The AI Second Brain

Your AI has amnesia

Most people use AI like a stranger they meet every morning. They explain who they are, what they do, and what they need. The chat ends, and all of it is forgotten. That is why the answers feel generic. Not because the model is weak, but because it knows nothing about you.

A second brain fixes that. You give the AI your context once, in files it reads every time, and every answer after gets sharper. This is the exact setup I run my business and life on. It is not locked to one model either, so you can switch between Claude, Codex, or whatever is best that week, and your context comes with you.

No coding. If you can install an app and fill in a form, you can do this. Set aside 30 to 45 minutes. Here is the whole thing in five steps.

Step 1: Install Claude Code

Claude Code is an app from Anthropic. It is Claude with the ability to read and write files on your computer. That is what lets it hold a second brain instead of forgetting everything when you close the chat.

The easy way, if you are non-technical:

1. Go to claude.com/claude-code.
2. Download the desktop app for Mac or Windows.
3. Install it like any other app and open it.
4. Sign in with your Claude account. You need a paid Claude plan or an Anthropic API key; the sign-in screen walks you through it.

If you are comfortable with a terminal, Claude Code also runs there and inside editors like VS Code. It is one command, npm install -g @anthropic-ai/claude-code (you will need Node.js first), then run claude. If those words mean nothing to you, ignore this and use the desktop app.

Step 2: Download the template

The template is the skeleton of a second brain: a set of folders and files Claude reads to understand you. Grab a copy.

1. Go to github.com/alexsidhu1/execassistant-template.
2. Click the green Code button near the top.
3. Click Download ZIP.
4. Find the ZIP (usually in Downloads) and double-click to unzip.
5. Move the execassistant-template folder somewhere you will remember, like your Desktop. Rename it to anything you like.

No GitHub account needed. You are just downloading a folder.

Step 3: Open the template in Claude Code

1. Open Claude Code.
2. Open the folder you just unzipped. There is an "open folder" option when you start. If you prefer VS Code, download it and open the folder there.

Claude can now see every file in that folder. When you start chatting it reads CLAUDE.md first, which tells it everything else to look at. To check it is working, type:

Read CLAUDE.md and the files in the context folder, then tell me in one sentence what you understand about me so far.

It will mention placeholders, because you have not filled anything in yet. That is the next step.

Recommended: run it inside VS Code

The desktop app is the easy on-ramp. But the way I actually run this, and the way most serious Claude users I know run it, is inside VS Code. VS Code is a free, open-source editor from Microsoft. Do not let the word "editor" scare you. You are not writing code. You are using it as the window your assistant lives in, and it is the same tool whether you touch code or not.

Two reasons it is worth it. You can see everything: every file in your second brain sits in a sidebar, so nothing is hidden. And you can watch it work: when your assistant creates a skill or updates a file, you see it happen live. That visibility is the difference between a demo and a workbench.

Here is the setup:

1. Download VS Code from code.visualstudio.com and install it.
2. Open your template folder: File → Open Folder, then pick your execassistant-template folder. The files appear in the sidebar on the left.
3. Open the built-in terminal: View → Terminal. A panel opens at the bottom.
4. Type claude and press enter. Claude Code starts right there, already pointed at your second brain.

From then on you talk to your assistant in the terminal on one side and watch your files change on the other. That is the whole setup, and it is the one I use every day.

Step 4: Make it yours

This is where the value comes from. The assistant is only as good as what it knows about you. Two ways to do it. Pick one.

Let Claude interview you (easiest). You do not have to edit files by hand. Paste this in:

I want to set up my second brain. Read CLAUDE.md and every file in the context folder. Then interview me one question at a time to fill them in with my real details. Ask about who I am, what my business does, my team, my current priorities, and my goals. After each answer, update the right file. Keep going until the context folder reflects my real life.

Then answer its questions in plain language. It writes the files for you. This is the fastest way, and honestly the best, because it asks about things you would not think to include.

Or edit the files yourself. Open these and replace the bracketed placeholders with your real information, in order of importance:

1. context/me.md — who you are, your role, how you work.
2. context/work.md — what your business does, your offers, your clients.
3. context/team.md — who is on your team and what to loop them in for.
4. context/current-priorities.md — what matters in the next 30 days.
5. context/goals.md — your targets for the quarter.
6. CLAUDE.md — change the name and top description so it is working for you.
7. .claude/rules/communication-style.md — set the tone so it sounds like you, not generic AI.

Be specific. "I run a business" is useless. "I run a 4-person marketing agency in Austin, most revenue from retainer clients, trying to land bigger accounts" is gold. The detail is the whole point.

One rule: do not put anything in here you would not want a stranger reading, unless you keep the folder private. By default it lives only on your computer and stays yours.

Step 5: Use it

Now it is an assistant. Talk to it like one:

• "Draft an email to a client who went quiet, in my voice."
• "What are my top priorities this week, and what should I ignore?"
• "I am thinking about hiring a salesperson. Talk it through with me using what you know about my business."
• "Summarize this messy note into three clear next steps." (paste the note)

Tell it to remember things. When you correct it or state a preference, say "remember that." For example: "Remember that I never want emojis in client emails." It saves that and applies it next time.

Log decisions. When you make a real call, say "log this decision." It writes it down with the reasoning, so future-you knows why.

Build skills as you go. The first time you do a repeatable task, do it once. The second time, say "turn this into a skill so we do it the same way every time." Now it is a reusable recipe.

What "gets smarter over time" actually means

Nothing here is automatic magic. It compounds because of you. Every time you correct it, add context, log a decision, or save a skill, the next conversation starts from a higher floor. After a few months it knows your business the way a good employee would, except it never forgets and it is there at midnight.

The hardest step is the first one: filling in the context honestly. Do that, and the rest takes care of itself.

Where this stops working

I will be honest about the ceiling. A personal second brain runs on your own discipline and lives in one folder, for one person. That is plenty for running your own work.

It is not enough to run a company. When your whole team needs the same context, when it has to pull live from your CRM, your inbox, and your project tool, and stay current without anyone maintaining it by hand, a folder of files hits a wall.

That is the line between a second brain and what we build at Whitehorse: a company brain. Same idea, industrial version. Trained on everything the business knows, connected to the tools you already use, kept current automatically so the whole team works from one source of truth. If you hit that wall, that is a good problem. It means the system is working and you have outgrown the DIY version. That is where we come in.

Stuck?

Ask Claude Code directly. It can see your files and your setup, so "I followed the guide and X is not working, help me debug it" usually sorts it out. That is the nice thing about an assistant that lives inside the tool: it can help you fix the tool.

Built by Alex Sidhu, Whitehorse AI.