Rahul Gupta

Claude Code is Magic

June 24, 2025

When I first saw people I trust rave about Claude Code, I couldn't understand how anyone would switch from an IDE to a terminal-based agent. Why would I leave the comfort of my AI-enabled IDE? Cursor was already a much better dev experience than an IDE without AI.

But when I had some free time, I decided to give Claude Code a shot, and it immediately clicked. The terminal is actually a better interface for agents than an IDE. Why? In an IDE, you're constantly managing the agent, watching it code, interrupting, making micro-corrections. But when the agent is good enough, like Claude Code, you actually want to give it more control and step back. The "relationship" shifts from micromanagement to delegation.

Claude Code feels like working with a mid-level to senior software engineer. I'll admit I haven't tried the new Cursor agent, but the one that lived in the sidebar felt like a junior developer. For those unfamiliar with software engineering, it's the difference between working with an experienced colleague versus managing an intern. Not only do you not have to manage Claude Code as much, but you can brainstorm and plan with it. And a terminal provides a better UX for that kind of collaboration. I don't want to watch a senior developer code. I'm just going to review their designs and code, and that's exactly how I use Claude Code. It's liberating to be able to give it a task, walk away, and then review the results. That's more or less what it's like working with a senior developer.

Thinking of Claude Code as a senior developer extends beyond just coding. You can have it manage and prioritize tasks, write up systems design docs or READMEs, and it will use markdown to manage a task list or write a document, which it will then commit and push. Since it feels like working with a senior developer, you don't need to limit yourself to one task or project. I now have multiple Claude Code agents running in parallel on different projects. I prompt one agent to tackle a task while reviewing the output of another, then go back and forth. This is incredible—it's the first time I feel like my productivity has truly 10x'd.

Cursor was amazing, but prior to their most recent release, I was running into issues and felt like it was making too many mistakes, really struggling with complexity. Since switching to Claude Code, I've cancelled my Cursor and v0 memberships and upgraded my Anthropic plan to the highest tier. It's totally worth it.

I've completely switched my dev workflow for Kalen, an email based accountability AI, fully to Claude Code. Most recently, I'm working on a project that allows users to share and convert playlists across streaming platforms like Spotify and Apple Music. I haven't written a single line of code myself. It's fully "vibe coded," but not in a reckless way. I still feel like I have a hand in the architecture and design. Unlike Cursor, I feel like I can trust Claude Code to go build something robust that works after just one or two attempts.

I'm still new to using Claude Code, but here are some of the ways I've found it to be valuable:

  • Building something from scratch: Starting a new project with minimal guidance
  • Test driven development: Write tests, write logic, test, fix logic to pass tests, rerun tests
  • Systems design: Let it come up with an architecture plan that you can review and tweak
  • Refactoring: It's really good at taking some logic and refactoring so it's more readable
  • Product and task management: Planning features, breaking down work, and organizing development workflows

This barely scratches the surface, and I haven't even properly explored MCP yet. I'll plan on writing another post on that.

If you want to dive deeper, here are resources with actual prompts, patterns, and examples for using Claude Code:


Here's a remix I made for getting this far: