3 Projects, 3 different tech stacks, 1 month of Claude Code

Claude Code Opens doors to unknown tech stacks

I didn’t think GenAI would catch up

I didn’t start as a GenAI-first engineer.
For a long time, I genuinely believed models wouldn’t reach a point where they meaningfully changed how real software is built.

GitHub Copilot was the first shift. Not because it wrote code, but because it understood context well enough to implement things that actually worked. That was my first real aha moment.

Claude Code came later. I was working on a client project, people kept talking about it, and I decided to try it for a month. I didn’t expect it to reshape how I approach building software.

Not just coding. Everything around it.

The real cost of unfamiliar tech stacks

New tech stacks are not hard because of syntax.
They are hard because of jargon, mental models, and understanding how the ecosystem fits together.

Before Claude Code, starting a project in an unfamiliar stack meant days of documentation, conventions, architecture decisions, and trial runs before writing anything meaningful.

Claude Code collapsed that entry cost, awakening my inner builder.

Project one: a VS Code extension

The first project was a VS Code extension built on top of the E2B ecosystem.

Earlier, just setting up a VS Code plugin would have meant reading multiple documentation pages to understand project structure, extension APIs, and JavaScript best practices. I am primarily a Python engineer, and JavaScript conventions are not something I actively keep up with.

With Claude Code, the initial setup was a breeze. I prompted it to follow modern JavaScript and VS Code extension best practices, and it scaffolded a clean, working project without friction. I could focus on what I wanted to build instead of how to begin.

Project two: a native Swift PDF reader

The second project was a native PDF reader built using Swift and PDFKit. I was not even sure Claude Code would handle Swift well.

It did. And surprisingly well.

Most of the core functionality came together quickly. Along the way, I still had to deep dive into PDFKit to understand some nuances and edge cases, but Claude Code covered the majority of the groundwork.

I did not become a Swift expert, but I learned enough to understand, modify, and maintain what I built. That difference matters.

Project three: replacing GitBook with Astro

The third project was migrating my website from GitBook to Astro.

GitBook increasingly paywalls essential features, and I wanted more control. I also didn’t want a documentation-style site. Astro’s Starlight theme was close, but it was not designed for the kind of blog I wanted.

A few well-directed prompts later, I had a GitBook-like experience built on Astro, tailored to my needs. Within two days, the site was live. That speed was unthinkable earlier.

If you start Claude Code today, try this

If you are new to Claude Code, don’t start with clever prompts. Start with intent.

  1. Start with the problem, not the stack
  2. Ask for a plan before code
  3. Constrain early
  4. Let it scaffold aggressively
  5. Intervene when judgment matters
  6. Switch it into teacher mode later

Claude Code amplifies clarity.
Vague intent leads to vague output.
Precise thinking leads to unfair speed.

The takeaway

If you are a techie in 2026, spend the twenty dollars and get a Claude Code subscription.

Use it for a month. Not for dependency, but to understand the amount of leverage this tool provides.