TL;DR
I’m a marketer, not a developer.
Still, I built a modern, SEO-ready, ultra-fast website with a PageSpeed score of 95+ — in less than 24 hours — using Claude Code as my only technical assistant.
👉 David Lopes — live, minimal, performant.
No WordPress. No Framer. No Webflow. No bloated themes. Just a clear spec and Claude as my pair programmer.
Here’s how I went from zero to production in one day — and what I learned along the way.
The Problem: From Side Projects to Nowhere
I’ve started dozens of side projects. Most never saw the light of day.
Not because of lack of ideas — but because of execution paralysis.
I don’t code. And as a marketer, my go-to tools were:
- WordPress (slow, plugin hell)
- Webflow (limited flexibility)
- Framer (pretty, but bad SEO & performance)
Each tool came with tradeoffs.
Every time I wanted to go beyond the default, I hit a wall.
So I gave myself a new challenge:
What if I could build a fast, production-ready, custom site in <24 hours… without writing code myself?
Enter Claude Code: My AI Pair Programmer
I didn’t pick a stack. I started with a spec.
Here’s what I told Claude:
“I want to build a responsive, SEO-ready site with 3 static pages, a blog, markdown content, good performance on mobile, hosted on a CDN, and zero CMS.”
Within seconds, Claude replied with:
- Vite + React + Tailwind CSS
- Express backend
- PostgreSQL with Drizzle ORM
- MDX for content
- Hosting on Railway + Cloudflare
- Umami for analytics (no cookies, no GDPR banners)
Not only did it suggest the stack — it scaffolded the full project:
files, folders, components, scripts, deploy setup. Everything.
Why This Stack Worked (Even If I Didn’t Choose It)
I didn’t arrive with strong technical opinions. I had functional constraints.
Claude mapped those into a lightweight, modern stack:
- Frontend: React 18 + Vite (fast builds, tiny bundles)
- Styling: Tailwind CSS (purged to 8KB)
- Routing: Wouter (1.6KB vs 9KB for React Router)
- Content: MDX files (easy to edit, flexible)
- Backend: Express.js
- Database: PostgreSQL via Railway
- ORM: Drizzle (minimal, type-safe)
- Analytics: Umami (privacy-focused, 2KB script)
- Hosting: Railway + Cloudflare CDN
No CMS, no plugin bloat, no black-box framework.
What Claude Actually Did For Me
Claude helped me:
- 🏗 Scaffold the full project (frontend + backend)
- 🎨 Build 40+ UI components (buttons, cards, forms) with accessibility
- 🔍 Set up meta tags, Open Graph, JSON-LD structured data
- 🖼 Optimize Core Web Vitals: hero image preload, font loading strategy
- 📦 Reduce JS bundle from 450KB to 85KB via lazy loading
- 🧪 Debug layout shifts, memory leaks, unused dependencies
- 🛠 Create database schema, migrations, and CRUD routes
I was running 3 Claude sessions in parallel:
- One for frontend (components + layout)
- One for backend/API
- One for SEO and performance optimization
The Results (Metrics After 23h Build Time)
PageSpeed Insights
| Metric | Mobile | Desktop |
|---|---|---|
| Performance | 95 | 98 |
| LCP | 1.1s | ✅ |
| INP | 45ms | ✅ |
| CLS | 0.02 | ✅ |
| FCP | 0.8s | ✅ |
| TTFB | 180ms | ✅ |
Bundle & Build
- Initial JS: 85KB (gzipped)
- Tailwind CSS: 8KB (purged)
- Build time: 12s (Vite magic ✨)
Hosting
- Railway: Git-based deploy, PostgreSQL native, $5/mo
- Cloudflare: Free CDN, cache rules, Brotli compression
- Domain: ~€10/year
Total cost: ~€10/month
What I Loved Most
🔓 Freedom – Full control, no CMS lock-in.
🧠 Learning – Understood the stack thanks to Claude.
🚀 Velocity – From idea to live in a day.
What I’d Improve Next Time
- Claude Credits – The free tier rate limits hit fast. I used the API (cheap and faster).
- Prompt Precision – The better your prompt, the better Claude’s output.
- Multi-Terminal Setup – Requires some juggling, but makes you feel like you have a dev team.
What I Learned
- You don’t need to be a developer to ship like one.
- Start with a clear spec, not a framework.
- Claude is a force multiplier — not a replacement, but a productivity rocket.
- Vite + React + Tailwind + Railway + Cloudflare is a killer stack for solo builders.
What’s Next
- Add i18n (multilingual support)
- Make it a PWA (offline-ready)
- Automate content via AI/Markdown pipelines
- Reuse the stack for landing pages, SEO projects, SaaS MVPs
Final Thoughts
I’ve never shipped this fast, this cleanly, with this much control.
You don’t need to code to build fast.
You just need the right assistant, and the right mindset.
— David Lopes
Top comments (2)
For blogs or flat business sites... go for it.
For marketing, as you did, go for it. Landing pages are fine to make with AI, consistency isn't super important and longterm you can wipe em and build new ones from scratch.
AIs however, are not ready for large scale complex sites with user accounts, ecommerce, or sensitive data.
Not because the AI tools cannot generate them, but someone without any experience doing it will not know how to properly secure it and will leave their system wide open to hackers.
Also, your performance score is now a 93... nothing's probably changed, except you've added more content.
Nothing beats actual dev experience for the long run, but for today, to market quickly and sell quickly... go for it.
For stable, large-scale future work, reach out to a serious developer.
I completely agree with your points. This project was never about replacing developers or pretending AI can handle production-grade complexity (ecommerce, auth, security...). I’m a marketer, and this was a pure marketing/SEO site, no dynamic backend, just content.
The goal was to test how far I could go solo using Claude as a pair programmer, and for a lightweight, flat marketing site… it worked incredibly well. Fast to ship, performant, and gives me total ownership.
You’re right that real dev experience is irreplaceable when it comes to long-term stability, scaling, or security. And yep the 93 score is expected, now that I’ve added more content and components. It’s still within my acceptable range for now, but I’ll keep optimizing.