Try this right now. It takes 30 seconds and it will teach you more than the rest of this issue.
Open Claude and type this:
Write an email to my boss about the project delay.
Read what comes back. It is fine. It is also nothing. Formal, hedged, four paragraphs of throat-clearing, and it sounds like it was written by a stranger about a company that does not exist.
Now open a new chat and type this:
Write an email to my boss about the project delay. She is the CEO of a private equity portfolio company. She hates long emails. She wants bullets, the number, and a clear ask. She already knows the project is late. She wants to know what I am doing about it.
Read that one.
Same machine. Same question. Wildly different answer.
Nothing got smarter in those 30 seconds. You just told it what it could not possibly know on its own. That is context, and context is the entire ballgame.
WHY THIS MATTERS MORE THAN IT SOUNDS
Part two of the Business Body made the case. The model is the processor. It is the same processor your competitor is renting. Context is the only thing that makes your answer different from theirs.
Every good AI output you have ever gotten had context in it. Every generic one was missing it.
So here is the habit. Before you hit enter, ask yourself one question. What does this thing not know that I know?
Then type that part. That is it. That is the skill.
NOW HERE IS THE PROBLEM
Go look at the good prompt again. My boss is a CEO. She hates long emails. She wants bullets and a number.
You are going to type that again tomorrow. And next week. And the week after.
The same facts about your company. The same customers. The same standards. The same way you decide things. Over and over, for the rest of your life, into a box that forgets you the second you close it.
That is stupid, and it is the only reason a second brain exists.
Your second brain is your context, written down once. That is the whole idea. Not a system. Not a productivity religion. Just the stuff you are tired of retyping, saved in one place, so the machine already knows it.
WHAT ACTUALLY BELONGS IN IT
Now the part people get wrong. They build a second brain and stuff it with the wrong things, and then it does nothing for them.
There are three layers of intelligence.
Layer one is what everybody knows. Definitions. Best practices. Frameworks. The stuff in every business book. The machine read all of it before breakfast. Putting layer one in your file is like handing a chef a bag of salt and telling him you brought dinner.
Layer two is what your company knows. Your numbers. Your customers. Your contracts. Real information, and it helps. But your competitor has their own version. Layer two gets you to the table. It does not win the hand.
Layer three is what only you know. The call you made that everybody hated and you were right. The deal you walked away from twice. Why you fired your biggest customer. What you will never do again, no matter how good the math looks.
Layer three never got written down anywhere, because it lives in your head. It is the only layer the machine cannot get from somebody else. Go straight at it. Everything else is filler.
So the document has six sections.
Your points of view. What you believe about a topic that most people in it do not. Here is the test. If everybody nods along, it is not a point of view. It is generalist knowledge. A real one is the thing a smart competitor would fight you about at dinner.
Mine is that departments are a constructed fallacy. They were invented to work around the limits of the human brain, not because the work needs them. Plenty of people would argue with me. Great. That is what makes it worth writing down.
Your rules of thumb. The things you say in every meeting. The lines you repeat so often that your team finishes them for you. That is your operating system, and nobody ever wrote it down.
Your non-negotiables. What you will never do, no matter what the numbers say.
Your examples. Real ones. A board update that was good and one that was bad, and one line on why. A machine learns what you want faster from two real examples than from a page of adjectives.
The things only you know. Which customer is fragile right now. Which metric is vanity. Which project is already dead while everybody keeps pretending.
Your gaps. Where you are weak. Where you are guessing. Write those down too, so the machine quits pretending you have it all figured out.
STEP 1. RUN THE INTERVIEW.
Here is the problem. Nobody sits down and writes six essays about themselves. You will open a blank page, stare at it, feel stupid, and go answer email.
So do not write it. Get interviewed.
Open Claude or ChatGPT and start a normal chat. Copy everything in the box below. Paste it in. Then answer the questions.
Give it 45 minutes. If you get pulled away, tell it to summarize where you left off and pick it up later.
ROLE
You are interviewing me to build my second brain. This is a working document that you and other AI tools will read before tasks that need my judgment, so you can think like me instead of like the average of the internet.
