Have you noticed every business idea is either saturated or needs 10 years you don’t have?

Start an AI automation agency, says the guy whose entire field is two years old and who somehow has a decade in it.

Flip peptides. Trade crypto.

Every one of them shows up as a course, sold by someone whose only proven income is the course.

You quietly wonder whether they’ve ever done it.

There is something you could sell. You just can’t see it, and there’s a good reason for that.

Why you can’t see what you’re good at

Try this on somebody. Pick a song everyone knows. Mary Had a Little Lamb. Tap the rhythm on the table with your finger and ask them to name it.

A Stanford student ran that as an experiment in 1990. She took a group of people and split them into pairs. One person taps, the other guesses. Between them, the group tapped out 120 songs.

Before each turn, she asked the tapper how likely their partner was to name it. They said about half the time.

The group got 3 of the 120 right.

The tapper hears the whole song in their head while they tap. The other person hears knocking on a table.

That gap has a name. It’s called the curse of knowledge. Once you know something, you can’t go back and imagine not knowing it, so you assume everyone else knows it too.

That’s why what you know feels worth nothing. In your head it’s obvious. Everybody knows this.

Truth is, nobody knows this.

So the prompt never asks what you’re good at. It asks questions about what you do.

What this prompt actually does

Put the prompt in, and it starts to interview you. You go through 11 questions over 3 rounds. After each round, it stops and waits for you to answer.

Then it goes into its memory and looks at older chats. It looks for something you asked about.

What you get back is 3 business ideas. Each one is named and based on your answers, and 1 of the 3 comes from suggestions from the memory. With each idea you get 5 topics you already know enough to write about today.

Why topics? Because writing one is how you find out whether you really know the subject or just like it.

Before the ideas, you get 2 paragraphs about yourself based on what you told it. The last sentence of that section matters because it points out something you think is normal, but other people would pay to learn.

If that part is wrong, the 3 ideas under it are wrong too, so run it again and answer differently.

Skip this if you already know what you’re selling and who you’re selling it to. This is for the part before that.

The business idea prompt

I wish I could say this gives you a business to run. It won’t. It gives you a subject, your ideal customer, and 5 topics to start with. You still have to write them.

Turn dictation on and talk your answers out. Give it as much context as you can, and don’t edit yourself.

Help me find what I could build something around. I do not have an idea yet, or I have a vague one.

Ask me these one round at a time. Wait for my answers before the next round. Do not comment on my answers as we go. Tell me before Round 1: answer fast and do not edit yourself, and if you notice yourself steering away from a subject, name it anyway.

Round 1
1. What do friends or family text you about, the thing they ask you specifically?
2. What do you feel you are genuinely an expert in, even if nobody pays you for it?
3. What do you catch yourself excited to talk about, past the point where the other person is still interested?
4. What have you explained more than three times in your life?

Round 2
5. Name the thing you spend the most time inside: a hobby, a craft, a scene, a subject, a community. Be specific. Not "crafts" but which craft. Not "games" but which game and which part of it. It can be anything at all: knitting, birdwatching, pickleball, metal detecting, sourdough, restoring furniture, rock collecting, competitive spreadsheet building, keeping bees. Weird is fine. Weird is better.
6. What is the last thing you researched for more than an hour before buying?
7. What did you last tell someone they were doing wrong?
8. What did you get obsessed with for a few months, even if you dropped it?

Round 3
9. What did you figure out the hard way that you wish someone had told you?
10. What do you know that would take a beginner six months to work out alone?
11. Who do you want reading you: people in your own country, or somewhere else? Name the country.

After Round 3, before you write anything else:

Step one. List for yourself the three subjects I bring you most often, the projects and topics that dominate our history together. Those three are banned. You may not build an idea on any of them and you may not mention them anywhere in your answer. I already know about those. They tell me nothing.

Step two. Use your memory tool and search what you remember about me. Do this as an actual lookup, not from what I typed in this conversation. Go old and prefer the one-off: the small thing I brought you once and never came back to, the question from a year ago, the problem I solved and forgot. List three things you found this way, numbered, then continue. If you have no stored memory of me, say so in one line and work from my answers alone.

Step three. Treat where I have lived, moved between, or had to navigate as expertise. A city I know from the inside, a country I moved to, a system I had to learn because I had no choice, a place I can get around and outsiders cannot. I will not have listed this as a skill, because it is just my life. Count it anyway.

Then write two things, in this order.

FIRST, a section headed exactly: YOU ARE AN EXPERT IN THIS AND YOU THINK IT IS NORMAL

Five or six sentences, split into two short paragraphs. Every sentence under 25 words. Second person, present tense, declarative. No hedging, no "you may sometimes," no "it seems."

