# The 50 Prompts
### The ones a one-person business actually uses in a week. Build Leaner.

**Read this first.** These are written to be pasted *after* the business context doc is already set up (see `business-context-doc.md` in this pack). The context doc tells the AI who you are, how you write, and when to shut up and admit it does not know. Every prompt below assumes that page is already loaded, which is why they are short. If you paste these into a cold chat with no context doc, they still work, they are just more generic.

**How to use this file.** Do not read it top to bottom. Find the job you are doing right now, copy the prompt, replace anything in `[SQUARE BRACKETS]`, send it.

**Jump to:**
1. [Writing that doesn't sound like AI](#1-writing-that-doesnt-sound-like-ai) (1–8)
2. [Inbox and clients](#2-inbox-and-clients) (9–16)
3. [Selling: proposals, pricing, follow-up](#3-selling-proposals-pricing-follow-up) (17–24)
4. [Building and shipping](#4-building-and-shipping) (25–31)
5. [Research and fact-checking](#5-research-and-fact-checking) (32–37)
6. [Numbers and decisions](#6-numbers-and-decisions) (38–43)
7. [The honesty prompts](#7-the-honesty-prompts) (44–50)

---

## 1. Writing that doesn't sound like AI

**1. Notes into posts**
*Monday. You have a week of scrappy notes and nothing published.*
```
Here are my raw notes from this week:
[PASTE NOTES]

Write 5 standalone posts from these. Follow my voice rules exactly. First line of
each has to work on its own with no context. No hashtags. Under 280 characters.
Do not invent anything that isn't in the notes.
```

**2. The de-AI pass**
*You wrote it yourself and it still reads like a press release.*
```
Here's a draft I wrote:
[PASTE]

Find every sentence that sounds like AI or corporate filler and rewrite only those.
Leave the rest alone. For each change, show the before and after on one line so I can
see what you did. Do not improve anything I didn't flag as broken.
```

**3. Cut it in half**
*It's good and it's twice as long as it should be.*
```
Cut this to [N] words without losing a single idea:
[PASTE]

Delete first, rewrite second. Tell me at the end which idea you were closest to
losing, so I can check you didn't.
```

**4. The first line**
*The rest of the post is fine. The opener is dead.*
```
Here's a post. The first line is weak:
[PASTE]

Give me 10 alternative first lines. Each one has to work with zero context, be under
15 words, and lead with a number or a named problem. No questions, no "imagine if,"
no "here's the thing."
```

**5. Same idea, five angles**
*You know the point. You don't know how to come at it.*
```
Core idea: [ONE SENTENCE]

Write this five ways: as a number, as a mistake I made, as a thing everyone gets
wrong, as a before and after, and as a single question. One short paragraph each.
Label them.
```

**6. Voice match from a sample**
*A new AI, a new chat, and it's writing like a stranger again.*
```
This is a paragraph I actually wrote:
[PASTE A REAL PARAGRAPH]

Study the sentence length, the rhythm, and the words I avoid. Now write [WHAT YOU
WANT] in that voice. Before you write, list the three rules you extracted from my
sample so I can tell you if you read it wrong.
```

**7. The newsletter draft**
*Same structure every week. Stop retyping the brief.*
```
Build notes for this week:
[PASTE]

Write this week's issue in my newsletter structure from the context document.
Give me 3 subject line options under 45 characters. Write it so a busy person reads
every line. Mark anything you had to assume with [assumed].
```

**8. Kill the tells**
*The one-line cleanup before you hit publish.*
```
Scan this for AI tells and list them with line references. Do not rewrite:
[PASTE]

Tells I care about: em dashes, "it's not X, it's Y" constructions, tricolon lists,
"delve," "leverage," "unlock," "game-changer," "in today's," fake enthusiasm,
and any sentence that could appear in anyone's blog post about anything.
```

---

## 2. Inbox and clients

**9. The follow-up that isn't needy**
*Eight days of silence and you need to send something.*
```
Situation:
[3 BULLETS: what they asked, what you sent, how long it's been]

Write the follow-up. Under 120 words. One clear ask. Warm and direct, zero corporate
filler. It must not sound like a template or a nudge sequence.
```

**10. Reply triage**
*Forty unread and ninety minutes.*
```
Here are the subject lines and first lines of my unread email:
[PASTE]

Sort into: reply today, reply this week, delete, and needs a decision from me.
For the last group, say in one line what the decision actually is.
```

