A Journey Down the Claude Code Rabbit Hole

I’ve been using GitHub Copilot for a while now at Under Armour, mostly for the usual stuff — refactoring, writing unit tests, and sometimes laying down the groundwork for new features. It’s become second nature in my workflow. I can start typing and Copilot quietly fills in the gaps, helping me move faster and stay in the zone. It’s not perfect, but when you’ve worked with code long enough, you learn to appreciate tools that remove friction instead of adding it.

Recently, though, a few people on my team started talking about Claude Code. I’ll be honest — I ignored it at first. I wasn’t too excited about a CLI-based workflow when I already had something that fit neatly inside my IDE. Copilot just feels more native — it understands my open files, picks up on naming conventions, and even leverages MCP (Model Context Protocol) servers in VS Code to provide richer suggestions. But curiosity got the best of me, and once I discovered a few community plugins that allow Claude Code to access the same VS Code context (open files, project data, etc.), I decided to give it a shot.

The Technical Differences That Matter

After a few days of switching between the two, it became clear that Copilot and Claude Code serve two very different roles — and honestly, they complement each other more than compete.

GitHub Copilot runs on OpenAI’s Codex and GPT-4-turbo models and is designed for micro-level assistance — the fast, inline kind of help that keeps you moving. It thrives inside your IDE, predicting your next few lines, suggesting syntax, or completing a function before you’ve finished typing. It’s pattern-based, responsive, and context-aware on a local scale. Copilot feels like an extra set of hands that works in real time — a partner who’s quietly there to handle the boring stuff while you think about architecture and logic.

Claude Code, on the other hand, uses Anthropic’s Claude 3.5 Sonnet model and takes a completely different approach. It’s built for macro-level collaboration. Instead of working at the speed of your typing, it works at the scale of your project. Claude can handle entire repositories in context — literally hundreds of thousands of tokens — which means it can read and reason over your full codebase. You can ask it to plan, refactor, or even coordinate multi-file changes. Through its agentic CLI workflow, it can manage branches, commits, and diffs automatically.

In practice, Copilot feels like your pair programmer, while Claude feels like your tech lead — the one you hand a problem to and come back later to see results. Copilot helps you code faster. Claude helps your team build smarter.

A Different Way to Work

Claude Code requires a different mindset. You don’t treat it like a fancy autocomplete; you treat it like a creative collaborator. You step into more of a director role — the person who defines the architectural goals, outlines priorities, and assigns the work.

I read an article recently that compared using Claude Code to being a surgeon entering an operating room: everything is prepped, instruments are laid out, and all you have to do is perform. The agents — your digital assistants — handle setup, cleanup, and background coordination. You just focus on the core vision.

The first time I spun up multiple Claude agents running in parallel, each cutting its own feature branch and working on independent tasks, it felt like I was provisioning a virtual engineering team. One agent was updating a design system package, another was writing Cypress tests, and a third was cleaning up some API routes. It was wild — like running your own small dev shop inside your terminal.

What Surprised Me the Most

After just a few days of using Claude Code, I noticed something surprising: I was tapping out my daily token budget — every single day. That’s how much I found myself relying on it. The more I used it, the more I realized how powerful it was to have agents that don’t just react but actually collaborate.

It didn’t take long before I caught myself thinking: I’d have a hard time going back to anything else. Copilot still has its place — it’s fast, frictionless, and great for local coding — but Claude Code has changed how I think about building software. It’s not just a tool anymore; it feels like a team.