Back to blog
4 min read

I Rebuilt My Personal Site with Claude Code: Here's What I Learned

AIClaude CodeCodexGeminiWeb DevelopmentLearning

TL;DR

  • The AI productivity gap is real and compounds fast
  • You're still an engineer - you've shifted from artisan to orchestrator
  • Claude Code's autonomy surprised me - it understood the full picture
  • Legacy code and messy codebases still need heavy human intervention

Do I really deserve to call myself a software engineer if AI writes my code?

That question kept nagging at me. I've been a software engineer for a while, but recently I made a decision: it's time to get serious about AI.

Here's the thing - I felt like I was falling behind. The AI train was moving fast, and at work, I'd seen the split clearly. Some people use AI for nearly 100% of their code. Others barely touch it. The productivity gap is real, and it compounds quickly.

But learning to use AI came with mental hurdles. That question above? It has a name: the "Software Engineering Identity Crisis." When "doing" gets replaced by "directing," it feels like the soul of the craft has vanished.

This isn't new though. Technical experts have felt this way before, and they adapted.

Think of it like this: a master mason spends 20 years learning to lay bricks perfectly. Then a machine appears that lays 1,000 bricks an hour. The skill shifts from laying the brick to designing the blueprint. AI can lay bricks, but it doesn't know if the house will fall down.

Why You're Still an Engineer

The word "engineer" comes from Latin ingeniator, meaning "one who devises." Anyone can type "make me an app." Only an engineer can:

  • Define constraints: knowing what to ask is 90% of the battle
  • Audit output: AI lies confidently, and you're the Validator-in-Chief
  • Integrate the mess: real code lives in messy, human systems AI doesn't fully grasp

So that's what I'm doing. Really learning how to work with agents, LLMs, and the tools that are changing how we build software. This site is my starting point, and this post is the first of many where I'll share what I learn.

Why I Started Here

I needed to revamp my personal site anyway. Instead of doing it the traditional way, I decided to build the entire thing with Claude Code, Anthropic's CLI tool that gives Claude full context of your codebase.

The goal wasn't just to ship a website. It was to learn how to work with AI effectively.

What Surprised Me

I've used ChatGPT, Gemini, and older Claude models before. Helpful, but slow. I spent a lot of time correcting mistakes, providing context, and handholding.

Claude Code felt different.

The autonomy caught me off guard. Instead of generating snippets I had to stitch together, it understood the full picture: the file structure, the dependencies, the patterns I was using. I did less correcting and more directing.

One example: the hover preview system on this site. When you hover over a project card, you see a live preview. That feature - the component architecture, the animation timing, the lazy loading - came together in about 15 minutes. Scaffolding that myself? Probably a couple hours minimum.

Where It Struggled

Not everything was smooth. I tried modernizing a portfolio project I built six years ago - a codebase I'd never touched since.

That was rough.

The legacy code, outdated dependencies, and unconventional patterns threw Claude off. One painful example: Claude confidently told me the package modernization was done. The site didn't load. I asked it to recheck the imports and versioning. It "fixed" them. Still broken. This happened several times before everything actually worked.

AI isn't magic. Context matters, and messy codebases are still messy.

I'm actively figuring out how to reduce these correction cycles - better prompting, cleaner project structures, knowing when to just do it manually.

What's Next

This blog is where I'll share what I learn. Expect posts about:

  • AI-powered tools and apps I'm building
  • Automation workflows that save time
  • Experiments with different AI tools and techniques
  • Practical tips from real usage, not theory

If you're also figuring out how to work with AI - not just use it, but get good at it - follow along. I'll share what works, what doesn't, and the projects I build along the way.

The tools are here. The question is whether you'll learn to use them, or watch others pull ahead.

githublinkedinx