Build a Second Brain in Claude: Set It Up Once, Never Prompt From Scratch Again

Here's the tax nobody talks about with AI: you explain yourself over and over. Every new chat, you paste the same context. Who you are. What your business does. Who you're talking to. How you like to write. What you already decided last week. You spend the first ten minutes catching Claude up, and then the output still comes back generic, because the version of you it's working from is a thin summary you typed while impatient.

Then you close the tab and it forgets all of it. Next task, you start over.

The problem was never your prompting. It's that a blank chat has no memory of you. The fix is to stop prompting from scratch entirely and build a second brain: one Claude Project that permanently holds the context of who you are and what you're working on. Set it up once, and every task after that starts with Claude already knowing you. Your best ideas stop scattering across a hundred dead conversations and start living in one place that gets smarter over time.

This guide is the whole setup. It takes about half an hour, and it's the highest-leverage thirty minutes you'll spend with AI all year.

What You'll End Up With

A single Claude Project that knows who you are, your business, your goals, and your voice, without being told each time
The ability to say "draft the launch email" or "what's my next move on this" and get an answer in full context, in your voice, with zero setup
A home for your ideas, notes, and decisions, so nothing good gets lost in a closed tab
An anti-AI style guide and a voice profile, so everything it writes sounds like you and not a robot

What a "Second Brain" Actually Means Here

It's not complicated. A second brain is just a Claude Project loaded with a few documents about you and your work. Claude reads them as background on every task inside that Project. So instead of you being the context each time, the files are. Four kinds of knowledge go in:

Knowledge	What it does
Who you are	Your work, your audience, your goals, your offers, so advice is actually about your situation
How you sound	Your voice profile and anti-AI style guide, so writing comes out as you
What you know	Your notes, ideas, and past decisions, so nothing gets re-litigated or lost
The standing order	Instructions that make Claude use all of it by default, every time

The steps below build each one. You only do this once.

Grab the starter files

Don't build these from a blank page. Download the ready-made Markdown templates, drop them into your Claude Project, and fill in the blanks. Each one includes the prompt to complete it.

⬇ about-me.md — who you are + interview prompt ⬇ style-guide.md — the anti-AI writing rules ⬇ current-context.md — what you're working on ⬇ voice-samples.md — your best writing ⬇ custom-instructions.md — the standing order ⬇ README.md — the 4-step quickstart

📦 Download all 6 files (.zip)

What You Need Before You Start

A Claude account (Projects are a Pro feature, so you'll want Claude Pro)
About 30 minutes and honest answers about yourself and your work
Optional: any notes, docs, or brain dumps you already have lying around

No technical skill required. If you can paste text into a box and answer questions, you can build this.

Step-by-Step Setup

1 Create the Project (2 minutes)

This is your second brain's home. A Project keeps its own instructions and knowledge files, and Claude remembers them across every chat you start inside it.

Go to claude.ai and sign in
In the left sidebar, click Projects, then Create Project
Name it something like "My Second Brain" and create it

From now on, start your work inside this Project instead of a blank chat. That one habit is what makes the rest pay off.

2 Load Who You Are (10 minutes)

This is the file that ends the re-explaining tax. Rather than describe yourself from a blank page (hard, and you'll leave out the important parts), let Claude pull it out of you with an interview. Open a new chat inside your Project and paste this:

You're going to help me build an "About Me" profile so you always know who I am and how I work, and I never have to re-explain myself in a new chat again. Interview me one question at a time. Ask a single question, wait for my answer, then ask the next. Dig deeper when I'm vague or generic. Cover, at minimum: - Who I am and what I do - My business or work: what I sell or make, and exactly who it's for - What I'm trying to build or achieve over the next year - The people I'm talking to, and the reaction I want from them - How I write and talk: words I actually use, words I'd never use, how formal or casual I am, my humor, my sentence rhythm - What you should always know about me before helping (my context, constraints, and preferences) - The kinds of tasks I'll bring you most often Ask at least 12 questions. When you have enough to represent me accurately, stop interviewing and output ONE Markdown document titled "About Me" with these sections: 1. Who I Am 2. My Work & Offers 3. My Goals 4. My Audience 5. My Voice (5 rules, plus words I use and words to avoid) 6. How to Help Me 7. Tasks I'll Bring You Often Write it so that, reading it cold, you could pick up any task from me with full context and sound exactly like me. Start with your first question now.

Answer in your normal typing voice. Don't polish it, because the messiness is the data. When Claude outputs the "About Me" document, read it, fix anything off, then add it to your Project knowledge (use Add content or the paperclip) as a file named about-me.md.

Why the interview beats a form. You know your own context the way you know your accent: you can't hear it, but it's obvious to everyone else. The questions surface things you'd never think to write down, and your unedited answers are the truest sample of how you actually sound.

⬇ Download about-me.md (prompt included)

3 Load How You Sound (5 minutes)

Your About Me file has a voice section, but a dedicated anti-AI style guide is what keeps Claude from sliding back into robot mode on longer writing. Copy the starter below, tweak anything that isn't you, and add it to the Project knowledge as style-guide.md.

