What's the point of using Claude's new Cowork?
Why fear the terminal? And why only use Claude Code for code?
The usual post on a new AI release is about how it’s lifechanging and you should quickly hop on it. I'm kind of rebellious. So here's one that goes the other way.
Some background. I took a while to get started on Claude Code a while back. Even though I’m quite used to coding, using an LLM this way takes some getting used to. But once I started, there was no going back. And my workflow has evolved.
Anyway, back to what sparked this post.
Anthropic just launched Cowork. Basically Claude Code with a friendly GUI. It allows one to use Claude Code for writing and anything else you can imagine. For now it’s MAX-only and in research preview mode.
But why waste money, or wait for it to be available?
Everything Cowork does, Claude Code does. And more. And with Claude helping, the terminal really isn't that scary.
I use Claude Code for more than coding. And I pair it with the free Obsidian app.
Why? Obsidian is a powerful note app that you can simply add your folder (or folders) to. No database. No proprietary format. Just folders and markdown files (which are just text files with a .md extension and fancy formatting).
My workflow. See what works for you. I did it this way as my memory is no longer great at 47.
1️⃣ The Views. A folder with my projects, areas, resources and archives (the famous PARA system) visible from both VS Code and Obsidian. Claude Code in VS Code for research and bouncing ideas. Obsidian as my editor for the final output. Same files, two lenses.
2️⃣ The Memories. Every key folder has three key files. CLAUDE.md is Claude’s instructions. How I want it to behave for this project, what context matters, what to ignore. PIN.md is the project state. What’s decided, where we are, what’s next. LESSONS.md captures what I learned for future projects.
3️⃣ The Workflows. Choose your own labels. But for me, this is my usual flow - 0-PLAN.md, 1-RESEARCH.md, 2-SYNTHESIS.md. The names and numbers vary by project. But what matters is that there’s a flow, and it works for you. My experience is that the quality is much better when you work with Claude on it step by step.
4️⃣ The Tools. Use MCPs to add tools like web search (Brave), academic papers (arXiv), or persistent memory. Just ask Claude for the instructions.
5️⃣ And how it comes together. Just talk to Claude Code in the terminal - “Read CLAUDE.md for context, refer to this [folder or file] then help me flesh out 0-PLAN.md.” “Search for papers on [topic] and save your synthesis to 1-RESEARCH.md.” “Update PIN.md with where we are.” Next session: “Read PIN.md and continue drafting 2-SYNTHESIS.md.” When done: “What should I add to LESSONS.md?” Then I switch to Obsidian to write the final output, the way I like it (I actually enjoy writing).
So, don't fear the terminal. Try it. It will grow on you.
Getting started with Claude Code (for the uninitiated)
1️⃣ Download VS Code and install it.
2️⃣Add the folder you are working on (see my workflow above)
3️⃣Open Terminal → run npm install -g @anthropic-ai/claude-code
4️⃣Type claude, then /login
5️⃣ Ask Claude for help adding MCPs.
#ClaudeCode #AI #GenAI #AIinWork