**11. Say no without burning it**
*Wrong fit, good person.*
```
I need to decline this:
[PASTE THEIR REQUEST]

Reason I'm declining: [REAL REASON]

Write the reply. Say no clearly in the first two sentences. No fake scheduling
excuses. If there's an honest referral or alternative, offer it. Under 100 words.
```

**12. The hard update**
*Late, over scope, or broken, and they don't know yet.*
```
What happened: [FACTS]
What it means for them: [IMPACT]
What I'm doing about it: [PLAN]

Write the email. Lead with the fact, not the apology. One apology sentence maximum.
End with the specific next date. Do not use the word "unfortunately."
```

**13. Meeting notes into a decision log**
*The call is over and nothing is written down.*
```
Raw notes from a client call:
[PASTE]

Extract: decisions made, open questions, who owes what by when, and anything I
agreed to that wasn't in scope. Flag that last group clearly. If something is
ambiguous in my notes, say so instead of guessing.
```

**14. Translate the jargon**
*They're technical. Your client isn't.*
```
Rewrite this for someone who is smart but not technical:
[PASTE]

Keep every fact. Lose every acronym or define it in four words. If a 12-year-old
couldn't follow a sentence, rewrite it.
```

**15. The kickoff message**
*New client, day one, set the frame.*
```
Project: [ONE LINE]
Scope: [BULLETS]
Timeline: [DATES]

Write the kickoff email. Confirm what's in scope AND name one thing that's out of
scope, plainly, so it can't be assumed later. Give them exactly one thing to do next.
```

**16. Reply in my voice, short**
*The generic one you'll use most.*
```
They wrote:
[PASTE]

I want to say: [YOUR ROUGH POINT, MESSY IS FINE]

Write the reply in my voice. Under 80 words. Do not add pleasantries I didn't ask
for. Do not thank them twice.
```

---

## 3. Selling: proposals, pricing, follow-up

**17. Brief into proposal**
*The most expensive hour of your week, compressed.*
```
Client situation: [BULLETS]
What they think they want: [THEIR WORDS]
What I think they need: [YOUR READ]
Price: [NUMBER]

Write a one-page proposal: problem, plan, price, next step. Their problem in their
words at the top. No methodology section. No "about us." One page means one page.
```

**18. Price it out loud**
*You always undercharge and you know it.*
```
Job: [DESCRIBE]
Hours I think it takes: [N]
What I was going to charge: [NUMBER]

Argue with my number. Tell me what I'm not counting: revisions, meetings, the thing
that always goes wrong, the cost of saying yes to this instead of something else.
Give me a floor I shouldn't go below and why.
```

**19. Objection pre-mortem**
*Before you send it, not after.*
```
Here's my proposal:
[PASTE]

You are the client. List every objection you'd raise, ranked by how likely it is to
kill the deal. For the top three, tell me whether the fix is a change to the proposal
or a change to the price. Be harsh.
```

**20. The scope-creep reply**
*"While you're in there, could you also..."*
```
Original scope: [BULLETS]
What they're now asking for: [PASTE]

Write the reply. Say yes to the work and no to doing it for free, in that order.
Give a number or a timeline for the extra. Do not sound annoyed. Under 100 words.
```

**21. Case study from a real job**
*You did good work and nobody knows.*
```
Project facts:
[WHAT THEY HAD BEFORE / WHAT YOU BUILT / WHAT CHANGED, WITH REAL NUMBERS]

Write a short case study: the situation, the specific problem, what I built, the
result. Use only the numbers I gave you. If a number would strengthen it and I
didn't give you one, write [need number] instead of inventing it.
```

**22. Discovery questions**
*So the call isn't you talking for 40 minutes.*
```
Prospect: [WHO THEY ARE, WHAT THEY DO]
What I sell: [ONE LINE]

Give me 8 discovery questions for the first call. Each one has to surface something
that changes the price or the plan. No questions whose answer I could have found on
their website. Order them so the uncomfortable ones come after the easy ones.
```

**23. The last-chance email**
*Two follow-ups in, still nothing.*
```
Context: [WHAT WAS QUOTED, WHEN, HOW MANY FOLLOW-UPS]

Write the final email. Give them a clean way to say no. Do not guilt, do not use
false deadlines, do not say "circling back" or "just bumping this." Under 70 words.
```

**24. Rewrite their brief**
*Their brief is a wish list, not a job.*
```
Client brief:
[PASTE]

Rewrite it as a real scope: what's actually being delivered, what's explicitly not,
and the three assumptions the whole thing rests on. Then list the questions I have
to ask before I can price it.
```

