Ep. 29: An Introduction to Claude Code and Claude Cowork
In this episode I break down Claude Code and Claude Cowork, two tools from Anthropic that have taken the AI space by storm and allow Claude to actually execute tasks on your computer instead of just chatting with you. This episode covers what each tool does, how they’re different from each other (and from regular LLMs like ChatGPT), what “agentic” AI actually means, who each tool is for, and what to know about safety and usage limits before jumping in.
Quick but Important Note
Real quick before we hop into it, I realize that perhaps not everyone who subscribes to this newsletter follows me on IG or subscribes to my other newsletter, Maestro Mail. I speak out a ton on those platforms, and I want to make sure my stance on everything currently going on in the US is crystal clear: Fuck Trump, fuck ICE, I stand for none of this.
Please note, I’m not all anger though; I do have action items. You can CLICK HERE for a link to a resource I sent to my other newsletter with 4 action items that you can start doing right now.
By the time this episode drops the Senate will have voted on ICE funding, but it’s still great to be aware of 5calls.org, and there are plenty of other things you can contact your reps about.
Disclaimer
Today we’re talking about Claude Code and Claude Cowork because there’s a good chance you’ve heard these terms floating around the socials, and I want you to be informed.
I know the name of this podcast is ChatGPT Curious, but like I’ve said a zillion times before, my goal is to create educated users and consumers, not die-hard LLM loyalists.
Disclaimer: I haven’t used either one yet.
I’ll likely get into Claude Code sooner than later given my vibe-coding endeavors, but that time has not yet come, and thus this episode is about sharing information, not reviews or recommendations.
Claude Code and Cowork: The Simple Version
To set the stage: Anthropic is the parent company of Claude. Claude itself is an LLM, just like ChatGPT.
Claude Code and Claude Cowork are tools that let Claude actually work on your computer, instead of just chatting with you. You describe what you need, Claude figures out the steps, and then handles it.
Think of them as programs that wrap the Claude LLM.
Important note: both are only available to paid Claude subscribers.
Claude Code
Claude Code is a computer program, and was launched in February 2025. In tech-speak, it’s a command-line tool that runs directly on your computer and allows Claude to:
- Read files
- Write and modify files
- Run commands
- Execute multi-step workflows across a codebase
It’s really designed for developers, engineers, and people comfortable with code.
Originally, it could only be used through the terminal, via a command-line interface, which make sense because it was built for codebases and software projects,
Since then, it’s been adapted to work through IDEs (integrated development environments), the Claude app, and a web browswer version. Still, it’s very much coder-biased.
Enter: Claude Cowork
Claude Cowork was released on January 12, 2026 and is marketed as “Claude Code for non-developers.”
It handles things like:
- File management
- Document creation
- Data organization
Cowork is accessed through the Claude desktop app and is currently MacOS-only.
What’s The Difference?
Now, if you’re not totally clear on the difference between Claude Code and Claude Cowork, you’re not alone. At first glance, they seem very similar.
Two important takeaways here:
- That confusion is a good indicator that all you probably need is Claude Cowork.
- The real difference is execution capability.
Claude Code can operate the system, not just touch files. It can execute code, run commands, install things, talk to other apps, and connect to services.
Claude Cowork can read, write, and organize files, but it’s limited to file management.
Fun fact: the Anthropic team used Claude Code to build Claude Cowork in about 10 days.
Kitchen Analogy (Very Literal)
Despite my general distaste for Chat-generated analogies, I had it come up with a few so that I too could better understand and articulate the difference between Claude Code and Claude Cowork. I found it’s kitchen analogy to be quite helpful, so I’m share it with you.
Claude Cowork
You can:
- Chop ingredients
- Reorganize the pantry
- Rewrite a recipe
You cannot:
- Turn on the oven
- Use appliances
- Cook anything that requires heat or power
Claude Code
You can:
- Turn on the oven
- Use the blender
- Cook the recipe
- Adjust while cooking
- Go get missing ingredients
- Run the dishwasher
- Install a new appliance
Hopefully you can see just how much more powerful Claude Code is than Claude Cowork, and how the two actually differ.
How is This Different from ChatGPT (or Claude?)
Hopefully by now you’re starting to see just how different these tools are from ChatGPT, and regular LLMs in general.
Claude Code and Claude Cowork can actually do things, not just answer questions.
The big shift here is agency.
These tools let Claude take action on your behalf, and there’s that word we’ve talked about before: Agent.
Claude Code and Claude Cowork are true examples of AI agents, or agentic AI tools: AI that takes action on your behalf instead of just chatting with you.
Some ways Claude Cowork is being used
So what are people actually doing with this thing? Here are some examples (per ChatGPT). Of note, I only asked for Claude Cowork examples because unless you’re headed down the coding rabbit hole, Cowork will be more relevant to what you do:
- Organizing a messy desktop into folders by project or topic
- Renaming hundreds of files with consistent naming conventions
- Sorting downloads by file type, date, or source
- Turning a folder of PDFs into a single summarized report
- Extracting key data from documents into a spreadsheet
- Converting receipts into an expense tracker
- Creating slide decks from notes, docs, or PDFs
- Drafting reports from raw research files
- Compiling meeting notes into a clean document
- Batch-converting file formats
- Cleaning up shared team folders
- Creating tables of contents from document sets
- Pulling highlights and quotes into one document
- Preparing deliverables from scattered inputs
Safety
If you read that list and you’re like, whoa whoa whoa, there are guardrails in place for both Claude Code and Claude Cowork:
- Claude requires explicit permission before permanently deleting files.
- You control which folders it can access.
- Claude will ask before taking significant actions.
That said, Claude can take destructive actions if you explicitly instruct it to. With great power comes great responsibility. Or something like that.
Important Things to Know: Usage Limits
If you’re used to ChatGPT Plus where you can basically chat all day, Cowork works differently.
You can choose which Claude model you use: Opus, Sonnet, or Haiku, but cowork is extremely compute-intensive. A single task can burn through as many tokens as dozens of regular chat conversations.
Pro subscribers at $20/month will hit usage limits much faster when using Cowork or Code as compared to regular Claude chat.
Max subscribers at $100–200/month get roughly 5x to 20x more usage.
Once you hit your limit, Anthropic offers extra usage on a pay-as-you-go basis at API rates.
Your usage allocation does reset monthly, but if you’re running Cowork regularly, the Pro plan may not give you enough runway.
Translation: Don’t assume you can run Cowork all day like ChatGPT. A few complex tasks can eat your entire month’s allocation.
My Take
If it sounds interesting, give it a try. We’re all about curiosity here.
Yes, you need a paid plan, but $20/month is enough to get started.
Again, I haven’t tried either yet (though Claude Code is likely in my near future), and I made this episode simply for information and awareness, not recommendations or reviews.
For more info, check out Anthropic’s blog post about Claude Cowork.
How I Used ChatGPT This Week
Each episode, I share a quick example of how I used ChatGPT that week.
This week, I used ChatGPT to help me create an LLM-portable memory document that I can upload to any LLM so it “knows me” and I can hit the ground running.
I’m not leaving ChatGPT anytime soon, but I stay thinking about worst-case scenarios and I always want to be prepared. Better to be safe than to be sorry
I’ll be breaking this down in next week’s episode, so stay tuned!
Da Wrap-up
Claude Code and Claude Cowork mark a shift from AI that talks to AI that acts. What exactly this means for…well…everything, time will tell. In the meantime, I think the best thing we can do is get curious and stay informed.
As always, endlessly appreciative for you and your curiosity.
Catch you next Thursday.
Maestro out.