From Zero to Code Hero? My First Weeks Using Claude Code

I have no background in coding. No CS degree, no bootcamp, not even a dusty copy of Python for Dummies on my shelf. So when I started seeing a surge of references to Claude Code, vibe coding, and the idea that AI tools were fundamentally changing what it meant to build software, I was curious in the way most non-technical people are: intrigued, slightly skeptical, and mostly convinced it probably wasn't for me.

Then I decided to just try it.

Step One: Getting My Bearings (Read: Surviving the Terminal)

My journey started with installing Claude Code, which, for someone who had never so much as opened Terminal on their MacBook, was its own adventure. There's something genuinely intimidating about staring at a command line for the first time. It feels like you've opened a door to the internals of your machine and one wrong keystroke might break everything.

It doesn't, of course. But that fear is real, and getting past it took a little time.

I downloaded VS Code as my editor, set up Python, and got Claude Code running as an extension. Each of those steps involved looking things up, reading documentation I only half understood, and occasionally retracing my steps when something didn't work. But piece by piece, it came together.

Step Two: Understanding GitHub

Once I had my environment set up, I spent a few days getting familiar with GitHub. If the terminal felt foreign, GitHub felt like arriving in a country where you don't speak the language and can't read the street signs.

The terminology alone is a lot: repositories, branches, commits, push, pull, merge, pull requests. And the underlying mental model—understanding why you'd work on a separate branch instead of your main codebase, understanding how version history works and why it matters, understanding the pipeline from local changes to a shared repository—takes some time to click into place.

It did eventually click. And once it did, I could see why developers treat it as second nature. It's an elegant system once you understand what it's trying to do.

Step Three: Building Something Real

With the foundation in place and a $20/month Claude Code subscription active, I wanted to test what this was actually capable of. The best way to do that, I figured, was to throw a real problem at it.

At work, there's a task our team does manually that takes hours: tracking how peer firms to Capital Group mark private investments over time using NPORT filings on the SEC EDGAR website. This involves sifting through hundreds of filings, extracting specific data points across countless securities, manually entering everything into Excel, and producing charts and tables for committee reviews. It's tedious, error-prone, and time-consuming.

It's also, I suspected, exactly the kind of thing that should be automatable.

So I described the problem to Claude Code. I explained what the data was, where it lived, what I needed extracted, and how I wanted it structured. And then it just… started building.

What happened next genuinely surprised me. Within an hour, I had a working tool. It pulled from the SEC EDGAR API, parsed files with hundreds of thousands of lines of text, identified the relevant data points, charted them over time, and organized everything into clean, structured tables. There was back and forth, some iteration on features, some testing and tweaking. But the core of it was there almost immediately.

A task that used to take hours of manual work was now running in minutes.

So, Where Does That Leave Me?

I'm not a developer. I still don't know how to code in any traditional sense. But what I found over these first few weeks is that the barrier to building something useful has dropped significantly. You don't need to know the syntax. You need to understand the problem, communicate it clearly, and be willing to iterate.

The learning curve on the tooling is real, and I won't pretend otherwise. Getting comfortable with the terminal, understanding Git, navigating a development environment—none of that is trivial if you're starting from zero. But it's learnable, and the payoff once you get there is pretty remarkable.

I'm curious to see where this goes. More projects are already in the queue.