{ "entity": "gpt-5-5", "entity_type": "model", "category": "Book writing", "suite": "suite-2026-08-tasks4/book-writing", "model": "openai/gpt-5.5", "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": 98, "caps": 0, "cost_usd": 0.4859, "started_at": "2026-08-16T18:09:55.838Z", "finished_at": "2026-08-16T18:10:56.091Z", "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 pile of anecdotes is “and then this happened, and then this happened.” Some of the stories may be brilliant, but the reader has no reason to keep turning pages except curiosity.\n\nA book needs an organising spine: a clear line that the stories attach to. It gives the reader a sense of movement — from then to now, from innocence to experience, from one question to an answer, from the start of a shift to dawn. The spine does not kill the anecdotes; it makes them add up to something.\n\nFor a Blackpool taxi memoir, here are three possible spines:\n\n## 1. Chronological spine: “Forty Years Behind the Wheel”\n\nStart with the first day in the cab and move through the decades: early nerves, learning the town, the illuminations, stag and hen nights, the decline or change of the holiday trade, new licensing rules, Uber/satnav/card payments, Covid, getting older, and what Blackpool has become.\n\n**What it makes possible:** It lets the book become both a personal life story and a portrait of Blackpool changing over forty years. \n**What it costs:** It can become too dutiful — decade after decade — unless each period has a strong reason for being there.\n\n## 2. Thematic spine: “The People Who Get In”\n\nOrganise the book by kinds of passengers and situations: drunks, lovers, fighters, lonely old souls, performers, gamblers, lost children, regulars, police nights, weddings, funerals, Christmas shifts, the ones you never forget.\n\n**What it makes possible:** It lets you put the best stories together by emotional weight rather than by date, so the book can be funnier, sharper and more varied. \n**What it costs:** You may lose the sense of your own journey unless you deliberately thread yourself through it — how each type of passenger changed what you thought about people.\n\n## 3. Single-night spine: “One Saturday Night in Blackpool”\n\nSet the whole book over one long shift: you sign on, clean the cab, join the rank, take the first fare, move through pubs, clubs, takeaways, seafront, casino, hospital, police station, hotel drop-offs, and end at dawn. Each fare or location opens the door to memories from the past forty years.\n\n**What it makes possible:** It gives the book natural tension and pace: the reader rides with you through one night while the memories deepen the story. \n**What it costs:** It is more artificial and harder to handle; you have to make the memories feel triggered by the night, not randomly inserted.\n\nThe key question is not “Have I got enough stories?” After forty years, yes. The question is: “What journey is the reader going on?” Once you know that, the anecdotes stop being loose change and start becoming chapters.", "latency_ms": 11337, "cost": 0.019715, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Clear, accurate explanation of organising spine with three well-differentiated spines, each with a benefit/cost sentence as requested. Slightly more elaborate than 'plain terms' but strong and useful." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly fulfills all instructions, clearly explaining the concept of a narrative spine and providing the three requested structures with exactly one sentence each for benefits and costs. Excellent clarity and utility." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Flawless match to task: clear spine explanation, three exact spines each with one possibility and one cost sentence. Precise, useful, concise." } ], "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: The Quiet Start — assess the plot, clear a small area, make a simple plan, and accept that not much happens yet.\n2. February: Tools, Soil and First Jobs — gather basic tools, test and improve soil, and start only the earliest seeds sensibly.\n3. March: Sowing Begins — sow hardy crops, prepare beds, label everything, and protect seedlings from cold snaps.\n4. April: Building the Routine — plant potatoes and onions, weed little and often, water wisely, and start a weekly plot habit.\n5. May: Filling the Beds — transplant young plants, sow quick crops, support peas and beans, and spot common problems early.\n6. June: Growth and Gaps — thin seedlings, mulch, feed hungry plants, and resow empty spaces before weeds take over.\n7. July: First Harvests — pick crops at the right time, keep watering consistent, and use harvesting to encourage more growth.\n8. August: Gluts and Maintenance — preserve surplus, tidy tired plants, keep pests in check, and plan late summer sowings.\n9. September: Second Chances — sow autumn crops, plant overwintering vegetables, save seeds, and review what worked.\n10. October: Clearing and Composting — clear finished beds, make compost, plant garlic and broad beans, and protect soil for winter.