Sign in Subscribe
AI

Claude Cowork Is the Digital Assistant I Always Wanted

Anthropic's new Cowork feature transforms Claude from a chatbot into an autonomous colleague that actually gets work done. After years of waiting for AI assistants that do more than chat, this might be the real thing.

Johan Kölhi

Johan Kölhi

— 4 min read
Laptop displaying an AI workspace interface with autonomous digital assistant capabilities

I've been waiting years for this. An AI that doesn't just answer questions but actually does the work. Last week, Anthropic released Claude Cowork, and it feels different from anything that came before.

From Chatbot to Colleague

Every digital assistant I've tried has shared the same fundamental limitation. You type a question. It types an answer. You copy-paste the output somewhere useful. Repeat endlessly.

Cowork breaks that pattern. You point Claude at a folder on your computer, describe what you want done, and walk away. It reads your files. It creates new ones. It organizes, analyzes, and produces actual output. Anthropic describes it as "much less like a back-and-forth and much more like leaving messages for a coworker."

That mental model shift matters. With a chatbot, you're constantly context-switching between asking and doing. With Cowork, you delegate and move on to something else. You can queue multiple tasks and let Claude work through them independently.

How It Actually Works

The technical setup is straightforward. Cowork runs inside a sandboxed environment on your Mac, using Apple's Virtualization Framework. You designate specific folders Claude can access. Within those boundaries, Claude plans and executes tasks autonomously.

Want your cluttered downloads folder sorted and renamed intelligently? Done. Need expense spreadsheets generated from a pile of receipt screenshots? Handled. Have scattered notes that should become a first draft of a report? Claude reads everything and produces something you can actually use.

The Gmail connector lets Claude send emails. The Chrome integration enables browser automation. This goes beyond file organization. Claude becomes a capable worker with access to real tools.

Built by AI in 10 Days

Here's the thing that makes developers' heads spin. Anthropic built Cowork in approximately 10 days. Claude Code itself wrote the entire codebase, according to engineer Boris Cherny.

This isn't just a parlor trick. The achievement shows where software development is heading. When AI can build AI tools this fast, the bottleneck shifts. The constraint becomes having good ideas about what to build.

Anthropic ate their own dogfood. The result shipped to paying customers in under two weeks.

The Unexpected Origin Story

Cowork exists because users kept surprising Anthropic. Claude Code launched as a developer tool. People started using it for everything else.

Cherny explained: "Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven."

That's a wild list. People recognized Claude Code's power. Executing commands and manipulating files works for more than programming. They repurposed a developer tool for life administration.

Cowork packages that capability in a friendlier interface for everyone else.

The Safety Question

Anthropic deserves credit for being upfront about the risks. Their documentation warns about destructive actions if instructions are unclear. Prompt injection attacks remain a vulnerability. Malicious content in files could manipulate Claude's behavior.

The built-in safeguards help. Sandboxed filesystem. Permission requests before significant actions. User controls over access. But security researcher Simon Willison raises a fair point. He wrote: "I do not think it is fair to tell regular non-programmer users to watch out for 'suspicious actions that may indicate prompt injection'!"

That tension exists. Cowork's power comes from autonomy. That same autonomy creates risk. Keep sensitive financial documents away from Claude. Monitor actions during critical operations.

During this research preview phase, treat it accordingly.

What This Means for Everyone

Cowork represents something bigger than one product launch. It marks a transition in how AI works.

From passive chatbots to agentic systems. From reactive responses to proactive task completion. From developer tools to everyone tools. From conversation to delegation.

Microsoft has Copilot in Office. Google has Gemini in Workspace. Now Anthropic has Claude working with your local files. Simon Willison predicts: "I would be very surprised if Gemini and OpenAI don't follow suit with their own offerings in this category."

The competition will benefit users. Each company brings different strengths. Local file autonomy is Anthropic's angle.

Current Limitations

Cowork is macOS only for now. Anthropic plans Windows support. You need a Claude Max subscription at $100 to $200 per month. This remains a research preview. Some features don't work yet: Projects, chat sharing, and Memory.

These limitations will shrink over time. What matters is the underlying capability.

The Feeling When Things Change

One reviewer captured it perfectly: "There's a very specific feeling I get when I try a new tool and immediately realise, oh... this is going to change how people work. It's not excitement, exactly. It's more like quiet discomfort mixed with curiosity."

That's honest. Major technology shifts feel uncomfortable. They force adaptation. The discomfort signals something real is happening.

I've tried every AI assistant that's come along. Most felt like incremental improvements on the same chat interface. Cowork feels like something else. An AI that does work instead of describing how you could do it yourself.

That's what I always wanted.

Sources

Read more