Building ayoubabed.xyz with AI

The website is the work

I built ayoubabed.xyz as the website I can share when someone wants to know who I am, see my work, find my channels, or view photographs from an event. I own the direction and review the result while AI writes and revises the implementation.

I used Claude and Codex during development. I now use Codex for the ongoing work. The important part of my role is choosing what the site needs, supplying context, reviewing the output, and checking that the result works for the people using it.

A portfolio and a working gallery system

The site uses Astro for its pages and content, Tailwind CSS for styling, and shadcn/ui components for shared interface controls. Portfolio entries are written in MDX so each project can have its own story and media while sharing a consistent layout.

The site also includes server-side gallery services through Cloudflare Workers, with event photographs stored in R2. It is more than a static collection of pages: visitors can browse photos, keep favorites in their browser, and download individual images or a collection.

I also use a personal Android companion app to upload photographs to the website galleries. It supports my delivery workflow; the website remains the place visitors go to see the work.

Directing and checking the AI

I connect documentation tools such as Astro MCP and shadcn MCP so the agent can consult references while working. Other tools, including context-mode and Caveman, help manage long development sessions.

The work is iterative. I explain the visitor’s needs, review the interface, ask for changes, and check the implementation. A concrete visual reference helps establish the direction; inspecting real pages and interactions helps keep subsequent changes consistent.

Owning photo delivery

The gallery system gives me control over how photographs are presented and delivered. The Sanad wedding gallery is one example of that workflow.

As of October 2026, my Cloudflare usage remains within the free allowances. That describes the site’s current usage rather than a promise that hosting every future feature or volume of photographs will always be free.

Continuing the work

The recurring challenge is keeping new features consistent with what is already there. I review typography, spacing, interactions, and content as the site grows, alongside checks for errors and regressions.

I apply the same AI-directed development approach to yahalausa.net, which I continue to update for Ya Hala with Haithum. Its audience and publishing needs shape that site separately from my own portfolio.

The source code is public on GitHub. I also share what I learn on my AlphaBravoMedia channel.