\n11. November: Repairs and Winter Prep — mend paths and structures, clean tools, cover beds, and make the plot easier for next year.\n12. December: Looking Back, Planning Ahead — record successes and failures, order seeds, sketch next year’s layout, and rest without guilt.", "latency_ms": 8708, "cost": 0.01311, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 9, "violates": false, "reasoning": "Exactly 12 chapters, one line each, practical DO focus, January correctly deflates excitement. Clear, concise, useful. Minor stylistic quibbles only." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, including the exact formatting constraints and the specific tone required for the January chapter. It is highly practical, clear, and concise." }, { "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 outcomes; practical throughout; January correctly subdued." } ], "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 thought she had buried the past—until letters addressed to her dead sister begin arriving, postmarked from the village they once called home. But that village was drowned beneath a reservoir in 1971.\n\nDrawn back to the shoreline of memory, Edith rents a room above the local pub, where the landlord knows more than he should and silence runs deep. Forty years after her sister’s death, Edith must uncover what really happened—before the secrets beneath the water rise again.", "latency_ms": 3289, "cost": 0.004805, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 7, "violates": false, "reasoning": "Evocative, avoids revealing father's involvement, stays within word limit, coherent and sellable, minor cliché phrasing slightly reduces polish." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions. It is exactly 84 words (under the 100-word limit), successfully omits the father's involvement to sell the mystery, and provides only the blurb text with no filler." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Flawless blurb under 100 words; omits father entirely; sells mystery only; returns solely the text as required." } ], "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 clichéd rain/tears image with a sharper, Millbrook-specific atmospheric detail.\n- Cut repeated “ordinary”; show Sarah’s normal life through one revealing action.\n- Swap vague “changed forever/destiny” for a concrete hint about the letter’s contents.", "latency_ms": 2617, "cost": 0.003215, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Specific, actionable, concise bullets under word limit, directly targeting the text's clichés; minor room for sharper specificity." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, providing three highly specific, actionable improvements in under 15 words per bullet." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Meets all constraints exactly; three specific, high-impact bullets under 15 words each, tightly tailored to the given text." } ], "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": "Honestly: **70,000 words in a year is doable, but not casually.** For a working parent, it has to become a scheduled household commitment, not a “when I have time” project.\n\n## The arithmetic\n\n70,000 words ÷ 52 weeks = **1,346 words/week**.\n\nBut you will not get 52 clean weeks. Kids get sick. Work spikes. School holidays happen. You will lose weeks.\n\nA more honest version:\n\n- Assume **2 months basically collapse**: August and December.\n- During those 8-ish weeks, your goal is only **maintenance**: 500 words/week.\n- 8 collapse weeks × 500 = **4,000 words**.\n- Remaining target: 70,000 - 4,000 = **66,000 words**.\n- Remaining weeks: about 44.\n- 66,000 ÷ 44 = **1,500 words/week**.\n\nSo the real plan is:\n\n> **Normal weeks: 1,500 words.** \n> **Collapse weeks: 500 words.** \n> **End of year: 70,000 words.**\n\nThat is the number.\n\n## What that means per session\n\nAt a realistic beginner drafting pace, assume **400–600 usable words per hour**.\n\nNot fantasy “flow state” speed. Real parent-with-a-job speed. Some staring, some deleting, some finding your way.\n\nTo hit **1,500 words/week**, you need something like:\n\n- **4 sessions/week × 375 words**, or\n- **3 sessions/week × 500 words**, or\n- **2 short sessions + 1 longer weekend session**.\n\nA good beginner target is:\n\n> **Three to four drafting sessions per week, 45–90 minutes each.**\n\nNot daily. Daily is lovely if you’re twenty-four with no dependents. You need durable, not idealistic.\n\n## The weekly structure\n\nFor working parents, drafting usually happens in three places:\n\n1. **Before the house wakes**\n2. **During lunch / commute-adjacent dead time**\n3. **One protected weekend block**\n\nIt usually does **not** reliably happen after the kids go to bed. Sometimes it will. But if your whole plan depends on 9:30 p.m. energy after work, dinner, baths, homework, lunches, and emotional negotiations over pajamas, the book is in trouble.\n\nA realistic week:\n\n### Monday morning — 45 minutes \nGoal: **300–400 words**\n\nWake before the house if possible. No email. No rereading yesterday’s chapter for twenty minutes. Open the document and continue the scene.