---

## 4. Building and shipping

**25. The feature brief**
*The prompt that gets working code instead of a demo.*
```
My stack: [STACK]
Here is the existing file:
[PASTE FILE]

Build: [THE FEATURE, ONE PARAGRAPH, INCLUDING WHAT IT LOOKS LIKE WHEN IT WORKS]

Give me complete working code for the changed files only. Match the existing code
style. Do not refactor anything I didn't ask about. List any new dependency and why
it's needed. If my request is ambiguous, ask before writing.
```

**26. Explain this error like I'm busy**
*Stack trace, no patience.*
```
Error:
[PASTE]

Relevant code:
[PASTE]

Tell me in this order: what broke, why, the smallest fix, and whether the smallest
fix is the right fix. If you're guessing about the cause, say you're guessing.
```

**27. The 90% wall**
*It demos fine and it isn't shippable.*
```
Here's what I've built:
[DESCRIBE OR PASTE]

It works when I test it. List everything that will break when a real user touches it:
empty states, bad input, slow network, two people at once, mobile, and whatever else
I haven't thought of. Rank by how likely it is to happen in the first week.
```

**28. Smallest shippable version**
*You're building the wrong amount of thing.*
```
What I want to build: [DESCRIBE]

Cut this down to the smallest version that a real person could actually use this
week. Tell me what to drop and what to keep. Do not give me phases or a roadmap.
Give me one version, and be aggressive about what gets cut.
```

**29. Review my code like you don't like me**
*Before it goes live.*
```
[PASTE CODE]

Review this. Lead with the worst problem. Cover correctness first, then security,
then readability. Do not compliment anything. If it's fine, say "no issues" and stop.
```

**30. Name the thing**
*Naming is 20 minutes you'll never get back.*
```
It does this: [ONE SENTENCE]
It's for: [WHO]

Give me 15 names. Short, sayable out loud, no invented compound words, no dropped
vowels, no "-ly" or "-ify." Mark any that are probably taken.
```

**31. Write the docs from the code**
*You'll forget how this works in three weeks.*
```
[PASTE CODE OR CONFIG]

Write the setup instructions someone else could follow. Assume they know nothing
about this project. Comment the parts a non-developer would trip on. If anything in
the code is unclear enough that you had to guess its purpose, list it separately.
```

---

## 5. Research and fact-checking

**32. Research with receipts**
*The one you'll use most, and the one most likely to lie to you.*
```
Research this for a decision: [QUESTION]

I need the top [N] findings, each with where you found it. Then one paragraph on
what this means for [YOUR SITUATION].

Follow the honesty rules: no invented sources, no invented numbers. Mark anything
you're not certain of with [uncertain]. If you can't find real sources, say so
instead of producing plausible ones.
```

**33. Check your own work**
*Run this on the answer it just gave you.*
```
Take your last answer. Pull out every factual claim, number, date, and source.
For each one, tell me: certain, likely, or guessed. Be honest about the guessed ones.
Do not re-argue the answer, just audit it.
```

**34. Find the counter-argument**
*You only found evidence for what you already thought.*
```
Here's what I've concluded: [YOUR CONCLUSION]

Make the strongest case against it. Use real evidence, not devil's advocacy. If the
best counter-argument is weak, say that too, and tell me why my conclusion holds.
```

**35. Compare properly**
*Two options, and every article about them is an affiliate page.*
```
Compare [A] and [B] for someone who [YOUR SITUATION].

Do not give me a feature table. Give me: the three ways they actually differ in
daily use, who each one is wrong for, and what it costs to switch later. Mark
anything you're not certain of with [uncertain].
```

**36. Summarize a long thing without losing the point**
*A 40-page PDF and 10 minutes.*
```
[PASTE OR ATTACH]

Summarize in three layers: one sentence, five bullets, and one page. Then tell me
the single thing in here that would change what I do, and the single thing that
looks important but isn't.
```

**37. What am I not asking?**
*Before you spend a week on the research.*
```
I'm trying to decide: [THE DECISION]
Here's what I've been researching: [WHAT YOU'VE LOOKED AT]

What questions should I be asking that I'm not? List them, ranked by how badly
getting them wrong would hurt.
```

---

## 6. Numbers and decisions

**38. Read my data**
*Real numbers, real decision.*
```
Here's my data as CSV:
[PASTE]

Question: [THE ACTUAL DECISION]

Show the calculation before the recommendation. If the data can't answer the
question, say so instead of guessing. Flag any row that looks wrong or incomplete.
```

