Shape the answer. These words decide what form the information arrives in — and form decides whether you'll use it.
Information has a shape before it has a meaning. The same twelve facts about streaming services are helpful in a table, tedious in prose, and useless in a poem — and the machine will happily deliver any of the three, because you never told it which you needed. Most people never think to ask for a shape at all. They accept whatever container the answer arrives in, then struggle to pour it into their actual lives: the plan that isn't in order, the summary whose dangerous clause is hiding in sentence four, the big job that won't fit in a weekend because it has never been taken apart.
Structuring words are container words. They don't change what the machine knows; they change the form it arrives in — sequence, outline, table, layers, chunks — and form is not decoration. Form is the difference between knowledge you have and knowledge you can use at 7 a.m. with coffee going cold. Ask for the shape your life can hold.
What it makes the AI do. It puts steps in true order, with dependencies respected — what must happen before what — instead of a pile of unordered advice.
Weak prompt: "How do I apply for a passport?"
Same prompt, with the word: "Sequence the steps for a first-time U.S. passport application in exact order, note which steps can't start until an earlier one finishes, and give the realistic wait at each stage."
Why it worked. Most processes fail at the ordering, not the steps. A sequence with dependencies is a plan; a list of steps is a rumor of one.
Another example.
Weak prompt: "How do I switch to a new phone?"
Same prompt, with the word: "Sequence switching phones so nothing is lost: what to back up first, what must be finished before the old card comes out, and what to check before the old phone gets wiped."
Why it worked. The danger in a phone swap is not the steps — it is doing them in the wrong order. The sequence exposed the trap doors before he fell through one.
One more.
Weak prompt: "How do I bake sourdough bread?"
Same prompt, with the word: "Sequence sourdough for a beginner: each step in order, with the wait between steps, which steps can flex, and the one step where timing actually ruins the loaf."
Why it worked. Baking is choreography. The sequence, and its single unforgiving moment, were the recipe.
Watch for. A sequence is only as good as its dependencies, and machines occasionally order by habit instead of logic — forms before the appointment that issues them. Read a sequence backward once, asking of each step: what before me makes me possible? The gaps show themselves.
Placed well: "Sequence the passport application in exact order, and mark which steps can't start until another finishes."
What it makes the AI do. It gives you the skeleton before the flesh, so you can fix the shape of a long answer while fixing is still cheap.
Weak prompt: "Write an essay about the history of lighthouses."
Same prompt, with the word: "Outline an essay on the history of lighthouses first — headings and one line under each — and stop there so I can adjust the shape before you write it."
Why it worked. A finished essay with the wrong structure gets thrown away whole. An outline gets rearranged in a minute, and everything written afterward inherits the fix.
Another example.
Weak prompt: "Write a letter to the airline about my lost bag."
Same prompt, with the word: "Outline the letter first: a heading for what happened, one for what I want, one for my deadline — a single line under each — and stop so I can approve the shape."
Why it worked. The outline took a minute to rearrange. The finished letter would have taken ten — and might have buried the deadline where no agent would ever see it.
One more.
Weak prompt: "Help me plan my mother's 80th birthday party."
Same prompt, with the word: "Outline the party plan first: guests, food, program, and budget — one line each — so we can argue about the shape before spending anything."
Why it worked. The outline revealed the real argument early: it was never about cake. It was about who gives the toast.
Watch for. The failure here is falling in love with the skeleton. An outline approved too quickly locks a weak shape in place, and everything written later inherits it. Argue with the outline while arguing is cheap — that is the point of asking for one.
Placed well: "Outline the letter first — what happened, what I want, my deadline — and stop for my approval."
What it makes the AI do. It turns prose into a table, aligning the same facts about different things into rows you can scan in seconds.
Weak prompt: "What are the differences between the main streaming services?"
Same prompt, with the word: "Tabulate the main streaming services: columns for monthly price, number of profiles, offline downloads, and whether they show ads, one service per row."
Why it worked. Prose hides mismatches — each service described in its own flattering paragraph. A table makes the missing cells and the odd numbers impossible to miss.
Another example.
Weak prompt: "Which phone plan should I get?"
Same prompt, with the word: "Tabulate these three phone plans: monthly cost, included data, hotspot allowance, and what happens when I go over — one plan per row, same facts in every column."
Why it worked. In prose, each plan sounded reasonable. In rows, the cheap plan's expensive overage stood in plain sight, lined up against its rivals.
One more.
Weak prompt: "Compare these laptops for me."