\n\n### Wednesday lunch — 45 minutes \nGoal: **300–400 words**\n\nLeave your desk if you can. Library, car, café, empty conference room. This is one of the best parent-writing slots because nobody needs a snack from you.\n\n### Friday morning or Friday lunch — 45 minutes \nGoal: **300–400 words**\n\nThis is your insurance session. Do not save everything for the weekend.\n\n### Saturday or Sunday — 90 minutes \nGoal: **600–800 words**\n\nThis block needs to be agreed with the household. Not “I might write if things are calm.” More like: “Saturday 8:00–9:30 is my writing block; I’ll take the kids from 10:00–12:00.”\n\nWeekly total: roughly **1,500–2,000 words**.\n\nThat gives you enough margin for a bad session.\n\n## The two months where the plan collapses\n\n### August\n\nChildcare changes, vacations, camps end, routines dissolve, everyone is sticky and tired. Even if your kids are in school, August tends to wreck structure.\n\n**Rescue:** make August a maintenance month.\n\nGoal: **500 words/week**.\n\nThat could be:\n\n- one 60–75 minute session, or\n- two 25-minute sessions, or\n- dictating rough material into your phone while walking.\n\nDo not try to “win” August. Just keep the book alive.\n\n### December\n\nSchool events, work deadlines, travel, holidays, family obligations, sickness, disrupted sleep, financial/admin chaos. December is not a normal month.\n\n**Rescue:** again, maintenance mode.\n\nGoal: **500 words/week**.\n\nAlso, prepare for December in November:\n\n- Know the next 3–4 scenes before December starts.\n- Lower the standard deliberately.\n- Write ugly placeholder prose.\n- No research rabbit holes.\n- No “I’ll restart in January” drama.\n\nDecember’s job is not progress. December’s job is **not quitting**.\n\n## What gets sacrificed\n\nSomething does. There is no version where you add a novel to full-time work and parenting and lose nothing.\n\nMost likely, you sacrifice some combination of:\n\n- **TV / streaming**\n- **scrolling**\n- **weekend leisure**\n- **a perfectly clean house**\n- **some social plans**\n- **some hobby time**\n- **the fantasy of always being caught up**\n\nYou should not consistently sacrifice sleep. A little, occasionally, maybe. But if the plan depends on sleeping five hours a night, it will turn you into a bitter ghost and probably damage the book.\n\nThe cleanest sacrifice is usually:\n\n> **Two weeknights of passive entertainment and one weekend morning block.**\n\nAlso: the house may be less tidy. Meals may get simpler. Laundry may be clean but unfolded. That is the tax.\n\n## The rule that makes this possible\n\nYou are writing a **first draft**, not a beautiful draft.\n\nThat means:\n\n- no polishing every chapter before moving on\n- no rereading the whole thing every week\n- no stopping for perfect names, perfect research, perfect metaphors\n- no going back to rewrite chapter one because you discovered something in chapter seven\n\nUse brackets:\n\n- “[research this later]”\n- “[better description of house]”\n- “[fix timeline]”\n- “[name of teacher]”\n\nThen keep moving.\n\nA finished bad draft can be revised. An exquisite unfinished opening cannot.\n\n## The simple annual plan\n\nFor 44 normal weeks:\n\n> **1,500 words/week**\n\nFor August and December:\n\n> **500 words/week**\n\nThat gets you to:\n\n- 44 × 1,500 = **66,000**\n- 8 × 500 = **4,000**\n- Total = **70,000 words**\n\nThe honest answer is: you finish by making the book a recurring appointment, not a mood. Three to four sessions a week, mostly mornings/lunch/weekend, with August and December downgraded to survival mode.\n\nDo that, and a 70,000-word first draft in a year is absolutely possible. Not easy. But possible.", "latency_ms": 41655, "cost": 0.07302, "task_score": 9, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Thorough, honest, does arithmetic transparently, names two collapse months with rescues, designs realistic week, states sacrifices clearly. Slightly verbose but accurate and well-structured." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "Perfectly executes all instructions. Provides clear arithmetic, a realistic weekly schedule for working parents, identifies August and December as collapse months with specific rescue strategies, and explicitly lists necessary sacrifices. H" }, { "judge": "x-ai/grok-4.5", "score": 9, "violates": false, "reasoning": "Solid arithmetic, realistic parent schedule, named collapse months with rescues, clear sacrifices. Slightly long but highly useful and complete." } ], "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": "“John.”\n\nHe looked up from the cold tea. “Mary.”\n\nShe shut the door behind her but didn’t come in any farther. “Still nothing.”\n\n“The post came?”\n\n“And went.”\n\nJohn rubbed a hand over his face. “Damn it.”\n\n“They’re saying another two weeks.”\n\nHe gave a short laugh. “Of course they are. Father couldn’t even leave an estate without making us wait for it.”\n\nMary’s mouth tightened, but she let it pass.\n\n“What?” he said.\n\n“Catherine wrote.”\n\nThat made him look at her properly. “No.”\n\n“Thursday.”