Claude Code

“Documentation is no longer a record of what was built; it is the infrastructure through which we build.” — Matt Gierhart

I’ve been using Claude Code within my Obsidian vault since late summer 2025. Unlike chatbots that start fresh every conversation, the context persists: files stay, and knowledge builds. It’s the difference between talking to someone with amnesia and someone who actually remembers what you’ve been working on. I use it for research, writing, project management, and building small tools. It became the interface to my work. What fascinates me is that this hands-on experience feels fundamentally different from the abstract AI discourse. The hype around “agents” loses its mystique when they’re just a tool in your workflow. “AI is whatever doesn’t work yet. When it does, it’s just software.” This is what it feels like when AI turns into software.

Claude Code is a terminal-based tool by Anthropic that works directly with local files and folders. Unlike browser-based chat interfaces, it can read, edit, and create files. It understands project context and executes multi-step tasks autonomously. Originally built for developers, more and more knowledge workers are discovering it for non-technical tasks: research, writing, data analysis, and personal knowledge management.

The strangest outcome for me: I just talk to Claude about how to solve problems, and then I’m developing code all of a sudden, and now I’m publishing stuff on GitHub. Did not see that coming.

Why It Matters

Getting Started


In January 2026, Anthropic released a preview of Claude Cowork, essentially Claude Code packaged for non-technical users. It brings the persistent, task-delegation capabilities to a graphical interface. Katie Parrott explains it well: Claude Cowork Is Claude Code for the Rest of Us. For a practical beginner’s guide, see Arman Hezarkhani’s thread.

Of course, people are also starting to use Claude Code for foresight work.

Notes mentioning this note


Here are all the notes in this garden, along with their links, visualized as a graph.