# Anti-AI Writing Style Guide Rules for writing that doesn't sound like a robot wrote it. Follow these on everything. ## Banned words and phrases Never use: leverage (as a verb), unlock, unleash, elevate, empower, supercharge, delve, seamless, robust, game-changer, tapestry, beacon, testament to, navigate the complexities, realm, foster, embark. Never use these openers: "In today's fast-paced world," "In the ever-evolving landscape," "Whether you're a [X] or a [Y]," "Ever wondered…?" Never use the flip: "It's not just [X], it's [Y]." Never use hype transitions: "Let's dive in," "buckle up," "here's the kicker," "the best part?" ## Punctuation and formatting Stop using em dashes as a default connector. One every few paragraphs at most. Prefer a period or a comma. No emoji unless I ask for them. Don't bold a phrase for emphasis in every paragraph. Let the words carry it. Vary sentence length on purpose. Short ones. Then a longer one with a real clause inside it. Robots make every sentence the same length. ## Structure Get to the point in the first sentence. No throat-clearing, no "picture this," no dictionary definitions. One idea per paragraph. Blunt is fine. Do not end with a summary paragraph that repeats the piece. End on the sharpest line. Cut hedging: arguably, it's worth noting, in many ways, quite, very, really, just. ## Voice Write like a smart person talking to another smart person, not a brand talking to a market. Use concrete nouns and real examples instead of abstractions. Contractions are good. Sentence fragments are fine. If a sentence could sit in anyone's blog, rewrite it so it could only be mine.

Make it yours. The banned list is a starting point. Every time you catch a word you personally never say, add it. Ten words you actually hate beat fifty generic ones.

⬇ Download style-guide.md

4 Load What You Know (5 minutes, then ongoing)

This is the "never lose your best ideas" part. Anything you'd normally re-explain or dig for, put it in the brain: current projects, decisions you've already made, offers and pricing, key facts, half-formed ideas. Start with a brain dump. Paste this into your Project:

Here's a brain dump of everything on my plate right now. It's messy on purpose: [Paste everything: current projects, what you're working toward, decisions you've made, ideas you don't want to lose, offers and prices, anything you keep re-explaining.] Turn this into a clean Markdown file called "current-context.md", organized by project or theme. For each one, capture: the goal, where it stands right now, the next actions, and any decisions I've already made so we don't relitigate them. Ask me up to 5 questions to fill the biggest gaps, then give me the finished file.

Save the result as current-context.md in the Project. From here on, treat the brain like a notebook: when you have an idea or make a decision, drop it in. The Project is the one place it won't get lost.

⬇ Download current-context.md (prompt included)

5 Set the Standing Order (2 minutes)

Finally, tell Claude to use all of this by default. In your Project, find Set custom instructions (or "What should Claude know") and paste:

You are my second brain. Before helping with anything, treat the files in this project's knowledge as background you already know: about-me.md (who I am and my voice), style-guide.md (how I write), and my context files (what I'm working on). Rules: - Never make me re-explain who I am, what I do, or how I write. You already know. - Match my voice from about-me.md and follow style-guide.md on anything you write. Before showing me a draft, silently check it against the style guide and give me the corrected version, not your first attempt. - If a task needs context you genuinely don't have, ask me one sharp question instead of guessing in a generic voice. - When I share a new decision or idea, tell me which file to add it to so the brain stays current.

That's the whole machine. Five files, one habit of working inside the Project, and Claude stops being a stranger you brief every morning.

⬇ Download custom-instructions.md

How You Actually Use It: Never Prompt From Scratch Again

This is the payoff. Because the context already lives in the Project, your prompts get short and the output gets sharp. You stop briefing and start delegating:

"Draft the launch email for my new offer." (It knows the offer, the audience, and your voice.)
"I'm stuck on pricing. Given what you know about my business, what would you do?"
"Turn my messy notes from this call into next actions and add them to current-context.md."
"What should I focus on this week to move toward my goal?"
"Write a LinkedIn post about this idea, in my voice, 150 words, no hashtags."
"Reply to this email as me." (paste the email)

No pasted background. No "for context, I'm a…" preamble. You ask, it answers like someone who's worked with you for a year.

Other Recommendations That Make a Real Difference

Keep one brain, not ten. A single Project you refine over months beats a fresh chat every time. The compounding is the entire point.
Feed it your greatest hits. Add two or three samples of writing you're proud of as voice-samples.md and label it "match this." Real examples teach voice better than any description. Download the template ⬇
Capture in the moment. The brain is only as good as what's in it. Make it a reflex: idea, decision, or lesson, drop it in the same day.
Correct it once, then update the file. When Claude slips a "leverage" back in, don't just fix the draft. Add it to the style guide. The brain gets smarter every time you catch it.
Give it the job, not just the topic. Even with full context, "write a post for founders behind on AI, my voice, 150 words, one idea" beats "write a post." Specifics get you a usable first pass.

Get the whole brain in one download

All six Markdown templates, ready to drop into your Claude Project.

📦 Download the Claude Brain Starter Pack (.zip)

Want Your Second Brain Built For You?

If you'd rather have your whole setup, profile, style guide, context files, and instructions, built and dialed in so it just works, that's exactly the kind of thing I do. Drop your email and I'll be in touch.

Troubleshooting

"It still asks me for context I already gave it." Make sure your files are attached to the Project (in the knowledge panel), not just pasted into one old chat. Anything in the knowledge panel applies to every chat; anything in a single chat is forgotten when you close it.

"The 'About Me' file feels generic." Your interview answers were probably too clean. Redo Step 2 and answer like you're texting a friend, not filling out a form. Let it be messy. That mess is you.

"It still slips into AI voice." Add the specific offending word or phrase to style-guide.md the moment you see it, and end your request with "now give me a sharper version that sounds more like me." The second draft is almost always better.

"I don't have Claude Pro / can't use Projects." You can do a lighter version by pasting your About Me and style guide at the top of each new chat. It works, you just re-paste. The Project is what makes it permanent.

The Bottom Line

Most people use AI like a stranger they re-hire every morning, briefing it from zero, getting generic work back, and closing the tab on anything good it produced. A second brain flips that. You invest thirty minutes once, and from then on you're working with something that already knows you, in your voice, holding your ideas. You stop prompting from scratch, and you stop losing the good stuff.

Get the AI Starter Kit

The seven prompts I'd start with to turn what you already know into income, built for people who don't want to become "tech people."

Get the Free Kit