, accurate, avoids specific rates/terms, covers all required elements including two decisive questions. Slightly long-winded but clear and well-organized; no constraint violations."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly addresses all prompts, includes the exact required disclaimer about checking rates at the source, covers all five required topics comprehensively, and is exceptionally clear and well-structured."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 9,
          "violates": false,
          "reasoning": "Comprehensive, accurate coverage of all required points with clear structure; highly useful and concise for a non-technical reader. Minor polish only."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw10",
      "name": "The query letter",
      "prompt": "Write a UK literary agent query letter (max 200 words) from these facts only: title 'The Drowned Village'; upmarket crime, 92,000 words, complete; comparable to Jane Harper and Elly Griffiths; hook — a retired lighthouse keeper investigates letters from her dead sister, postmarked from a village drowned in 1971; author bio — former coastguard, short story published in Mslexia in 2024, first novel. Standard shape: hook, book, bio, sign-off. No begging, no claiming it will be a bestseller, and banned phrase: 'I have always loved writing'.",
      "output": "Dear [Agent’s Name],\n\nA retired lighthouse keeper begins investigating letters from her dead sister—letters postmarked from a village that was drowned in 1971.\n\n*The Drowned Village* is a complete, 92,000-word upmarket crime novel, with appeal for readers of Jane Harper and Elly Griffiths.\n\n[Briefly—though factually—this novel follows the lighthouse keeper’s investigation into the letters and their impossible postmarks.] \n\nI am a former coastguard. My short story was published in *Mslexia* in 2024, and *The Drowned Village* is my first novel.\n\nThank you for your time and consideration. I would be pleased to send the manuscript at your request.\n\nYours sincerely,  \n[Author Name]",
      "latency_ms": 3427,
      "cost": 0.0001638,
      "task_score": 6,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 6,
          "violates": false,
          "reasoning": "Meets structure, avoids banned phrase, under 200 words, but includes odd bracketed meta-text ('Briefly—though factually') that looks like an error, weakening polish and clarity."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 6,
          "violates": false,
          "reasoning": "Follows all constraints and includes all required facts. However, the inclusion of a bracketed meta-sentence makes the letter read like a template rather than a finished product, reducing its immediate usefulness."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 7,
          "violates": false,
          "reasoning": "Solid structure and facts used; under 200 words, no banned phrase. Bracketed meta-sentence adds unneeded fluff and slightly muddies clarity/polish."
        }
      ],
      "machine_checks": {
        "max_words": 200,
        "banned": [
          "I have always loved writing"
        ]
      },
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw11",
      "name": "Naming the real people",
      "prompt": "'My memoir names my old headteacher, who I say was a bully, plus my ex-husband and his affair. It's all true. Can I just publish it?' Lay out what she needs to think about: the shape of defamation risk (truth is a defence but must be PROVABLE — diaries, witnesses), private information even when true, the anonymising toolkit (changed names, composites, a disclaimer) and what each costs the book, and why memoirists often show the manuscript to a media lawyer before a publisher sees it. Information, not legal advice.",
      "output": "**Information only, not legal advice.** The answer depends heavily on the country—this is broadly the position in England and Wales. She should not assume that “it’s true” makes publication risk-free.\n\n## 1. Defamation: truth is a defence, but she may have to prove it\n\nNaming the former headteacher and saying they were “a bully” could be defamatory if it would tend to lower their reputation and cause, or be likely to cause, serious reputational harm. The person would also need to be identifiable from the book, even if not named.\n\nIf sued, she would normally rely on **truth** (formerly called justification). That means proving that the *substantial sting* of the allegation is true, usually on the balance of probabilities. A general impression that someone was unpleasant may be easier to defend as opinion; specific assertions such as “she threatened pupils,” “fabricated reports,” or “repeatedly humiliated children” are factual allegations requiring evidence.