WHAT I ACTUALLY WANT
My judgment. Not a nice summary of my judgment. The specific, opinionated, arguable way I run my business. If this document could have been written by any competent executive in my industry, it is worthless and we both wasted an hour.
HOW TO RUN THIS INTERVIEW
- Ask ONE question at a time. Wait for my answer. Never batch questions.
- Do not compliment me. Do not say "great answer." Do not repeat my answer back to me. Just go to the next question.
- Follow up before you move on. Assume my first answer is my shallow answer. It usually is.
- Demand specifics. If I give you a principle, ask for the time it actually happened. Ask who, what number, what it cost, what I did.
- Call out weak answers. If my answer is vague, hedged, safe, or sounds like a LinkedIn post, tell me so and ask again harder.
- Push for the arguable version. If what I said is something most people in my industry would agree with, say: "Everyone believes that. What do you believe that they do not?"
- If I am stuck, do not let me off the hook. Ask me something concrete instead. Examples: "Tell me about the last time you overruled your whole team." "What is the dumbest thing everyone in your industry does?" "What did you get wrong last year and what did it teach you?"
- Do not move on until the current section is actually done.
- Tell me when we are halfway. If I need to stop, summarize where we are so I can pick this back up later.
WHAT TO COVER, IN ORDER
1. POINTS OF VIEW. What I believe about my market, my customers, my industry, and how to run a company that smart people would argue with. Push until I have TEN. For each one ask: who disagrees, and why are they wrong? If it is not arguable, throw it out and make me try again.
2. RULES OF THUMB. What I say in every meeting. The lines my team can finish or me. Ask what I always ask first when a deal comes in, when something breaks, when a hire is not working.
3. NON-NEGOTIABLES. What I will never do, no matter what the numbers say. For each one, ask when I was tempted and what happened.
4. EXAMPLES. Real ones. A piece of work that was good and one that was bad, and why. A decision that went right and one that went wrong. Ask what the difference was. Save my exact wording, not your cleaned-up version of it.
5. WHAT ONLY I KNOW. Which customer is fragile right now. Which metric is vanity. Which project is dead while everyone keeps pretending. Who is carrying more than their title says.
6. GAPS. Where I am guessing. Where I am weak. What I avoid because I do not understand it. Ask me directly.
HOW TO WRITE IT UP
When we are done, write the whole thing out as ONE clean document I can copy straight into Word or Google Docs. Plain headings. No special formatting.
- Use my words. Keep my phrasing and my bluntness. Do not translate me into corporate language.
- Every point of view gets its reasoning and the real example behind it.
- End with a section called OPEN QUESTIONS listing everywhere I was vague, contradicted myself, or dodged.
- Do not pad it. If I only gave you seven real points of view, write seven and tell me I owe you three.
Start with question one. Nothing else.
Now the part that decides whether this is worth anything.
Do not butter it up. Do not dress your answers in corporate language to sound reasonable in front of the machine. Say it the way you would say it to a friend who runs a company, with the profanity if that is how it comes out.
Every time you soften a thought to sound professional, you sand the point off it. And a sanded answer is worth nothing to you six months from now. Blunt in, sharp out. Polished in, average out.
STEP 2. SAVE IT AS A DOCUMENT.
When the interview is done, ask it to write the whole thing out.
If it stops partway through, just say keep going. Expect somewhere between two and five pages. If you got one page, you were not honest enough, and you already know it.
Select it. Copy it. Open a blank Word document or a blank Google Doc. Paste it in. Name it My Second Brain. Save it.
Read it once. If a line does not sound like you, change it. This is your document, not the machine's.
One warning. This document could have real customer names, real numbers, and real opinions about real people in it. Treat it like a board document, because that is what it is. Keep it wherever your company already approves.
STEP 3. BUILD THE PROJECT.
Open Claude. Left sidebar, click Projects, create a new one. Name it after your company.
A Project has two sides, and almost everybody puts things in the wrong one.
The knowledge side is what the machine should know. Upload My Second Brain here. That is it. Just upload the file.