Same prompt, with the word: "Tabulate these three laptops: price, weight, battery hours, screen size, and warranty length — nothing else, no paragraphs."
Why it worked. Stripping the paragraphs stripped the salesmanship. The table's lightest row won on the facts.
Watch for. Tables flatter messy information: ragged, arguable facts march in tidy columns and start looking equally solid. Watch the empty cells and the footnotes that didn't fit — a table can hide a "nobody knows" more elegantly than any paragraph ever could.
Placed well: "Tabulate these three plans: monthly cost, data, hotspot, and what happens when I go over."
What it makes the AI do. It compresses the long into the short while keeping the claims that matter — the executive function every busy reader actually needs.
Weak prompt: "Here's a 12-page HOA document. What's in it?"
Same prompt, with the word: "Summarize this HOA document in five sentences, then list separately anything in it that costs me money, restricts what I can do, or has a deadline."
Why it worked. A summary alone can smooth over the one clause that bites. Asking for the costly parts as their own list means compression can't quietly bury them.
Another example.
Weak prompt: "Read this lease and tell me about it."
Same prompt, with the word: "Summarize this lease in six sentences — then list separately what I pay beyond rent, what I'm not allowed to do, and every date that matters."
Why it worked. Summaries smooth; leases bite in the details. Naming the costly parts as their own lists kept the deposit rules and the notice deadline above water.
One more.
Weak prompt: "What's in this employee handbook?"
Same prompt, with the word: "Summarize this handbook in five sentences, then list separately: how time off actually accrues, what can get me fired, and anything with a deadline in my first 90 days."
Why it worked. Handbooks are written to be unread. The separate lists are the parts written about him.
Watch for. Every summary is an act of triage performed by someone else's judgment — in this case, the machine's guess about what matters to you. If a document can bite (money, deadlines, prohibitions), never let it travel as summary alone; demand the biting parts listed separately.
Placed well: "Summarize this lease in six sentences, then list separately everything that costs me money or has a date."
What it makes the AI do. The stricter cousin of summarize: it squeezes an answer to a hard size while demanding nothing essential be lost — a test of what the machine thinks is essential.
Weak prompt: "Shorten this cover letter."
Same prompt, with the word: "Compress this cover letter to exactly half its length. Keep every specific achievement and number; cut compliments, filler, and anything the resume already says."
Why it worked. "Shorten" gets random deletions. Naming what must survive and what may die turns compression into editing with a policy.
Another example.
Weak prompt: "Make this message shorter."
Same prompt, with the word: "Compress this voicemail script to twenty seconds spoken: keep the callback number, the reason for calling, and the deadline; cut the apology and the backstory."
Why it worked. Naming what must survive turned a hatchet into a scalpel — and the message that remained still sounded like a person.
One more.
Weak prompt: "Shorten this product description."
Same prompt, with the word: "Compress this description to 40 words: keep the material, the size, and the one unusual feature; cut every adjective that isn't information."
Why it worked. The adjectives were the fat; the facts were the flavor. Forty words held all three facts comfortably.
Watch for. Compression kept the facts and cut the manners — a message can be half as long and twice as cold, because warmth is usually stored in the words labeled filler. Tell the machine which quality must survive the squeeze: a number, a name, a tone.
Placed well: "Compress this letter to half its length: keep every number and achievement, cut compliments and filler."
What it makes the AI do. The reverse gear: it takes the one terse point you care about and grows it — more detail, more reasoning, more examples — exactly where you point.
Weak prompt: "Give me tips for negotiating a raise. Tell me more about timing."
Same prompt, with the word: "Expand only the timing point: when in the year, when in the week, and when in the conversation a raise request lands best, and the reasoning behind each."
Why it worked. "Tell me more" expands everything a little. "Expand only this" expands one thing a lot — depth where you asked for it, silence where you didn't.
Another example.
Weak prompt: "Give me steak tips. Tell me more about resting the meat."
Same prompt, with the word: "Expand only the resting step: why it matters, how long for a thick steak, what happens if I skip it, and how to keep the meat warm meanwhile."
Why it worked. One step of the recipe got a full lesson while the rest stayed usefully brief — depth delivered exactly where the curiosity was.
One more.
Weak prompt: "Give me a history of the railroad. More on the strikes."
Same prompt, with the word: "Expand only the railroad strikes: what the workers wanted, what happened at the worst one, and how it changed the law. The rest can stay a sketch."
Why it worked. History is a continent; he got one state, mapped properly, instead of a postcard of the whole.
