Claude Cowork: three things nobody explains

Most people open Claude Cowork for the first time, look at it, and think: this looks exactly like a regular chat window. So they use it like one. And then they wonder why it feels like a slightly fancier version of something they already have.

That's not a user problem. It's a framing problem. Nobody explains the three things that make Cowork genuinely different, and until you understand all three, the tool never quite clicks.

Here they are.

Thing One: You stop bringing files to Claude. Claude comes to your files.

With regular Claude chat, getting files into the conversation is a manual process. You find the files on your computer, you attach them one by one, you wait, you get an output, you download it, you move it to the right folder, and next time you start over.

Cowork removes that entire loop.

Instead of attaching files, you point Cowork at a folder on your computer. It reads everything inside directly. When it creates something, a spreadsheet, an HTML report, a summary document, it saves that file straight into your folder. No downloading. No moving things around. The output lands where it belongs, immediately, and it stays there so you can find it later.

The practical shift is this: with Claude chat, you bring your work to the AI. With Cowork, the AI works where you already are.

To set this up properly, create one central folder on your computer that becomes your permanent Cowork hub. Give it a name, point Cowork at it every time you start a task, and put all your subfolders inside it. Think of it as your AI's home base on your machine.

Thing Two: The claude.md file is the most important thing in your entire setup.

This is where most people miss the real power of Cowork.

Inside your central folder, you create a plain text file and name it claude.md. That name is not arbitrary. Every single time Cowork starts a task in your folder, the first thing it does is read that file. Before it looks at anything else. Before it starts working on whatever you asked it to do.

This file is your standing instruction manual. Write it once and it applies to every task from that point forward.

What goes in it? Everything Claude should know about you without you having to say it out loud every session. Who you are. How your folder is structured. The tone you want. How you prefer outputs formatted. Any specific rules you want it to follow, like always save reports to the reports subfolder, always build HTML files rather than PDFs, never delete anything without asking first.

You can also build specific workflows into it called skills. A skill is a set of step-by-step instructions for a task you do repeatedly. You write the instructions once inside the file. Then when you want that task done, you just say the name. Cowork reads the skill, knows exactly what to do, and executes it without you typing out the process again.

Here is what that changes in practice. Without a claude.md file, every new Cowork task starts completely fresh. You explain who you are. You describe your workflow. You specify how you want things formatted. Then you do it all again next time.

With a claude.md file in place, you just say what you want done. Claude already knows everything else. The instructions live in the folder, not in your prompt. That is the shift that makes Cowork click.

The good news: you do not have to write this file yourself. Open Cowork, make sure your central folder is selected, and paste in this prompt:

Answer the questions honestly, one at a time. When it's done, both files will be sitting in your folder ready to go. You can open either one, read it in plain English, and edit anything that doesn't look right.

Thing Three: The memory.md file means your assistant actually gets better over time.

The second file is memory.md, and it works differently from claude.md.

While claude.md holds your standing instructions, memory.md is where Claude writes down what it learns about you as you work together. Tell it once that you want reports sorted by date. It writes that down. Next week, next month, it still knows. Not because of some invisible background process you can't see or control, but because it's sitting there in a text file in your folder.

Open it any time. Read exactly what Claude has retained about your preferences. Edit anything that's wrong. Delete anything that's outdated. It is completely transparent, completely in your control, and completely unlike the black box memory in regular AI tools where you have no idea what the AI has or hasn't remembered.

Between these two files, your assistant compounds. The longer you use it, the more precisely it knows how you work, what you care about, and how you want things done. That is not something you get with a chatbot.

What all three of these add up to

Local file access means the AI works where your files already live. The claude.md file means the AI knows your workflows, your preferences, and your folder structure without being told every time. The memory.md file means the AI gets better and more aligned with how you work the more you use it.

At that point it stops being something you talk to when you need a quick answer. It becomes something that executes your work, the way you want it done, without you having to manage every detail.

That is the shift. That is what Cowork actually is.

Set up your central folder this week. Run the setup prompt above. Get both files in place. The whole process takes under 20 minutes and the first time you run a task and realize you didn't have to explain anything, you will immediately understand why this is different.

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

AI will eliminate 300 million jobs in the next 5 years.

Yours doesn't have to be one of them.

Here's how to future-proof your career:

Start learning AI now

One question this week and it's a genuine one.

You've read about the Slack Kanban skill. You've seen how Cowork works. Both are tools that shift AI from something you talk to into something that works alongside how you already operate.

What's the biggest workflow in your day that still feels manual and repetitive - the thing you keep doing by hand that you know shouldn't require a human?

Hit reply and tell me. The best answers become the next Inner Circle Edge.

Stay sharp.