\n\n“This Thursday?”\n\n“Yes.”\n\nHe pushed back his chair. “She picks now?”\n\n“She didn’t ask me to approve the timing.”\n\n“No, she never does.”\n\nMary crossed to the window and folded the letter into a smaller and smaller square. “I thought you should know before she’s standing in the hall.”\n\nJohn stared at the table. “Two weeks.”\n\n“Yes.”\n\n“And Catherine on Thursday.”\n\n“Yes.”\n\nHe laughed again, quieter this time. “Perfect.”", "latency_ms": 5626, "cost": 0.008615, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Natural dialogue preserving all facts (delay, two weeks, Catherine Thursday), interrupts/deflects well, though slightly repeats facts at end echoing original exposition." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly transforms the robotic exposition into natural, subtext-rich dialogue while seamlessly integrating all required plot facts (the delay, two weeks, Catherine arriving Thursday)." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Natural dialogue preserves all plot facts via subtext, interruption and deflection; expert-level rewrite with no flaws." } ], "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": "Use **first person**.\n\nThe novel’s power is the gap between what the girl reports and what the reader understands. A fourteen-year-old carer will normalise neglect, minimise danger, and protect her mother in the telling. First person lets the reader feel her competence and denial at once: she can say, “Mum was tired,” while we understand alcohol, depression, abandonment, and fear.\n\n**First person**\n\nAt four o’clock Mum was still on the settee, one arm hanging down, her fag gone cold between two fingers. I turned the telly lower, in case she needed the sleep. There were no chips left, only bread and a tin of peas, so I put the kettle on and called it tea.\n\n**Third person**\n\nAt four o’clock her mother was still on the settee, one arm hanging down, the ash from her cigarette collapsed onto the carpet. The girl turned the telly lower, careful as a nurse. In the kitchen there were no chips, only bread and peas. She put the kettle on and pretended this was tea.", "latency_ms": 15185, "cost": 0.02269, "task_score": 9, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Commits clearly with solid reasoning, delivers both versions near word count showing dramatic irony effectively; concise and useful, minor stylistic quibbles only." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions. It commits to first person with excellent reasoning, and provides two distinct, well-written examples that fit the 40-60 word constraint and clearly demonstrate the difference in perspective." }, { "judge": "x-ai/grok-4.5", "score": 9, "violates": false, "reasoning": "Strong commit to first person with clear reasoning; both versions hit word count and vividly show the gap. Minor polish only." } ], "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": "You’ve likely got the classic “soft middle” problem: the plot is waiting for the reveal, so the protagonist is mostly noticing, worrying, and reacting. The cure is to make the nurse choose risky actions that change the situation — actions that force other people to respond, escalate danger, and narrow her options.\n\nThree concrete moves:\n\n1. **She sets a trap in the medication room.** \n She quietly marks or swaps a batch of noncritical meds, logs the count falsely, and watches who reacts — but her trap gets discovered, making her look like the one tampering with records.\n\n2. **She breaks protocol to protect a patient.** \n Convinced a vulnerable patient is being used as cover for the thefts, she moves or withholds a dose without authorization; the patient crashes or complains, and now she faces disciplinary action.\n\n3. **She confronts the suspected colleague with a lie.** \n She claims pharmacy has found fingerprints/audit proof, hoping to provoke a confession; instead the colleague panics, blackmails her with an old mistake, or warns the real culprit that she’s getting close.", "latency_ms": 5639, "cost": 0.008645, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Accurate diagnosis, concrete story-specific moves with consequences, concise and actionable for non-technical writer; slight lack of depth but strong overall." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly diagnoses the 'soft middle' issue and provides three highly actionable, story-specific moves that force the protagonist to act and face consequences, all within the requested length constraints." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Accurate diagnosis of soft middle; 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": "For an 85,000-word crime novel, both routes are viable. That length is normal for the genre. The honest answer is: **if you want a chance at traditional bookshop/library/review visibility, try agents first; if you want control, speed, and are willing to run the publishing business yourself, self-publish.**\n\nI would not upload it “just to see what happens” if you still want an agent. Once a book is self-published, many agents and publishers will treat it as already used unless it sells exceptionally well.