Watch for. Expansion inflates. Point "expand" at a minor step and the machine, eager to fill the space, promotes trivia to doctrine and pads the rest with weather. Expand one step, and cap it — "three paragraphs, no more" — so depth doesn't become sprawl.
Placed well: "Expand only the resting step: why it matters, how long, and what happens if I skip it."
What it makes the AI do. It breaks a big, foggy task into named parts small enough to start, exposing the hidden sub-jobs every big job conceals.
Weak prompt: "How do I start a podcast?"
Same prompt, with the word: "Decompose 'starting a podcast' into its separate jobs — format, equipment, recording, editing, hosting, artwork, first three episode topics — and give each job a realistic first step and cost."
Why it worked. Big goals intimidate because they're secretly six goals wearing a trench coat. Decomposition separates them, and small jobs get started.
Another example.
Weak prompt: "How do I move house without going crazy?"
Same prompt, with the word: "Decompose the move into its separate jobs — declutter, pack, utilities, transport, cleaning, and a first-night kit — with a start date and a cheap first step for each."
Why it worked. "The move" is dread wearing one word. Six named jobs each fit on a calendar square, and calendar squares get done.
One more.
Weak prompt: "I need to get my finances in order."
Same prompt, with the word: "Decompose 'finances in order' into separate jobs — bills, debts, savings, insurance, papers — and give each one a first step that takes under an hour."
Why it worked. "Finances" is weather; an hour-long first step is a door. He walked through the bills door the same evening.
Watch for. Taken apart, a job can look finished when it has only been described. Decomposition's trap is mistaking the parts list for progress — twelve neat sub-jobs, none started. End every decomposition by choosing the one part that begins this week, with a date on it.
Placed well: "Decompose the move into named jobs, each with a start date and a first step under an hour."
What it makes the AI do. After something is taken apart, it rebuilds it in a better arrangement — new order, new grouping, dead parts removed.
Weak prompt: "My resume feels messy. Fix it."
Same prompt, with the word: "Recompose my resume: regroup the experience under three skill headings instead of by date, drop anything older than fifteen years, and lead each section with its strongest result."
Why it worked. Editing polishes what's there. Recomposing questions the arrangement itself — and most messy documents are good material in a bad order.
Another example.
Weak prompt: "My week feels chaotic. Help."
Same prompt, with the word: "Recompose my weekly schedule: batch the errands into one block, move focused work to mornings, protect one evening completely, and show me the new week side by side with the old one."
Why it worked. He did not need more hours; he needed the same hours in a better order. The rearrangement, not the effort, was the fix.
One more.
Weak prompt: "My desk is a disaster area."
Same prompt, with the word: "Recompose my desk by function instead of habit: a work zone, a paper-in zone, a paper-out zone, and one small zone for things with no home yet. Describe the new layout."
Why it worked. The "no home yet" zone was the honest invention — clutter is usually just items awaiting a category.
Watch for. Recomposing rearranges; it does not repair. A chaotic week rebuilt in a better order is still a chaotic week if it holds nine hours of work in six hours of space. After the new arrangement, ask the rude question: what, in total, did we remove?
Placed well: "Recompose my week: errands batched, deep work in the mornings, one evening protected completely."
What it makes the AI do. It divides a long effort into phases with gates between them — each stage finished, checked, and only then the next begun.
Weak prompt: "How do I train for a 5K run?"
Same prompt, with the word: "Stage a beginner 5K plan into three stages — walking and easy jogging, then steady running, then distance — and tell me what I should be able to do before moving to each next stage."
Why it worked. A single plan hides its own beginning. Stages with gates made progress measurable — and quitting harder, because there was always a next small door.
Another example.
Weak prompt: "Help me write a book."
Same prompt, with the word: "Stage writing a short book: outline, rough draft, honest critique, revision, polish. For each stage, tell me what 'done' looks like before I'm allowed to move on."
Why it worked. The gates stopped the classic failure — polishing chapter one for a month while the book stayed unwritten.
One more.
Weak prompt: "How do I lose 20 pounds?"
Same prompt, with the word: "Stage the 20 pounds: stage one is habits only, no scale goals; stage two adds the kitchen; stage three adds activity. Give each stage a graduation test that isn't a number."
Why it worked. The gates measured capability instead of weight — and capabilities, unlike Mondays, mostly move in one direction.
Watch for. Stages can become a waiting room: stage one polished forever, its graduation test always nearly passed. The gate exists to be walked through, not admired. If a stage has lasted twice as long as planned, the honest move is shrinking it, not decorating it.
