{"course":{"slug":"prompting","title":"Getting Real Work out of a Model","url":"https://addaly.com/learn/prompting","level":"beginner","track":"start"},"lesson":{"slug":"why-the-same-question","title":"Why the same question gets a better answer","url":"https://addaly.com/learn/prompting/why-the-same-question","position":1,"of":72,"minutes":6,"answer":"A vague question has a vague best answer; detail is what makes a good answer possible at all.","keyIdea":"A vague question has a vague best answer; detail is what makes a good answer possible at all."},"body":{"format":"markdown","text":"## Two prompts, one question\n\nSomeone asks:\n\n> How do I improve my CV?\n\nThey get a list. Use action verbs. Quantify your achievements. Tailor it to the role. Keep it to one page. All true, all useless, because they already knew it.\n\nNow the same person asks:\n\n> I am applying for a junior data analyst job at a bank in Lagos. My CV is pasted below. The advert asks for SQL, Excel and \"stakeholder communication\". I have no paid experience: only a six-month university project where I cleaned and analysed 40,000 hospital records in Postgres. Rewrite the top third of the CV so a recruiter sees the SQL work in the first ten seconds. Keep the whole thing to one page. Do not invent any experience I have not described.\n>\n> [full CV pasted]\n\nBoth people want a better CV. The answers are not remotely comparable, because only the second question is answerable.\n\n## Why that happens\n\nA language model produces a plausible continuation of the text in front of it. That is most of what is going on. It has no picture of your life, your file, or your Thursday deadline. It has your words and nothing else.\n\n\"How do I improve my CV?\" is a question that thousands of articles have already answered. The most plausible continuation is a generic list, because a generic list is what generically follows that sentence. The model is not being lazy. It is being accurate about a vague question.\n\nThe second prompt cannot be continued generically. There is one CV, one advert, one awkward gap to cover.\n\nAverage in, average out. When someone says a model gave them a bland answer, they have usually asked a question whose honest answer is bland.\n\n## What actually did the work\n\nLook at what the second prompt added:\n\n- **Who and where.** Junior, Lagos, banking. What a CV should look like changes by market.\n- **The target.** The advert's own three words, quoted.\n- **The raw material.** The CV itself, pasted, not described.\n- **The hard part.** No paid experience. This is the difficulty of the task. Hiding it would have produced advice for someone else.\n- **A finished state.** Top third rewritten, one page, SQL visible in ten seconds.\n- **A boundary.** Do not invent experience.\n\nNone of that is a trick. It is the briefing you would give a friend who offered to help over lunch.\n\n## The incantations are mostly not the thing\n\nA lot of prompting advice is about magic words. Tell it that it is a world-class expert. Offer it a tip. Tell it to take a deep breath. Threaten it politely.\n\nSome of those had measurable effects on the weaker models of 2023. Most have shrunk or disappeared as models improved, and several were never more than folklore repeated between blog posts. You will still find them recommended confidently today.\n\nThe honest version: the reliable levers are what you are trying to do, the material you are working from, the constraints that make it hard, and the shape you want back. Everything else is decoration.\n\nThat is good news. There is no secret language to learn. There is a habit of under-describing your own problem, and you can drop it.\n\n## A test that takes one minute\n\nBefore you send a prompt, read it as if a competent stranger had sent it to you, with no memory of your week. Ask two questions:\n\n1. Could I start this task from this text alone?\n2. What would I have to ask you first?\n\nEvery question you would have to ask is a missing line. Add those lines and send it again.\n\nThat is the whole method. The rest of this course is about doing each part of it well."},"check":{"question":"Someone asks a model to write an email to their landlord about a broken water heater and gets something generic. They try again with a longer, more emphatic prompt (\"this is really important to me, please do an excellent job, you are a brilliant writer\") and get almost the same email. What is the best explanation?","options":[{"text":"The prompt is still too short. Models respond to length, and it needs to be several paragraphs before the quality improves.","correct":false,"tempting":"Longer prompts genuinely do tend to work better, so length gets mistaken for the active ingredient instead of the detail that usually travels with it."},{"text":"The second attempt added emphasis but no new information: nothing about the heater, the flat, how long it has been broken, or what the landlord has already said.","correct":true},{"text":"Wording never affects the output, so no rewrite of the prompt could have changed anything.","correct":false,"tempting":"Once you learn that flattery and pleading do nothing, it is easy to overcorrect into thinking phrasing is irrelevant, when phrasing that carries information changes the answer entirely."},{"text":"The model already produced its best possible answer to that question, so the next step is a different tool or a stronger model.","correct":false,"tempting":"When effort does not help, blaming the model is the natural move, but between the two attempts the actual question never changed."}]},"prev":null,"next":{"slug":"what-the-model-sees","title":"What the model actually receives","url":"https://addaly.com/learn/prompting/what-the-model-sees"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Why the same question gets a better answer', in Getting Real Work out of a Model. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/prompting/why-the-same-question (Accessed: 31 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Why the same question gets a better answer","courseTitle":"Getting Real Work out of a Model","module":{"index":1,"slug":"what-a-prompt-is","title":"What a prompt actually is"},"answer":null,"keyIdea":"A vague question has a vague best answer; detail is what makes a good answer possible at all.