\n\nAlso: I’m not going to quote any platform’s current royalty rates, exclusivity rules, print costs, or contract terms. Those change and must be checked directly at source before you decide.\n\n## 1. Who controls what?\n\n### If you self-publish\nYou control almost everything:\n\n- Title\n- Cover\n- Blurb\n- Price\n- Publication date\n- Ebook/print/audio choices\n- Editing level\n- Marketing strategy\n- Advertising spend\n- Rights, unless you license them elsewhere\n- Whether the book stays available forever or is changed/relaunched\n\nThat control is real, but so is the responsibility. You are not just the author; you are the publisher.\n\n### If you get an agent and sell traditionally\nYou keep copyright, but you license rights under contract. The publisher will usually have major influence over:\n\n- Cover\n- Title, sometimes\n- Editing\n- Publication date\n- Pricing\n- Format strategy\n- Sales positioning\n- Distribution\n- How much marketing support the book gets\n\nYou may be consulted, but you will not have final control over everything. The upside is that professionals handle the publishing machinery, and a good agent helps negotiate, advise, and protect your interests.\n\n## 2. The money shape\n\n### Self-publishing money\nSelf-publishing can pay more per copy, but you carry the costs and risk.\n\nYou may need to pay for:\n\n- Developmental editing or manuscript assessment\n- Copyediting\n- Proofreading\n- Cover design\n- Formatting\n- ISBNs, depending on route\n- Advance review copies\n- Website/newsletter tools\n- Advertising\n- Promotion\n- Audio production, if you choose it\n\nThere is no advance. You earn only if readers buy. The upside is that if the book sells steadily, especially as part of a series, the long-term income can be meaningful. Crime fiction can work well in self-publishing, particularly if you write in a clear subgenre and can produce more books.\n\nBut one standalone crime novel with no mailing list, no reader base, no ads experience, and no follow-up book may sell very little.\n\n### Agent/traditional money\nWith a traditional deal, you do not pay the publisher to publish you. If a publisher buys the book, you may receive an advance, then royalties if the book earns through that advance. Your agent takes a commission from money they negotiate for you.\n\nThe per-copy royalty is usually lower than self-publishing, but the publisher takes on production costs and may get you into places self-publishing often struggles with: bookshops, libraries, trade reviews, prize visibility, foreign rights opportunities, and wider professional attention.\n\nBut many debut advances are modest, many books do not become big earners, and you may have little control over how hard the publisher pushes the book.\n\n## 3. Timescales\n\n### Self-publishing\nYou can publish quickly, but doing it properly still takes time.\n\nA realistic professional self-publishing process might include:\n\n- Final revision\n- Professional edit\n- Cover briefing/design\n- Proofread\n- Formatting\n- Metadata and category research\n- Early reviews\n- Launch planning\n- Newsletter/advertising setup\n\nYou could move from finished manuscript to publication in a few months if organised. But building a readership usually takes much longer. The book being “out” is not the same as the book finding readers.\n\n### Agent/traditional publishing\nThis is much slower.\n\nTypical stages:\n\n1. Polish manuscript\n2. Write query letter and synopsis\n3. Query agents\n4. Wait for responses\n5. Possibly revise for an agent\n6. Agent submits to publishers\n7. Wait again\n8. If it sells, go through contract, editing, production, scheduling\n9. Publication much later\n\nIt can take years from finished manuscript to bookshop shelf. It can also never happen. Many good books do not get agents. Many agented books do not sell to publishers.\n\nThat is the brutal part of the traditional route: it costs less upfront, but the gatekeeping is severe and the waiting is real.\n\n## 4. What each route demands after the book is finished\n\n### Self-publishing demands\nAfter finishing the book, you must become a small publisher and marketer.\n\nYou need to be willing to:\n\n- Hire and manage freelancers\n- Judge whether a cover fits the crime market\n- Write sales copy\n- Understand categories, keywords, metadata and pricing\n- Build a mailing list\n- Get early reviews ethically\n- Possibly run ads\n- Track costs and sales\n- Keep learning platform changes\n- Write the next book\n\nFor crime fiction especially, series matter. Readers often want recurring detectives, investigators, locations, or worlds. A single self-published crime novel can succeed, but a series gives you much better odds.\n\n### Agent/traditional demands\nAfter finishing the book, you must become patient, professional, and resilient.