Claude Cowork Just Did 7 Days of Work in 15 Minutes

Hey everyone, I want to tell you about the ChatGPT moment — but for business work — that happened this week. It was when Claude released Claude Co-work.

Co-work is a new experimental product from Anthropic, the makers of Claude. And what it essentially is, is a web app frontend to their really famous and powerful Claude Code product. Claude Code is an agentic coding platform that developers have been using.

I’ve been teaching myself Claude Code, but Claude Co-work is way, way easier. So I want to explain why this is going to be a seminal moment in the history of AI for business use cases. I want to walk you through Co-work, show you what the heck it is, and how you should think about using it — all on today’s show.

The first thing I want to go through is what happened. What is Co-work? What is happening?

On January 11th, Anthropic released Co-work. And there’s a couple interesting things about Co-work, exclusively with Claude Code. Claude Code wrote the Co-work product in about a week and a half.

So we’re seeing a massive shift in velocity of products coming out through this new AI-driven coding model.

But really what Co-work is — as you can see in the video here — is it’s an easy, non-technical person’s way to use Claude Code.

What I’ve been seeing a lot is: where Claude Code might be your CTO — your chief technology officer — Claude Co-work might be thought of as your chief operating officer. It helps you operate your life a lot better.

I thought this was a very interesting tweet from Vib — who I don’t know — but he basically says: “Hey, I installed Claude Co-work yesterday. Two hours later, it produced 14 job descriptions. I’ve been getting around to a marketing strategy doc with budget allocation, partner emails, website copy, responses to 23 LinkedIn DMs. Two months of work, two hours.”

This is a hyperbole. This stuff wouldn’t have taken two months. This stuff would have taken like a day or two. But it is the next level of automation for what I’d say is personal productivity for business use cases and work use cases.

I’m sure you’re going to use some of this stuff in your personal life, but it’s going to be really valuable in your work life.

And guest friend of the pod — all-around amazing human — Lenny Rachitsky, talks about one of the most important things of Claude Co-work. What makes Co-work so interesting is that you can give Co-work access to the files on your computer.

It can read, write, it could organize things. It has a lot more access — and subsequently context and power — than the traditional version of Claude, where you can create a project and upload some files, but you’re limited to how much context you can actually upload.

For example, Lenny gave it 320 podcast transcripts and had it go through and give some of the top lessons and principles from that podcast. That’s a huge amount of context. That wouldn’t have been possible with the current Claude web app experience or desktop experience without Claude Co-work.

So I want to take you right now and walk you through Claude Co-work, show you a couple examples, and how it actually works.

To use Claude Co-work, the first thing you’re going to have to do is install the Claude desktop app.

The other thing I’ll say: if you want to try Claude Co-work right now, it’s only available to Claude Max subscribers, which is $100 a month. The reason for that is this is an early research preview, fresh off the presses. They’re trying to get feedback and it’s very, very new. So they’re trying to limit it to a smaller pool of users.

That is the caveat.

You need to use it on the desktop app, and you need to be a Max subscriber.

If you are a Max subscriber, what you’ll see is: you have your historical chats. You have your Claude Code experience — where if you’re running the actual Claude Code, you’re there. And then you have Co-work.

Co-work is a new tab there on your sidebar. And they show you, right from the beginning, some of the tasks you should do. You can code a prototype. You can send a message because it can integrate with your email services — your Gmail account, etc. You can crunch data, create files, organize files, prep for your day because it can access your calendar.

The other thing about Co-work: if you look here, it says these tasks are run locally and aren’t synced across other devices. What the hell does that mean?

Well, that means it’s processing them on your computer, not in the cloud — which means everything you do is going to be done locally to your computer. And if you’re using Claude across multiple machines, it’s not going to sync across them.

But it also means that because it’s processing on your computer, it can kind of happen in the background. You don’t have to do a lot of that annoying stuff where you accidentally close a browser tab or you walk away and it times out. All that kind of stuff. It’s running locally on your machine, which is a little bit of a better experience.

So you just saw what Claude Co-work can do. Here’s how you actually use it.

We built 12 prompts specifically for Claude Co-work that can access your folders, work in the background, and create real deliverables. They’re designed to help you start replacing days of work. Get it right now. Scan the QR code or click the link in the description. Now, let’s get back to the show.

The first thing you’re going to notice with Claude Co-work is this interesting different thing: you can add a folder, or you can add connectors.

The big thing we’re going to talk about today is folders because, again, you’re running this on your computer — it can access and interact with everything on your computer.

So, for the sake of the show today, I really liked what Lenny did, and this is just an example. I’ve got a folder here I downloaded of about a hundred transcripts from Marketing Against the Grain, and it’s going to ask for permissions to access that folder. And now it can basically access all the transcripts that are in that folder on my computer.

Additionally, I want to add one other file. I want to add a CSV that’s called “table data.” And table data is literally just a CSV export of all of our YouTube analytics and how the podcast has performed.

Now that I’ve done this, I can do a whole host of things.

I can simply say: okay, looking at the podcast data and the show transcripts, generate a new format of our podcast that will generate more subscribers on YouTube.

It’s a very basic query. I can say, “Let’s go.”

You’ll see it’s saying “starting agent,” but you do have to hit allow a few times. There’s some permissions and everything that comes up. And basically what this is now doing is: locally on my computer, it is running and analyzing.

It’s going to analyze the podcast data to help generate a new format that could drive more YouTube subscribers.

“Let me start by reading the uploaded file.” So it’s first looking at all of the data and understanding the CSV. “Now, let me read the CSV portions and also check the workspace folder for any transcript files.”

So it’s finding all the transcripts. “I have both your YouTube analytics data and many transcript files. Let me read a few more sections of the data and sample top-performing transcripts to understand what’s working.”

