For a very long time, my perception of the terminal was confined to what a movie or TV show showed you. A multi-monitor setup with terminal windows full of green text scrolling too fast to read, with some hoodie-clad guy muttering "I'm in." And while I definitely wanted to be, I had no real idea what any of it actually did. In fact, despite the fact that it looked "cool" to my younger self, it also appeared intimidating enough to have stuck with me for years.

Funnily enough, it wasn't majoring in computer science or writing for XDA and hearing my colleagues talk about running everything through the terminal that broke the mold. It was Claude Code — and I think that's the best thing these agentic AI tools have done: bring the terminal to the rest of us.

Plain English is the only command I need now

Now, it isn't like installing Claude Code was the first time I opened the terminal. I very vividly remember opening it when the Run button refused to show up in VS Code for a Java file, and some forum told me I had to check whether the JDK was even installed. I also remember opening the terminal when testing out a string of commands to customize the Mac Dock for an article I was writing for XDA's sister site, MUO. That said, I never willingly opened it for anything. Every one of the times I did open it, I was following instructions someone else had written, copy-pasting commands I didn't fully understand, and praying they'd work. Despite hearing endlessly about how powerful it was and how you could rip through tasks in seconds that would take ages clicking around a GUI, I stayed away because I knew that power wasn't really available to me. It was locked behind a wall of exact syntax and commands I'd have to go look up every single time, and one wrong flag away from breaking something I couldn't fix.

This is precisely what Claude Code changed for me. All I really need to do is cd into the folder I want to work in and type claude. That's all I really need to do when it comes to commands. From there, I can just tell it what I want in plain English. For instance, using the same macOS Dock example, I can simply tell it I want to add a few spacers and bump up the icon magnification, and it'll work out the exact defaults write commands. Beyond surfacing exactly what command I need (which any AI chatbot can), what makes agentic terminal-based tools like Claude Code special is that they can run these commands for me too. And if they don't work as intended, Claude Code can also read the error it gets back, figure out what went wrong, and try a different approach — without me having to copy the error into Google and start the whole hunt over again.

For instance, I should be embarrassed admitting this, but I'm still learning my way around GitHub. Given I'm not a professional developer by any means, I'm constantly looking up how to clone a repository, commit changes, and push them without accidentally turning my whole project into a tangled mess. With Claude Code, I don't have to. I can paste in a GitHub link and say "clone this and get it running," and it works through the whole chain for me. When I want to save my own work, I can just say "commit this and push it," and it handles the staging, the commit message, and the push without me having to remember the right order of git commands or panic that I'm about to overwrite something I can't get back.

I'm learning more from watching it than I expected to

You know how they say the best way to learn is by doing? Turns out watching works surprisingly well too, at least when what you're watching is a tool narrating its own every move. Depending on the permission mode you've set Claude Code to, it'll pause and ask before running anything. It'll show you the exact command it wants to execute and wait for you to manually approve it. While the permission prompt includes a description of the command it'd like to run, you can also ask it to explain what a particular flag or piece of syntax does before you hit the Approve button.

It'll then break it down right there, and you'll not need to do so much as switch to another tab and Google it. With time, I found myself picking up different terminal commands without ever sitting down to memorize them. This is similar to how vibe-coding has been helping me learn how to actually code too. I like comparing it to how learning a language works! When you begin conversing with a native speaker, you don't learn by drilling grammar rules in isolation. Instead, you pick things up in context. Hearing a word used correctly enough times that it just sticks! Watching Claude Code work is the same!

My favorite uses have nothing to do with coding

I've repeatedly said that the biggest benefit of vibe-coding is the fact that it's removed the barrier to entry non-technical people once had. If you had an idea for a tool or app you wanted to build before the AI era, you really only had two options: learn how to code or pay someone to build the app for you. There wasn't really a middle ground, and each option came with its own set of consequences. With Claude Code and similar tools now, you can go from a rough idea to a full-fledged working prototype within minutes. And while that's certainly the biggest headline, it's not the only thing this kind of access unlocks.

Once a tool like Claude Code lives in your terminal, it isn't limited to building apps. It can reach into your actual computer, read your files, and run commands on your behalf, which means it can take on all sorts of everyday tasks that have nothing to do with writing software. Think organizing a Downloads folder that's spiraled out of control, bulk-renaming a few hundred messily named files, converting a pile of images from one format to another, or digging through a folder to find and clear out duplicates.

I'm not intimidated by the terminal anymore

Anthropic recognized the use cases above and decided to build Claude Cowork for this exact reason. It brings Claude Code's agentic capabilities to non-developers and strips the terminal out of the equation entirely. However, given that I use it for both coding and non-coding use cases, I prefer keeping everything in one place. Claude Code in the terminal handles both perfectly well for me, so I've never felt the need to switch over.