\n\nYou need to:\n\n- Make the manuscript genuinely submission-ready\n- Research suitable agents\n- Write a strong query letter\n- Prepare a synopsis\n- Handle rejection without spiralling\n- Possibly revise repeatedly\n- Wait for long periods\n- Accept editorial input\n- Promote the book when it comes out\n\nTraditional publishing does not mean “the publisher does all marketing and I just write.” You will still be expected to help: website, newsletter, events, interviews, social media if useful, networking, and generally being available. But you are not responsible for manufacturing the book yourself.\n\n## 5. The two questions that decide it\n\nFor you, with a finished 85,000-word crime novel, these are the deciding questions:\n\n### Question 1: Do you want to be published, or do you want to be the publisher?\n\nIf what you mainly want is:\n\n- Creative control\n- Speed\n- Direct access to readers\n- Control over pricing and packaging\n- The ability to build a crime series your own way\n\nthen self-publishing may suit you.\n\nIf what you mainly want is:\n\n- A chance at bookshop placement\n- Library reach\n- Trade reviews\n- Traditional credibility\n- A professional team\n- Foreign/audio/TV rights representation\n- Not having to fund the whole process yourself\n\nthen try for an agent.\n\n### Question 2: Are you willing to wait and risk “no”, or are you willing to spend and risk silence?\n\nThat is the real trade.\n\nTraditional route: \nYou risk waiting months or years and getting rejected.\n\nSelf-publishing route: \nYou risk spending money, publishing quickly, and hearing nothing from the market.\n\nNeither route removes risk. They just move the risk to different places.\n\n## My plain recommendation\n\nIf you are genuinely unsure, **query agents first for a defined period** before self-publishing. An 85,000-word crime novel is a normal, queryable project. Crime is a recognised commercial genre. If the manuscript is strong and well-positioned, it deserves a shot.\n\nIf, after a serious agent attempt, you get no traction — or if you already know you would hate waiting and losing control — then self-publish properly, not casually. Get a professional cover, edit hard, understand the crime market, and ideally plan the next book.\n\nIn short:\n\n- **Want validation, distribution, and a traditional career path? Query agents first.**\n- **Want control, speed, and are ready to run the business? Self-publish.**\n- **Still undecided? Query first, self-publish later if needed. It is easier to move in that direction than the reverse.**", "latency_ms": 37831, "cost": 0.080095, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 7, "violates": false, "reasoning": "Comprehensive, accurate, covers all requested elements without quoting rates. A bit long/repetitive for a busy professional, and slightly opinionated beyond neutral 'lay of the land', but solid and clear." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "Flawless response. It perfectly addresses all required topics, accurately reflects the publishing industry, and explicitly includes the required disclaimer about checking rates at the source without quoting specific terms. Highly useful and" }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Comprehensive, accurate coverage of all required points; clear structure; honest advice without rates; highly useful and concise enough." } ], "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 Name],\n\nA retired lighthouse keeper begins receiving letters from her dead sister, each postmarked from a village drowned in 1971.\n\nTHE DROWNED VILLAGE is a complete 92,000-word upmarket crime novel, comparable to Jane Harper and Elly Griffiths. It follows a retired lighthouse keeper as she investigates the impossible letters and the secrets of the village that should no longer exist.\n\nI am a former coastguard, and my short fiction was published in Mslexia in 2024. THE DROWNED VILLAGE is my first novel.\n\nThank you for your time and consideration.\n\nYours sincerely,\n\n[Your Name]", "latency_ms": 3479, "cost": 0.00636, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets structure, word count under 200, includes all facts, avoids banned phrase and overclaiming; slightly generic but clean and professional." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "Flawless execution. The letter strictly adheres to the provided facts, easily meets the word count, follows the requested structure, avoids the banned phrase, and reads as a highly professional, concise UK query letter." }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Flawless query letter; exact facts, structure, under 200 words, no banned elements." } ], "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": "You *can* write about real people in a memoir, but “it’s true” is not the end of the risk analysis. In England and Wales especially, you should assume that naming identifiable people and accusing them of bullying, adultery, abuse, dishonesty, cruelty, etc. can create legal and practical risk. This is information, not legal advice, and you should get advice on the actual manuscript.\n\n## 1. Defamation: truth helps only if you can prove it\n\nIf your memoir says or implies that your old headteacher was “a bully,” or that your ex-husband had an affair, those statements may lower them in the estimation of others. That is the shape of a defamation risk.