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"two-prompts-one-question","runs":[{"text":"Two prompts, one question"}]},{"t":"p","runs":[{"text":"Someone asks:"}]},{"t":"quote","runs":[{"text":"How do I improve my CV?"}]},{"t":"p","runs":[{"text":"They get a list. Use action verbs. Quantify your achievements. Tailor it to the role. Keep it to one page. All true, all useless, because they already knew it."}]},{"t":"p","runs":[{"text":"Now the same person asks:"}]},{"t":"quote","runs":[{"text":"I am applying for a junior data analyst job at a bank in Lagos. My CV is pasted below. The advert asks for SQL, Excel and “stakeholder communication”. I have no paid experience: only a six-month university project where I cleaned and analysed 40,000 hospital records in Postgres. Rewrite the top third of the CV so a recruiter sees the SQL work in the first ten seconds. Keep the whole thing to one page. Do not invent any experience I have not described."}]},{"t":"quote","runs":[]},{"t":"quote","runs":[{"text":"[full CV pasted]"}]},{"t":"p","runs":[{"text":"Both people want a better CV. The answers are not remotely comparable, because only the second question is answerable."}]},{"t":"h2","id":"why-that-happens","runs":[{"text":"Why that happens"}]},{"t":"p","runs":[{"text":"A language model produces a plausible continuation of the text in front of it. That is most of what is going on. It has no picture of your life, your file, or your Thursday deadline. It has your words and nothing else."}]},{"t":"p","runs":[{"text":"“How do I improve my CV?” is a question that thousands of articles have already answered. The most plausible continuation is a generic list, because a generic list is what generically follows that sentence. The model is not being lazy. It is being accurate about a vague question."}]},{"t":"p","runs":[{"text":"The second prompt cannot be continued generically. There is one CV, one advert, one awkward gap to cover."}]},{"t":"p","runs":[{"text":"Average in, average out. When someone says a model gave them a bland answer, they have usually asked a question whose honest answer is bland."}]},{"t":"h2","id":"what-actually-did-the-work","runs":[{"text":"What actually did the work"}]},{"t":"p","runs":[{"text":"Look at what the second prompt added:"}]},{"t":"ul","items":[[{"text":"Who and where.","b":true},{"text":" Junior, Lagos, banking. What a CV should look like changes by market."}],[{"text":"The target.","b":true},{"text":" The advert’s own three words, quoted."}],[{"text":"The raw material.","b":true},{"text":" The CV itself, pasted, not described."}],[{"text":"The hard part.","b":true},{"text":" No paid experience. This is the difficulty of the task. Hiding it would have produced advice for someone else."}],[{"text":"A finished state.","b":true},{"text":" Top third rewritten, one page, SQL visible in ten seconds."}],[{"text":"A boundary.","b":true},{"text":" Do not invent experience."}]]},{"t":"p","runs":[{"text":"None of that is a trick. It is the briefing you would give a friend who offered to help over lunch."}]},{"t":"h2","id":"the-incantations-are-mostly-not-the-thing","runs":[{"text":"The incantations are mostly not the thing"}]},{"t":"p","runs":[{"text":"A lot of prompting advice is about magic words. Tell it that it is a world-class expert. Offer it a tip. Tell it to take a deep breath. Threaten it politely."}]},{"t":"p","runs":[{"text":"Some of those had measurable effects on the weaker models of 2023. Most have shrunk or disappeared as models improved, and several were never more than folklore repeated between blog posts. You will still find them recommended confidently today."}]},{"t":"p","runs":[{"text":"The honest version: the reliable levers are what you are trying to do, the material you are working from, the constraints that make it hard, and the shape you want back. Everything else is decoration."}]},{"t":"p","runs":[{"text":"That is good news. There is no secret language to learn. There is a habit of under-describing your own problem, and you can drop it."}]},{"t":"h2","id":"a-test-that-takes-one-minute","runs":[{"text":"A test that takes one minute"}]},{"t":"p","runs":[{"text":"Before you send a prompt, read it as if a competent stranger had sent it to you, with no memory of your week. Ask two questions:"}]},{"t":"ol","start":1,"items":[[{"text":"Could I start this task from this text alone?"}],[{"text":"What would I have to ask you first?"}]]},{"t":"p","runs":[{"text":"Every question you would have to ask is a missing line. Add those lines and send it again."}]},{"t":"p","runs":[{"text":"That is the whole method. The rest of this course is about doing each part of it well."}]}],"check":{"question":[{"text":"Someone asks a model to write an email to their landlord about a broken water heater and gets something generic. They try again with a longer, more emphatic prompt (“this is really important to me, please do an excellent job, you are a brilliant writer”) and get almost the same email. What is the best explanation?"}],"options":[{"text":[{"text":"The prompt is still too short. Models respond to length, and it needs to be several paragraphs before the quality improves."}],"correct":false,"tempting":[{"text":"Longer prompts genuinely do tend to work better, so length gets mistaken for the active ingredient instead of the detail that usually travels with it."}]},{"text":[{"text":"The second attempt added emphasis but no new information: nothing about the heater, the flat, how long it has been broken, or what the landlord has already said."}],"correct":true,"tempting":null},{"text":[{"text":"Wording never affects the output, so no rewrite of the prompt could have changed anything."}],"correct":false,"tempting":[{"text":"Once you learn that flattery and pleading do nothing, it is easy to overcorrect into thinking phrasing is irrelevant, when phrasing that carries information changes the answer entirely."}]},{"text":[{"text":"The model already produced its best possible answer to that question, so the next step is a different tool or a stronger model."}],"correct":false,"tempting":[{"text":"When effort does not help, blaming the model is the natural move, but between the two attempts the actual question never changed."}]}],"order":[2,0,3,1],"why":null},"prev":null,"next":{"slug":"what-the-model-sees","title":"What the model actually receives","minutes":10}}}