**39. Where do the hours go?**
*The most useful hour you'll spend this quarter.*
```
Here's roughly how I spend a week:
[LIST TASKS AND HOURS]

Here's what actually moves my scoreboard: [YOUR 1-2 REAL METRICS]

Tell me which hours are producing the number and which aren't. Be specific about
what to stop. Do not suggest I "batch" or "time-block" anything.
```

**40. Price the decision, not the task**
*"Should I build it or buy it?"*
```
Option A: [DESCRIBE, WITH COST AND TIME]
Option B: [DESCRIBE, WITH COST AND TIME]
My hourly value: [NUMBER OR "unknown"]

Do the actual math on both, including the cost of my time and the cost of maintaining
it in a year. Then pick one. Do not say "it depends."
```

**41. Forecast honestly**
*You want a number, not a story.*
```
Here's the trend so far:
[PASTE NUMBERS BY PERIOD]

Give me a low, likely, and high case for [NEXT PERIOD], and state the assumption
behind each. Tell me how much of this is a real pattern and how much is too few
data points to say. Do not smooth over a small sample.
```

**42. The pre-mortem**
*Before you commit the month.*
```
Plan: [DESCRIBE]

It's [THREE MONTHS] from now and this failed. Write the three most likely reasons,
most likely first. For each, tell me the cheapest thing I could do this week to find
out early.
```

**43. What's the one number?**
*You're tracking eight things and steering by none.*
```
My business: [ONE LINE]
What I currently track: [LIST]
What I'm trying to achieve in the next 90 days: [ONE SENTENCE]

Pick the one number I should steer by, and say why the others are noise for now.
Then tell me what that number would have to hit for the 90 days to count as a win.
```

---

## 7. The honesty prompts

These are the ones people copy from the video. They work best with the honesty block from `business-context-doc.md` already loaded, but they work alone too.

**44. Tell me it's bad**
*The core one.*
```
[PASTE THE THING]

Assume this is weak and find out why. Lead with the single worst problem. Do not
open with anything positive. If your honest view is that it's good, say that in one
sentence and then tell me the one thing that would make it better.
```

**45. Kill my idea**
*Before you spend three weeks on it.*
```
I want to build: [THE IDEA]
I believe it will work because: [YOUR ASSUMPTION]

Give me the strongest reason not to build this. Then tell me which of my assumptions
is the one most likely to be wrong, and the cheapest test I could run this week to
check it. Do not soften the answer.
```

**46. Stop agreeing with me**
*Mid-conversation, when it's gone soft.*
```
You've agreed with my last several messages. That's not useful.

For the rest of this conversation: challenge my reasoning before you extend it. If I
say something wrong, say so first. If I'm right, say "agreed" in one word and move on.
No preamble, no validation, no restating my point back to me.
```

**47. Where are you guessing?**
*Run this on any answer that felt too smooth.*
```
Which parts of that answer are you confident about, and which are you filling in?
Mark each claim: verified, inferred, or guessed. Be specific about the guessed ones.
I'd rather have a gap than a confident wrong answer.
```

**48. Two sharp questions**
*Use it before any big task, not after.*
```
Before you do this: [THE TASK]

Ask me the two questions whose answers would most change what you produce. Only two.
Do not start the work until I answer.
```

**49. What would you do differently?**
*The end-of-project one.*
```
Here's what I did and how it turned out:
[WHAT YOU DID / WHAT HAPPENED / THE NUMBERS]

What would you have done differently, and at which specific point? Do not tell me
what I did well. If the outcome was mostly luck in either direction, say that.
```

**50. Argue with yourself**
*When the answer matters enough to spend two extra minutes.*
```
Take the answer you just gave me. Now write the strongest argument that it's wrong.
Then tell me which of the two you actually believe and why.
```

---

## The three habits that matter more than any prompt

1. **Set the context doc once, not per chat.** ChatGPT custom instructions, a Claude project, or a Gemini Gem. Every prompt in this file is short because that page is doing the heavy lifting.
2. **Take three claims out of every research answer and verify them yourself.** Every time. That habit is worth more than the subscription.
3. **If the answer sounds like it's flattering you, it probably is.** Prompt 44 or 46, then continue.

---

*Build Leaner. I build AI tools in public and give them away. What I built, what broke, what I learned, weekly at buildleaner.com.*