- Open on a tension: two opposite things that are both true of me. "You will spend four hours on X and cannot make yourself spend ten minutes on Y."
- Describe what I do, not what I am. Banned words: curious, analytical, passionate, driven, pragmatic, detail-oriented, creative, thoughtful, resourceful, and every other word that names a personality type. If it would go on a CV, cut it and write the behaviour instead: the situation, and what I do in it.
- Use at least two specific details from my own answers, close enough that I know exactly which answer you took them from.
- Name one thing I do that I have probably never said out loud, and that I will recognise the moment I read it.
- Include exactly one criticism. State it plainly, then say what it costs me. Do not end the section on it.
- End on the point. Name the one thing I treat as ordinary that other people would pay to have explained to them. Say it flatly, as a fact about the world, not as encouragement. This is the last sentence.
- Do not explain why people cannot see their own expertise. Do not name any psychological effect. Do not mention studies. Just do it to me.
- Tone: someone who knows me well and likes me. Not a performance review, not a diagnosis, not a warning.

Then a line break, then three ideas, ranked, in exactly this format. Obey the word limits. No preamble, no reasoning, no extra sections.

IDEA 1: [name the newsletter, three or four words]
[One sentence: what lands in the inbox each week, and for whom.]

Why it is you: [Two sentences, max 40 words. Tie it back to what you wrote above.]

What you know that most people do not: [One sentence, max 20 words.]

Who would buy this: [One specific person by situation, not a demographic, max 25 words. Use the country I named. If it is not the US or UK, do not hand me a US market and do not assume my readers have US money or US problems.]

Why it is narrow enough:
- Wide version: [the crowded category and why it is crowded. Max 20 words.]
- Yours leaves out: [what you cut to make it different. Max 20 words.]

First five issues:
- [Max 12 words each. Something only someone who knows this could write. Every title must name a specific thing: a product, a place, a technique, a moment, a number. No title may be about me, my habits, or what I learned. If you could swap the subject out and the title still worked, it is too vague.]

Repeat for IDEA 2 and IDEA 3.

Rules:
- IDEA 3 must come from your memory of me, not from my answers, and it cannot touch the three banned subjects. In its "Why it is you" line, name the specific past conversation it came from, close enough that I remember it. If you cannot name one, say "no memory to draw on" and give me a fourth idea from my answers instead.
- The three ideas must each draw on a different answer of mine. If two trace back to the same answer, throw one out and find another. Before you finish, check idea 1 against idea 3: if the same reader would subscribe to both, they are one idea and you have only given me two.
- Make the three genuinely different, not three angles on one idea.
- If what I know is something I am still learning, do not tell me to teach it. Find what I can do now: what I see as I go that the people ahead have forgotten and the people behind have not reached.
- Every idea must be something I could put my name on and send to a stranger. If my answers point somewhere I could not, do not lecture me. Point the underlying skill somewhere I could.
- Nothing that needs money, credentials, or an audience I do not have.
- Do not tell me to start a YouTube channel or make videos.
- Do not mention business models, monetization, marketing, or platforms.

End by asking which of the three landed, and why.

Why it works: The 11 questions never ask you to rate yourself. They ask you to remember specifics.

You can answer those without thinking too much.

The instruction doing the heavy lifting: It bans your 3 biggest subjects before it writes a word. We don’t want your AI to tell you what you already know. Banning them pushes it toward something you asked about a couple of times, a subject it thinks you might be good at. It’s the same habit as writing AI prompts that get you past the obvious answer.

One gap: Sometimes it comes back with nothing from memory. It happened to me 3 times in a row with Claude, and it gave me 3 ideas out of my answers instead. The issue was that my system prompt was getting in the way. You’ll know because all 3 ideas came out of the answers you just gave. One of them should come from memory.

Simple fix: Try it in a plain chat with no custom instructions and no project settings loaded, and check that memory is on before you start.

What got cut: The first version let the model describe you, and it was pure glaze. Words like curious, analytical, driven. The same words almost every try, and they could have been about anybody. Those words are banned, so it has to tell you what you do instead of what you are.

What to do with the 3 business ideas

Read the paragraphs above the ideas first. That last sentence is the one to understand, and decide whether it’s accurate. The 3 ideas are 3 different ways to sell it, and even if all 3 are wrong, that sentence can still be right.

Pick the one you reacted to, not the one that sounds best. A strong feeling about what it wrote means you know something the model doesn’t.

Choose any idea it gave you. Then take the 5 titles under it and try to write one of them without looking anything up.

If you can, that’s your business idea.

If you can’t write about any topic, it’s a category you like, not a subject you know, so go to another idea. Don’t use excitement as the gauge either. People want to hear what you know best, and it might not be glamorous, but it could be a business.

You don’t have to be sure. This is a step, not a career choice.

How to find out if the idea is real

Write it as a newsletter, because that’s the only version of this you can start today with nothing. No product, no company, no money up front. Most free accounts include a website, so the topics have somewhere to live.

This isn’t a course about something you’ve never done. It’s the subject people already come to you about.

Not a writer? Talk out loud to the same AI you used to get your business ideas. Tell it your objective and have it clean up your words without changing them. Then cut anything you wouldn’t say out loud.

Send the first few issues to 10 people you know, then ask each one where they stopped reading. That tells you where you lost them. Work on that.

By issue 4, you know the subject better than when you started, because writing about it makes you find the parts you weren’t clear on. That’s your idea getting tested by something other than your own opinion.

Steal this: open a beehiiv account, take the first title from idea 1, and send that issue to 10 people this week. It’s free to start, and that’s where I write mine. It’s an affiliate link.