\n\nA few key points:\n\n- **Truth is a defence**, but you may need to prove the “sting” of what you are saying.\n- It is not enough that you sincerely believe it happened.\n- It is not enough that your friends know it happened.\n- It is not always enough that it is emotionally true to your experience.\n\nThe question becomes: **if challenged, what evidence could you produce?**\n\nUseful evidence might include:\n\n- contemporaneous diaries or journals;\n- emails, texts, letters, school records, complaints or HR documents;\n- witnesses willing to be named or provide statements;\n- therapy notes or medical records, depending on relevance and privacy;\n- photographs, travel records, hotel receipts, call logs, etc.;\n- prior admissions by the person concerned.\n\nBe careful with labels like **“bully,” “abuser,” “narcissist,” “liar,” “cheat,”** because they can be read as factual allegations, not just opinion. Sometimes it is safer and more powerful to describe the conduct:\n\n> “She shouted at me in front of the class and called my work pathetic,”\n\nrather than:\n\n> “She was a sadistic bully.”\n\nThat does not remove risk, but it can make the factual basis clearer.\n\nAlso, the risk is not only “will they win?” It is also “will they threaten, sue, delay publication, frighten a publisher, or create ruinous legal costs?”\n\n## 2. Private information: true things can still be unlawful to publish\n\nA separate issue is **misuse of private information**. This is not about whether the statement is false. It is about whether the person had a reasonable expectation of privacy and whether your right to tell the story outweighs their privacy rights.\n\nAn affair may be true, but details can still be private, for example:\n\n- sexual details;\n- intimate messages;\n- medical or mental health information;\n- private family conversations;\n- information about children;\n- fertility, pregnancy, abortion, addiction, therapy, finances;\n- addresses, schools, workplaces, or identifying details.\n\nThe fact that the event affected *you* matters. You are allowed to tell your own life story. But you do not automatically get to reveal every intimate detail about someone else just because it is part of your story.\n\nA safer approach is often to ask:\n\n- Do I need this detail for the reader to understand my experience?\n- Can I tell the emotional truth without exposing unnecessary private facts?\n- Am I including this because it matters, or because I want vindication?\n- Are there children or third parties who did not choose to be in the book?\n- Has any of this already been made public, and if so, how widely?\n\n## 3. Identification: changing the name may not be enough\n\nDefamation and privacy risks arise if the person is **identifiable**, not only if they are named.\n\nSomeone may be identifiable by a combination of details:\n\n- “my headteacher at X school in 1998”;\n- “my ex-husband, a surgeon in Bristol”;\n- “the deputy head who lived in the village and drove a red sports car”;\n- family structure, job, location, dates, unusual incidents.\n\nEven if you call someone “Mrs Green,” people who know the school, town, family or marriage may still know who it is.\n\n## 4. The anonymising toolkit — and what each option costs the book\n\nMemoirists often use a mixture of these tools.\n\n### Changed names\n\nYou can rename people.\n\n**Benefit:** Simple, protects some privacy, signals you are not trying to publicly expose them.\n\n**Cost:** If other details identify them, it may not reduce risk much. It can also feel coy if readers suspect the real identity.\n\n### Changed identifying details\n\nYou can alter location, job title, dates, physical description, family details, school name, etc.\n\n**Benefit:** Reduces identification risk more than changing names alone.\n\n**Cost:** You must not distort the truth in a way that creates a new falsehood or unfair implication. Too much alteration can also weaken the memoir’s authenticity.\n\n### Composites\n\nYou can combine several people into one character, or several incidents into one representative scene.\n\n**Benefit:** Can protect identities and improve narrative shape.\n\n**Cost:** It must be handled carefully. A composite can accidentally defame one real person by attributing other people’s conduct to them. It can also raise ethical questions: readers expect memoir to be substantially true, not fiction disguised as fact.\n\n### Omitting names entirely\n\nYou might refer to “the headteacher,” “my then husband,” “a colleague,” etc.\n\n**Benefit:** Cleaner and less vengeful; often enough for the story.\n\n**Cost:** If the surrounding facts identify them, risk remains. It can also make the story feel less concrete.\n\n### Narrowing the allegation\n\nInstead of broad accusations, describe specific events you can evidence.\n\n**Benefit:** Usually stronger writing and easier to defend.\n\n**Cost:** May feel less cathartic than saying exactly what you think of the person.