How I built this site with AI

2026-06-02

This very site you're reading was built with the help of AI, and I wanted to share how the process went.

The starting point

I had an old personal site on Express and wanted something more modern: a bilingual site, with a small CMS to write without touching code, deployed on my own homelab.

AI as a copilot

I used Claude as a development copilot. It wasn't "ask it to do everything", but rather steering it: deciding the architecture, reviewing every change and checking it worked. Some things it helped with:

Architecture and code in Next.js 16, with a custom CMS (content in a database, rendered with MDX in Spanish and English).
Authentication with Auth.js and mandatory 2FA (TOTP).
My journey: pulling my real experience from LinkedIn into the journey section.
Design: a nature-inspired theme with a relaxed but readable typeface.
A working contact form.

From code to deployment

My favourite part: the site deploys itself through a GitOps flow.

Push to GitHub → GitHub Actions builds the Docker image.
The image is published to GHCR.
ArgoCD picks up the change and deploys it to my Kubernetes (K3s) homelab.

All on albertoar.com, with a SQLite database on a persistent volume and secrets managed with Vault.

What I take away from working like this

AI speeds things up massively, but judgement stays human: deciding what to build, reviewing what it generates and understanding every piece. Used as a copilot — not autopilot — it's an incredible tool.

Want to talk AI, backend or deployment? Reach out from the home page.