Chef Nam Catering Website
Custom website built with modern stack—100% coded with Claude, zero prior HTML/CSS knowledge
Overview
Built a complete custom website for a local catering business using Astro and Sanity CMS, entirely through AI-collaborative development
Problem
Client needed a professional website to showcase their catering services. Template solutions didn't capture the brand personality, and I had no frontend development experience—didn't know HTML or CSS.
Constraints
- No HTML/CSS knowledge at project start
- Client needed full content control (CMS required)
- Performance-critical for local SEO
- Budget-conscious—couldn't justify traditional agency costs
Approach
Treated this as a test of AI-collaborative development for unfamiliar domains. Used Claude Code as my development partner—not to generate code blindly, but to work through architecture decisions, debug layout issues, and understand what I was building. Learned to work with my machine differently: command line, local files, deployment pipelines.
Key Decisions
Build custom with Astro rather than use WordPress or Squarespace
Modern stack = better performance = better SEO. Full control over design and functionality. Learning investment pays forward to future projects.
- WordPress with premium theme
- Squarespace or Wix
- Hire a web developer
Use Sanity CMS for content management
Headless CMS gives client full control over content without touching code. Generous free tier. Excellent developer experience for building custom schemas.
- Markdown files (no CMS)
- WordPress as headless CMS
- Contentful
Deploy on Cloudflare Pages
Free tier handles traffic easily. Global CDN for fast load times. Simple deployment from Git. Already using Cloudflare for DNS.
- Netlify
- Vercel
- Traditional hosting
Tech Stack
- Astro
- Sanity CMS
- Tailwind CSS
- Cloudflare Pages
- Claude Code
Result & Impact
- Near-perfect technical SEOLighthouse Score
- 3 weeks from zero knowledgeTime to Build
- $0/month (free tier)Ongoing Cost
Client has full control over content. Site loads fast, ranks well locally. Proved that AI collaboration can bridge significant skill gaps—not by replacing learning, but by accelerating it.
Learnings
- The 'how do I deploy this?' friction is where most side projects die. Claude helped me push through.
- Three sites later, I'm not a frontend developer—but I'm no longer someone who outsources anything technical
- Working with AI on unfamiliar domains requires learning to ask better questions, not just accepting answers
The Skill Acquisition Pattern
This project revealed a pattern I’ve seen repeat across domains:
Capability loading (AI gives you access to skills) is the easy part.
Sparring (working through problems, building judgment) is where actual learning happens.
I could have asked Claude to generate an entire website. Instead, I used it to understand what I was building—why this CSS property does that, why this component structure makes sense, how deployment actually works.
The result: I can now build and maintain websites. Not because AI did it for me, but because AI collaboration accelerated the learning I needed to do anyway.