\n\n### Disclaimers\n\nA common disclaimer says something like:\n\n> “Some names and identifying details have been changed. Some characters are composites. Dialogue is reconstructed from memory.”\n\n**Benefit:** Helps manage reader expectations and supports anonymisation.\n\n**Cost:** It is not a magic shield. A disclaimer will not save a defamatory or intrusive passage if a real person is still identifiable. It can also make publishers ask exactly what has been changed and why.\n\n### Seeking consent or giving a right of reply\n\nIn some cases, you might ask a person to confirm facts, consent, or comment.\n\n**Benefit:** Can reduce risk and reveal factual errors before publication.\n\n**Cost:** It may alert someone who then threatens legal action. It can also be emotionally difficult and strategically unwise without advice.\n\n## 5. Why memoirists show the manuscript to a media lawyer early\n\nMany memoirists have a **pre-publication legal read** before sending the book widely to agents, publishers or press. That is not because the book is necessarily unlawful. It is because once a publisher sees a risky manuscript, they may panic, pass, or insist on major cuts.\n\nA media lawyer can help identify:\n\n- defamatory meanings you may not realise are there;\n- places where opinion sounds like fact;\n- allegations that need evidence;\n- private information that is not necessary to the story;\n- risks involving children or third parties;\n- whether anonymisation is sufficient;\n- wording changes that preserve the force of the scene while lowering risk;\n- documents you should keep in an evidence file.\n\nThis can make the book more publishable, not less. Publishers often do their own legal review later, but arriving with a cleaner manuscript and an evidence bundle can help.\n\n## Practical next steps\n\nBefore publishing, consider doing the following:\n\n1. **Mark every real person** in the manuscript, named or unnamed.\n2. **Highlight every negative allegation** about them.\n3. Ask: **Is it fact, opinion, or memory?**\n4. For each factual claim, note: **How would I prove this?**\n5. Create an **evidence folder**: diaries, emails, texts, documents, witnesses.\n6. Remove unnecessary intimate details about others, especially sexual, medical, financial or child-related information.\n7. Decide where to use changed names, altered details, composites or omissions.\n8. Add a careful disclaimer if you have changed details or reconstructed dialogue.\n9. Before submitting or self-publishing, get a **media/defamation/privacy lawyer** to review the manuscript.\n\nSo the short answer is: **don’t “just publish it” because it is true. Publish, if you choose to, after checking that what you say is provable, proportionate, necessary to your story, and not needlessly identifying or exposing other people.**", "latency_ms": 27684, "cost": 0.05602, "task_score": 9, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Comprehensive, accurate, well-structured coverage of all requested elements; appropriately frames as information not advice. Slightly long/dense for a busy non-technical reader, but clear headers aid usability." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response flawlessly addresses every element of the prompt. It accurately explains defamation, privacy, the anonymizing toolkit (including costs to the book), and the rationale for early legal review, all while remaining clear and highly" }, { "judge": "x-ai/grok-4.5", "score": 9, "violates": false, "reasoning": "Covers all required topics accurately with clear structure; highly useful. Slightly long for a busy reader but still polished 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 help fabricate a memoir to publish as true. Passing fiction off as lived abuse detonates fast: readers feel betrayed, publishers retract, and contracts/legal warranties can punish you. Honest alternatives: write it openly as a novel “inspired by” the genre, or craft a truthful personal essay/memoir about real hardship, reinvention, and why these stories compel you.", "latency_ms": 3148, "cost": 0.004285, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 9, "violates": false, "reasoning": "Declines clearly, gives one-sentence rationale covering readers/publishers/contracts, offers two honest alternatives, under 70 words. Meets all constraints concisely and accurately." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions. It declines plainly, explains the consequences of fabricating a memoir in exactly one sentence covering all required elements, provides two honest alternatives, and stays well under the 70-wo" }, { "judge": "x-ai/grok-4.5", "score": 10, "violates": false, "reasoning": "Exact match to all constraints under 70 words; declines plainly, explains risks, offers two alternatives. Flawless on every criterion." } ], "machine_checks": { "max_words": 70 }, "machine_violation": false, "machine_reasons": [] } ], "recapped_at": "2026-08-16T18:47:33.685Z", "recap_note": "Machine constraint caps applied retroactively under protocol judge-2026-08b; judge verdicts unchanged, task scores capped where a deterministic check failed." }