Beyond Code: The Rise of the AI Agent Coworker

I’ve been on the AI tool merry-go-round for a while now, juggling the likes of ChatGPT, Perplexity, Claude, Gemini, NotebookLM, and Grok. But a few months ago, my work truly transformed with the adoption of Claude Code. Though I’m an engineer by trade, my day-to-day role had shifted away from the keyboard. Claude Code brought me back to my roots in a really delightful way. I estimate I’ve generated over a million lines of code in the last few months since I started using it, often while multi-tasking. It felt less like using a tool and more like having a talented, tireless intern (or three) working for me 24/7.

Last fall we let Claude Code lose with command line access to some networking gear (supervised), to help debug some gnarly local networking problems. It performed like a champ and found a configuration issue that could have taken many hours to track down. It really started me thinking about how a tool like Claude Code could be useful beyond just generating code.

Apparently we weren’t the only ones to recognize the power of more general uses of Claude as an agent. Last week Anthropic introduced Claude Cowork. It is a great and useful tool that foreshadows our future in several ways.

First is its origin story. This is a new product, or at the very least, significantly new functionality. The story is that it was created by four people at Anthropic over the course of a little over a week(!!). It was released on Jan 12, 2026 and it developed this year (since Jan 1, 2026) completely using Claude Code. That shows how quickly we can go from concept to product these days.

Second is what it can do. They call Claude Cowork, Claude Code for the rest of your work. For me the timing of this was perfect. As I establish my own business practice, Cowork is proving to be an indispensable partner. It takes the advantages I found in coding and applies them to a vastly broader set of tasks. It can control my Mac, use my browser, and leverage many other tools. The mental shift is significant: I no longer think about how to make an AI write code for me; I think about the tasks I want to complete.

In just a week, it has:

• Helped me brainstorm and finalize a name for my new business.

• Secured a domain name.

• Researched the necessary paperwork for county registration.

• Created a daily morning newsfeed curated only with the stories I need to track the breakneck speed of AI change.

• Set up and began syndicating this very blog.

• Even helped to create and debug a home kiosk system for distributing relevant infotainment around the house.

While some of these tasks involved a bit of code, many didn’t. The key takeaway—the feature that is hard to fully relate until you use it—is that the output of Cowork was not code, it was completed tasks.

It’s not a seamless experience yet; frustrating loops and moments where I have to re-explain the agent’s capabilities still happen. Nevertheless, it is an amazing glimpse into a future that is fast approaching. The era of the productive, autonomous AI Agent is upon us.

In an upcoming blog, I’ll dive into some of the powerful implications these changes have for our management structures, product development, and the very definition of a developer. Stay tuned!

What do you think? I’d love to hear your thoughts.