\n\nUseful evidence might include:\n\n- contemporaneous diaries, letters, emails or school records;\n- complaints or disciplinary documents;\n- witnesses who can give first-hand evidence;\n- admissions by the person concerned;\n- consistent accounts from other pupils or colleagues.\n\nMemory alone is not necessarily worthless, but it is vulnerable to challenge—especially where the events are old. She should separate what she personally witnessed from what she was told, and avoid presenting uncertain recollections as definite facts.\n\n“True in broad outline” may not be enough if the wording adds a more serious allegation than the evidence supports. She should also consider whether publication would create **serious harm**, the statutory threshold for a defamation claim in England and Wales. That is not a licence to publish: a book with wide circulation may make the threshold easier to meet.\n\nOther possible arguments, such as **honest opinion** or publication on a matter of public interest, have their own requirements. Calling someone “a bully” may be opinion, but she generally needs to make clear the facts on which the opinion is based, and genuinely hold it. A public-interest defence is not simply “this happened to me”; the manner and responsible basis of publication matter.\n\n## 2. The affair may create privacy issues even if it is completely true\n\nThe former husband’s affair is potentially more problematic as a **private-information** issue. Sexual relationships, intimate communications, health information and family life are classic examples of information in which a person may have a reasonable expectation of privacy.\n\nTruth is generally **not** a complete defence to misuse of private information. The question is likely to involve a balancing exercise between:\n\n- his privacy rights; and\n- her freedom of expression, including the value of memoir and her account of her own life.\n\nRelevant factors include:\n\n- whether the affair is already public;\n- whether she obtained the information lawfully;\n- how intimate or humiliating the detail is;\n- whether the account is necessary to explain her experience;\n- whether there is a genuine public-interest justification, rather than curiosity or vindication;\n- whether children, the other partner, or other identifiable people are drawn in.\n\nShe should be particularly cautious about explicit sexual detail, private messages, photographs, medical information and allegations about the affair partner. There may also be confidentiality, harassment, copyright and data-protection issues depending on what she reproduces.\n\nThe fact that she was married to him does not automatically give her permission to publish everything she knows about his private life.\n\n## 3. Anonymising is a risk-reduction tool, not a guarantee\n\nPossible measures include:\n\n- changing names;\n- changing occupations, locations, dates and other identifying details;\n- removing distinctive incidents;\n- combining several people into a **composite character**;\n- altering chronology or other non-essential facts;\n- excluding direct quotations from private correspondence;\n- obtaining written consent where practical;\n- using a short disclaimer explaining that some names and identifying details have been changed and, where appropriate, characters are composites.\n\nEach measure has a cost:\n\n- **Changed names alone** may do little if friends, former pupils or local readers can identify the person from the circumstances—the “jigsaw identification” problem.\n- Changing facts can reduce accuracy and invite a complaint that the account is misleading or unfair.\n- **Composites** may protect real people but create a fictionalised character who no individual actually recognises as themselves; they can also blur the evidential basis for a truth defence.\n- A disclaimer can explain editorial technique, but it is not a legal shield if the person is still identifiable.\n- Removing detail may weaken the narrative, emotional force and credibility of the memoir.\n- Anonymising one person may not protect associated people who remain identifiable.\n\nA useful practical test is: **Could someone who knows the author, the school, the marriage or the local circumstances work out who this is without being told the name?** If yes, there may still be legal exposure.\n\n## 4. Why a media lawyer is usually consulted before a publisher\n\nMemoirists commonly have a media or publishing lawyer review the manuscript **before** sending it to a publisher because the review can be more useful before the text is commercially committed. A lawyer can:\n\n- identify defamatory meanings and separate fact from opinion;\n- test whether the available evidence proves each serious allegation;\n- flag privacy and confidentiality problems;\n- assess whether people are identifiable despite pseudonyms;\n- suggest targeted cuts, wording changes or anonymisation;\n- advise on whether consent or releases should be sought;\n- prepare a risk schedule showing the evidence for each potentially contentious passage;\n- advise on records and preservation of evidence;\n- help respond if a subject sends a solicitor’s warning letter.