Okay — I didn’t tell it this much. It figured out everything that it had here.

“And now I have a good understanding of your podcast data and content. Let me read one more high-performing transcript and a couple short-form video transcripts to get the full picture.”

So it’s telling you exactly what it’s doing.

“What type of deliverable would be most useful for your new podcast strategy? Strategy document, presentation deck, one-page summary, or something else?”

You know what? Let’s see what a deck would be.

“What’s your primary constraint for implementing the new format? Time, budget, team size.”

Time.

Okay — now you’ll see it said “using to-do,” right? You can see that Claude Co-work is much more technical. It’s showing you that it’s using different services and scripts and it’s reading the files.

It’s giving you more transparency, but you can also feel that it is a lighter-weight version of Claude Code. They’ve built a nicer experience for non-technical people like me — likely you — on top of Claude Code.

Also pretty fast.

“Okay, so now let me create the presentation.”

Based on my analysis of your data, I see clear patterns in what drives subscribers. Top subscriber drivers: videos with “How I run a zero employee agency,” “World’s best landing page,” “GPT-4.0,” “Replace your creative teams,” “Best CTR video,” “Best short-form sweet spot,” “10 to 20 minute tutorials with live demos perform best,” “High impact titles specific around time savings.”

So now that it understands the problem, it’s going to go and build me a presentation about what to do.

You can see it’s sending a request. It’s actually writing the HTML for the individual slides. So what it’s doing is it’s using Claude Code to code the slides to give you this output.

This would be really valuable if you were working on a team and you wanted to take this presentation, share the learnings that you derived with Claude, and share them with other people — which is why I picked presentation. A strategy document is a little bit more of an obvious output — probably a little bit better if you’re just Kieran and I reading through it — but in this case I wanted you all to realize that Claude Code codes lots of things, including slides, which doesn’t get talked about that much.

What you’ll see here now is it’s building my slides. And now it’s creating a JavaScript file to build the presentation with charts. So it’s building me charts within my slides, which is pretty cool.

But you can see, again, it’s using very different interactions than traditional Claude. Claude wouldn’t tell you that it was using JavaScript to build presentations.

This is definitely a more technical version of Claude, but a much less technical version of Claude Code.

One of the good things about it is you can just go do this, set it to run, and go work on other stuff — and then come back when it’s done — because it’s self-contained to the desktop app. It’s running on your computer. You can go do things in your browsers and other applications while this is working in the background.

“All right, I’ve created your presentation. Here’s a summary of your new podcast format strategy.”

New format: AI Marketing Lab. This is something we had actually been talking about, so this is pretty cool.

Based on our analysis of 200+ videos and transcripts, here’s what drives the most subscribers. What’s working? What’s not? New format pillars: build-along tutorials, solo founder spotlights, interview AI-native entrepreneurs, tool showdowns.

Winning formula equals: time saved plus specific outcome plus method or tool.

And so here’s a presentation to get 100,000 subscribers this year. Let’s look.

You can open this in Keynote, but I’m going to preview it right here so you can see how it works.

Marketing Against the Grain: Data-driven format for 10x subscriber growth.

What the data reveals: the core sweet spot duration is 15 to 25-minute tutorials with step-by-step walkthroughs — drive 3x more subscribers.

New format: the AI Marketing Lab. Build-along tutorials. Solo founder spotlights. Tool showdowns.

High converting title formula: time saved plus specific outcome plus method or tool. It came up with a title formula.

“The 18-minute blueprint.” I like this.

It shows you exactly what you need to do: build part one, build part two, results, charts, here’s the context, and it starts the hook.

Quick wins to implement this week: add chapter markers. Three-second hook rule. Double down on Claude — Claude content has 6.26 CTR. Claude seems a little biased towards Claude. Pin comment strategy. Killer short-form lead magnet per video.

It’s pretty cool. It gives you the whole projection to get to 100k subs.

And: let’s build the first episode and pick a pilot topic. Select a blueprint, record, and ship. Because again, we told it time is the barrier here — we want to get started as soon as possible.

So in literally 15 to 20 minutes — I don’t know the full runtime there — what we’ve done is: we’ve looked through hundreds of transcripts. We’ve looked through thousands of rows of data. We’ve taken all of that. We’ve come up with insights against a given goal. And we now have a presentation that Kieran and I could show anyone around what we want to do next.

And I could have been doing two or three other things while this was all happening.

This is why I think this is the ChatGPT moment for business work. There’s just a level of complexity of thinking that exists now when you look at the Claude Opus model combined with the ability to really access a lot of information that Co-work makes much easier.

You can see its ability to programmatically create a lot of emails, to do a lot of the mundane tasks you have to do — much faster and at scale with Co-work.

At first you’re like, “Oh, 100 bucks a month — that’s a lot.” And then you’re like, “Wait a second. What would you pay for a growth strategy for an important project you’re working on?”

I’d pay way more than 100 bucks for a really good new show format to get us more subscribers on the podcast — and yet I just did that in 15 minutes this month, and I’m probably going to do 10 other things that are worth thousands of dollars this month for my $100.

So I do think it’s good value.

It’s a really cool product. It’s still early. There are bugs. You’ve got to hit allow. You need to be okay with the rough edges.

But to me, Claude Co-work is showing everybody what is truly possible with AI when you can get the right context with the right intelligence of the model.

So my call to action: all of you, go out, give it a try. Drop a comment with any of the cool use cases you’ve used Claude Co-work for. And I want your thoughts.

Are you excited to use it? What’s stopping you from using it? If you’re not excited to use it, would love to hear all that.

Drop those in the comments below. Please hit subscribe. Please hit like. We’ll see you real soon on Marketing Against the Grain.