The instructions side is how the machine should behave. That is step four.
On ChatGPT this works the same way. Projects in the left sidebar, upload the document as a source, and there is a project instructions box. Same two sides, different labels.
On Microsoft Copilot there is no instructions box that stays on for you. Keep the document in OneDrive and point Copilot at it at the start of anything that matters. More work, same idea.
STEP 4. TELL IT WHEN TO USE THE THING.
This is the step people get wrong in both directions.
Skip it entirely and the machine ignores your document. It does not automatically decide your second brain is the most important thing in the room. Ask about pricing and it may never think to open the file that explains how you think about pricing.
But do not swing the other way either. Your second brain has no business anywhere near most of what you do all day. Fixing a typo does not need your points of view. Formatting a table does not need your non-negotiables. Loading your whole brain into every trivial request is friction, and a waste of tokens!
So tell it when. Copy this into the instructions side. It works word for word in Claude or ChatGPT.
I have a document called My Second Brain. It holds my points of view, my rules of thumb, my non-negotiables, and what I know that nobody wrote down
USE IT when the task involves my judgment, my voice, or a real decision. That means strategy, board and investor material, anything that goes out under my name, hiring calls, pricing, customer decisions, and anything where being wrong actually costs something.
DO NOT USE IT for quick mechanical work. Typos, formatting, math, summarizing something I just handed you, pulling a fact. Do not slow those down. Just do them.
When you do use it:
- Reason from my points of view, not from industry consensus. If the two disagree, go with mine and tell me you did.
- Write in my voice. Short sentences. Plain words. No corporate language.
- If what I am asking for crosses one of my non-negotiables, say so before you do the work, not after.
- If my second brain does not cover something that matters here, say "your second brain does not cover this" and ask me. Do not fill the gap with generic answer and sound confident about it.
- At the end of anything substantial, tell me if we landed on a new point of view, rule, or decision worth keeping. Say "add this to your second brain" and give me the exact line to paste.
That last line matters more than all the others. It means the document grows on its own. You do the work, the machine flags what is worth keeping, and your context gets sharper every week instead of going stale in a folder.
STEP 5. TEST IT.
Do not take my word for any of this. Test it in five minutes.
Pick a real question you are sitting on this week. Something only you would know how to answer well.
Ask it in a regular chat, outside the Project. Read what comes back. It will sound fine. Polished, confident, and written by nobody.
Now ask the same question inside your Project. Read that one.
That gap is what everybody in your industry is missing.
WHEN IT DOES NOT WORK
Three things go wrong, and they are all fixable in a minute.
It still sounds generic. Your points of view were too safe. You wrote what everybody thinks. Go find the lines nobody would argue with and either sharpen them or cut them.
It seems to ignore your document. You are probably in a regular chat instead of inside the Project. Check the top of the screen. Or the task did not look like a judgment call to it, in which case just say use my second brain for this.
You changed the document and nothing changed. You did not re-upload it. Editing your copy does nothing. The machine reads the one you uploaded.
WHAT TO EXPECT
Nothing gets smarter. What changes is that the machine stops handing you the average.
You will feel it first in board prep, hard emails, strategy memos, and the hiring call you keep going back and forth on. Anywhere you were about to rewrite the output because it did not sound like you, you will not have to.
And on everything else, keep doing what you did in the first 30 seconds of this issue. Before you hit enter, ask what this thing does not know that you know. Then type that part.
The second brain handles the context you were tired of retyping. You still handle the rest.
WHAT COMES NEXT
This is the beginner version, and it works. A Word doc and a Project will get you most of the value.
There is a version that runs itself. The machine reads your board decks, your customer calls, and your notes, and keeps the whole thing current without you touching it. People are building that right now with tools like Obsidian, and I am going to write it up for you soon.
Do not wait for that issue. An advanced second brain you never built is worth exactly nothing. A simple one, sitting in a Word doc, will change your week and probably your month.
Here is your figure it out moment. Run the interview today. Not this weekend. Today. It takes an hour, and you already know every single answer, which is the entire point.
Kathy
THE AI BOSS