Placed well: "Stage this book: outline, draft, critique, revision — and tell me what 'done' looks like at each gate."
What it makes the AI do. It cuts big material into sittings sized to a real attention span, each with an ending point instead of a bookmark and a prayer.
Weak prompt: "This training manual is 200 pages. Help."
Same prompt, with the word: "Chunk this manual into six sittings of about 30 minutes each. For each chunk, give me its main ideas in advance and two questions to answer when I finish it."
Why it worked. Two hundred pages is a wall; six sittings are a path. The questions at each border turned reading into arriving.
Another example.
Weak prompt: "My inbox has 900 unread emails."
Same prompt, with the word: "Chunk the inbox cleanup: sort first into four piles — answer, file, delete, decide — and give me a 15-minute routine per pile that I can repeat daily."
Why it worked. The pile was never 900 problems; it was four small problems, repeated. Chunking made the first day finishable.
One more.
Weak prompt: "Teach me photography."
Same prompt, with the word: "Chunk photography into weekend lessons — one idea and one outing per chunk, ending each with a single photo assignment I can finish in an hour."
Why it worked. A hobby became a course with doors in it — and hour-sized assignments fit real Saturdays.
Watch for. Chunks fail at their borders. Cut a manual at page counts alone and one sitting ends mid-argument while the next opens mid-sentence, and both feel broken. Chunk by ideas, not arithmetic — each sitting should end where a thought completes, with a question answered.
Placed well: "Chunk this manual into six sittings by topic, each ending with two questions I should be able to answer."
What it makes the AI do. It stacks the same explanation at several depths — a sentence, a paragraph, a page — so you can stop at the layer you need and descend only where it matters.
Weak prompt: "Explain how the internet works."
Same prompt, with the word: "Layer the explanation of how the internet works: first in one sentence, then in one paragraph, then in a full page for where I want more — and label each layer so I can stop when I've got enough."
Why it worked. Most explanations pick one depth and dare you to be satisfied. Layers respect the reader's real need, which is usually "enough to follow the news, plus one deep dive."
Another example.
Weak prompt: "Explain my lab results."
Same prompt, with the word: "Layer this lab report: layer one, am I okay; layer two, what each number means; layer three, what questions to bring to my doctor."
Why it worked. The layers matched the emotional order of the questions — reassurance first, understanding second, action third.
One more.
Weak prompt: "Explain quantum computing."
Same prompt, with the word: "Layer it: one sentence for a neighbor, one paragraph for a curious teenager, one page for me — and be honest at each layer about what the layer is leaving out."
Why it worked. The honesty labels were the upgrade: he knew not just the explanation, but its thickness.
Watch for. Layers tempt skimming with a clean conscience: layer one read, layers two and three forever pending, and the reader certain they "basically understand." Sometimes that's fine. On lab results and contracts, decide in advance which layer you're obliged to reach — then reach it.
Placed well: "Layer this report: am I okay, what the numbers mean, and what to ask my doctor — labeled."
What it makes the AI do. It shows how the pieces connect — who affects whom, what feeds what — instead of describing each piece alone in its own paragraph.
Weak prompt: "I'm lost in this novel. Who are all these people?"
Same prompt, with the word: "Map the relationships in this novel: group the characters by family and alliance, and for each one give me their connection to the main character in a single line."
Why it worked. A cast list is alphabetical; a map is meaningful. Grouped by connection, the crowd became a structure he could hold in his head.
Another example.
Weak prompt: "Why does the office coffee fund keep running out?"
Same prompt, with the word: "Map the coffee fund as a system: where money enters, where it leaves, who decides purchases, and where the leak most likely sits."
Why it worked. Small mysteries are usually systems with one loose joint. The map pointed at the joint instead of at suspects.
One more.
Weak prompt: "Explain why our town's Main Street is struggling."
Same prompt, with the word: "Map Main Street as a system: the anchor stores, the parking, the highway bypass, the rents — and show which pieces pull on which."
Why it worked. Arguments about single causes ended. The map showed the bypass and the rents holding hands, which is where the real conversation starts.
---
Watch for. A map shows connections so confidently that invented ones slip in wearing the same lines as real ones — two departments "linked" because both were mentioned near each other. Ask the machine to mark which connections are documented and which are its reasonable guesswork.
Placed well: "Map the coffee fund as a system: where money enters, where it leaves, who decides, and where it leaks."