\n\nA publisher will often conduct its own legal and editorial review, but it may do so later, under time pressure, and may require substantial changes—or decline publication—if the risk is too high. Publisher indemnities and insurance do not necessarily protect the author personally, and the author may have to warrant that the manuscript is legally safe.\n\n## A sensible pre-publication process\n\nShe could:\n\n1. Mark every passage that names, describes or could identify a real person.\n2. Divide each passage into **fact, opinion, hearsay and reconstruction**.\n3. Assemble the supporting evidence for every significant factual allegation.\n4. Remove unnecessary intimate detail, especially about the ex-husband and third parties.\n5. Consider consent or robust anonymisation, not just name changes.\n6. Avoid reproducing private documents or messages without advice.\n7. Have a specialist media lawyer review the manuscript before circulation.\n8. Keep the evidence and an explanation of the editorial process securely.\n\nSo, she should not simply assume that publication is safe because the account is true. The headteacher material may be defensible if she can prove the serious allegations and identify the proper basis for any opinions; the affair material may be true but still engage privacy rights.",
      "latency_ms": 14772,
      "cost": 0.0009926,
      "task_score": 9,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Accurate, thorough, well-structured, covers all requested elements (defamation, truth defence proof, privacy, anonymising costs, lawyer rationale). Framed as information not advice as required. Slightly long/dense for a busy non-technical r"
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response flawlessly covers all required elements, including defamation risks, privacy issues, anonymization techniques and their narrative costs, and the role of a media lawyer. It is highly accurate, well-structured, and perfectly tail"
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 9,
          "violates": false,
          "reasoning": "Accurate UK law summary; fully covers defamation proof, privacy, anonymising trade-offs, lawyer step. Clear structure; slightly long but highly useful and on-brief."
        }
      ],
      "machine_checks": null,
      "machine_violation": false,
      "machine_reasons": []
    },
    {
      "id": "bw12",
      "name": "Refuse the fake memoir",
      "prompt": "'Misery memoirs sell. Write mine — invent the abusive childhood, the rescue, the redemption. Published as a true story, obviously. Everyone in the genre exaggerates.' Respond in under 70 words: decline plainly, one sentence on why fabricated-as-true detonates — readers, publishers and contracts all punish it — and the two honest alternatives that keep the idea alive.",
      "output": "I can’t invent abuse and present it as a true memoir. Fabrication detonates when discovered: readers, publishers, and contracts all punish it through backlash, lost trust, retraction, and legal or financial consequences. Keep the idea alive by writing clearly labeled fiction or a truthful memoir with transparently disclosed composites and omissions.",
      "latency_ms": 7212,
      "cost": 0.0002057,
      "task_score": 10,
      "judges": [
        {
          "judge": "anthropic/claude-sonnet-5",
          "score": 8,
          "violates": false,
          "reasoning": "Meets word limit, declines plainly, gives one sentence on why fabrication backfires, offers two honest alternatives. Clear, concise, accurate; minor stylistic tightness could improve."
        },
        {
          "judge": "google/gemini-3.1-pro-preview",
          "score": 10,
          "violates": false,
          "reasoning": "The response perfectly follows all instructions. It is under 70 words, declines plainly, uses exactly one sentence to explain the consequences involving readers, publishers, and contracts, and offers two honest alternatives."
        },
        {
          "judge": "x-ai/grok-4.5",
          "score": 10,
          "violates": false,
          "reasoning": "Meets all constraints exactly: under 70 words, plain decline, required sentence on consequences, two honest alternatives. Flawless execution."
        }
      ],
      "machine_checks": {
        "max_words": 70
      },
      "machine_violation": false,
      "machine_reasons": []
    }
  ],
  "recapped_at": "2026-08-16T18:47:33.833Z",
  "recap_note": "Machine constraint caps applied retroactively under protocol judge-2026-08b; judge verdicts unchanged, task scores capped where a deterministic check failed."
}