@DamiDefi: <i>The Cowork Setup That 10x m...

The Cowork Setup That 10x my productivity.

You're not getting bad output because Claude is weak. You're getting bad output because you start from zero every single session.

I used to spend 20 minutes at the start of every Claude session re-explaining who I am, what I'm building, and how I think. Then I built a Cowork. That dropped to zero. Claude already knows my work, my voice, and what I've built before I type the first word. The first message I send is the actual task.

Claude doesn't have a memory problem. You have a structure problem.

Here's how to fix it.

What Cowork Actually Is

Cowork is not a Claude feature. It's a system you build inside the Claude desktop app.

Regular Claude chat reads only what you type. Cowork gives Claude access to your actual files before you say anything. It reads your background, your projects, your templates, and your preferences automatically. Every session starts informed.

This is not about prompting better. It's about never having to prompt the basics again.

Step 1: Install the Claude Desktop App

Go to claude.com/download and install the desktop app. This is not optional. Cowork runs on the desktop app's ability to read local files. Claude.ai in your browser cannot do this.

Once installed, open it and log in. You're building here from now on.

Step 2: Choose the Right Model

Open the model selector. Use Claude Sonnet 4.5. It's the current model with the best balance of speed, quality, and context handling for daily work. Extended Thinking is not necessary for this setup and slows the workflow down. Turn it off.

Step 3: Create Your Cowork Folder

Create a single master folder on your desktop or in your documents. Call it COWORK.

Inside it, build exactly four subfolders:

ABOUT ME, PROJECTS, TEMPLATES, CLAUDE OUTPUTS.

This is the structure Claude will read from. Everything you build from here lives in one of these four places.

Step 4: Write Your About Me File

Inside the ABOUT ME folder, create a file called about-me.md.

Write these things: who you are, what you do, how you think, what you're building toward, and the tone you write in. Include your audience and what they care about. Be specific.

This file is what Claude reads before every single response. The more specific it is, the less you explain. Most people skip this step and wonder why Claude sounds generic.

Step 5: Build Your Projects Folder

One subfolder per project. No exceptions.

Inside each subfolder, create one file: project-brief.md

Write five things in it: what the project is and why it exists, the goal and deadline, tone and audience for this specific project, what's already been done, and what Claude should never suggest for this project.

After this, you never re-explain a project again. You open the subfolder, Claude reads it, work starts.

Step 6: Build Your Templates Folder

This is your reusable weapon bank.

Every time Claude produces output you'd use again, save it here as a named .md file. cold-email-template.md. proposal-template.md. weekly-report-template.md.

Next time you need that format, paste the template and say: "Use this structure." Output stays consistent. Quality compounds. You stop rebuilding the same thing.

Step 7: Set Up Your CLAUDE OUTPUTS Folder

After every session, copy the best output into a dated file. Name it by task and date. homepage-copy-apr15.md.

Over time this becomes a library of your best AI-assisted work. Claude can read your past outputs and calibrate its standard to what you've already approved.

Step 8: Write Your Anti-AI Style File

Create a file called anti-ai.md inside ABOUT ME.

List every word and pattern you refuse. Here's a real starting list:

Words to ban: leverage, evolve, space, ecosystem, empower, foster, unlock, seamlessly, revolutionize, game-changing, deep dive, at its core, in today's world, it's worth noting.

Patterns to ban: starting sentences with "Certainly" or "Absolutely", ending with "the future is bright", using the phrase "not just X but Y", any sentence that could appear in a press release.

Add to this file every time Claude produces something that sounds like a robot. After four weeks, the outputs stop needing cleanup.

Step 9: Set Global Instructions

Go to Settings, then Profile, then Global Instructions.

Write how you think, what formats you prefer, what you hate, and what you always want included. This runs across every Claude conversation, not just Cowork. It is the baseline layer everything else sits on.

Step 10: Always Use Projects, Not New Chats

In the Claude sidebar, create a New Project and point it at your COWORK folder.

This is what ties everything together. Projects retain context across sessions. New chats don't. Every time you open a new chat instead of your project, you're throwing away the system you built.

You've now built the system. Part 2 covers how to work inside it so the outputs match the setup.

Follow @damidefi on X for daily Claude AI tools, crypto analysis, and the full journey to 100K. Bookmark this. Share it with one person still starting from zero every session.