I build websites and web apps for organisations that need real software live—donation platforms, booking systems, multilingual storefronts. AI coding agents carry the build volume; that is the economic mechanism, disclosed on every project. Automated tests, staged deployments and one signature on the work are why it holds up.
The familiar problem
Ask a traditional agency for a serious website and the quote comes back in months and layers: discovery workshops, account managers, a team you never quite meet. A large share of the fee pays for coordination, not code.
The alternative ships fast but thin. Template builders produce a site—but no automated tests, no staging environment, no accessibility work, and nothing to lean on the day you need payments, four languages or a working back office.
The gap between the two is where this practice operates: agency-grade engineering—tested, multilingual, properly deployed—from one accountable builder, at the speed AI coding agents now make possible.
See the prototype-first processThe El-Sultan Web approach
Claude Code, OpenAI Codex and Google Antigravity produce and refactor code at a pace no small team matches—which is how one person delivers donation platforms, ticketing systems and multilingual back offices on short timelines. The method is disclosed on every project, because it is the honest reason the economics work.
Nothing the agents write ships on trust. Every build passes through Playwright and Vitest suites, staged dev–test–production deployments, WCAG accessibility work and GDPR-conscious analytics—and a single person scopes it, reviews it, tests it and answers for it after launch.
Services
Whether it is a launch build, a platform build, a website rebuild, campaign landing pages or ongoing care, every engagement runs the same way: a working prototype early, production discipline before launch, one accountable name throughout.
01
Website and brand, delivered as one
A multilingual business website and the brand collateral around it — logo suite, colour and type system, print items — designed and delivered as a single project. The same pattern sits behind 961.be, the SKM Élégance salon package and the Makkah Consultancy one-pager.
Explore service02
The flagship — web applications
Web applications that take payments, manage people and run operations: donation systems, ticketing, bookings, portals and the admin back-offices behind them. AI coding agents make builds like this economical for a founder-led practice; automated tests, staged deployments and versioned migrations make them dependable.
Explore service03
Strategy-led redesign and migration
A strategy-led rebuild for organisations whose current site no longer serves them. We audit what exists, keep what earns its place, rebuild the rest on modern foundations, and migrate with safeguards so nothing of value is lost in transit.
Explore service04
One page. One audience. One action.
A focused standalone page for a campaign, event or offer: one audience, one action, measurement wired in before launch. Every component is a contained subset of platforms already in production — Stripe ticketing, booking forms, locale routing — not something invented for the occasion.
Explore service05
Maintenance, monitoring, small improvements
Ongoing care for a live site: dependency and security updates, monitoring, and measured content and conversion improvements — every change shipped through the same staged pipeline that built the site. It is the routine our client platforms already run on: versioned migrations, automated tests, dev, test and production environments.
Explore serviceThe model
You click a working version of the real software in the first days, not a slide deck at the end of a phase.
The person you brief is the person who builds, tests and ships—nothing is lost in handover.
The tools are named on every project—Claude Code, OpenAI Codex, Google Antigravity—never dressed up or hidden.
No release rests on a read-through: automated tests and staged environments gate every deployment.
English, French, Dutch and Arabic—with full right-to-left layout—are designed in from the first commit, not bolted on.
Live client work
Five client properties are live today—a registered charity’s donation platform, a competition platform with ticketing, a media-production suite with e-learning, a trilingual restaurant site and a band platform. Everything else shown is a case study, and is labelled as one.
Three products on one platform — public site, Academy e-learning and a /manage back-office with bookings, e-signature contracts and CRM — running on Supabase with more than 20 edge functions and push-to-deploy CI gated on database migrations.
A full Stripe giving stack for UK Registered Charity 1208169 — one-off and monthly donations, Apple Pay, Google Pay and Gift Aid — alongside a unit-tested Zakat calculator with live Nisab prices, a custom admin CMS, role-based auth and 38 versioned database migrations.
A custom SEO pre-render build step, a full JSON-LD menu and hreflang alternates across three languages, shipped through a dev → test → production pipeline for a Lebanese restaurant in Anderlecht.
Process
Four stages that put working software in front of you within days—then the engineering discipline that makes it safe to launch.
01
Frame
A short call and a written scope: what the software must do, in which languages, for whom. Days, not discovery theatre.
02
Prototype
AI coding agents build a clickable version of the real thing first. You react to software in your browser, not to wireframes in a deck.
03
Harden
The working version gains its production spine: Playwright and Vitest suites, WCAG accessibility work, security review, and a dev–test–production pipeline.
04
Ship
The build goes live through the staged pipeline, with GDPR-conscious analytics and technical SEO in place. A care plan keeps it maintained after launch, if you want one.
Founder-led collaboration
El-Sultan Web is the founder-led practice of Abed El-Sultan—no layered team, no handovers, fewer steps between the question, the decision and the implementation.
Where a project needs photography, video or another specialism, it is planned with trusted production partners. Their role, responsibility and cost are made clear before they are included.
About the practiceThe method, in the open
AI-accelerated engineering is the economic mechanism of this practice, not a footnote. Claude Code, OpenAI Codex, Google Antigravity, Hostinger Horizons and Claude Design carry work that used to need a team—which is exactly why the timelines and the pricing look the way they do.
It is not a shortcut past engineering. Everything the agents produce passes the same gates: Playwright and Vitest test suites, staged dev–test–production deployments, WCAG accessibility checks and GDPR-conscious analytics. The speed comes from the tools. The trust comes from the verification.
The whole method is written down—which tools are used where, what stays under human review, and what never ships without a passing test.
Read how I buildQuestions, answered
No. It is a short qualification conversation — around 20 minutes — to understand your organisation, what you need built, your languages and your timing, and to establish honestly whether El-Sultan Web is the right fit. It is not a free audit or strategy session, and it carries no commitment on either side. If the practice is not the right fit, you will be told so directly.
Organisations that need real functionality, often in more than one language: charities and non-profits (alberr.org runs donations, Gift Aid and orphan sponsorship for a UK registered charity), cultural and events organisations (bestvoicebelgium.be handles applications and Stripe ticketing), and small businesses that need bookings, e-commerce or a back-office rather than a brochure (lubbadproduction.com, 961.be). If a simple template site would serve you well, you probably do not need this practice — and the fit call will say so.
No. AI coding agents write the code quickly — that is why the pace and pricing are possible — but generation is the start of the process, not the end. Abed specifies every feature, reviews every change, and ships through automated test suites (Playwright, Vitest) and staged dev→test→production pipelines. The method is AI-accelerated engineering with founder accountability, and it is documented in full on the How I Build page.
Claude Code, OpenAI Codex and Google Antigravity for engineering, Hostinger Horizons for rapid prototyping, and Claude Design for visual design work. They are named because you should know how your product is made. The How I Build page explains what each contributes and, more importantly, the review and testing discipline wrapped around all of them.
The projects are built on standard, widely used stacks — Next.js, React, Supabase, Stripe — so any competent developer can read, maintain and extend them. Discipline does the rest: automated test suites, versioned database migrations (the alberr.org database carries 38 of them), role-based access control and staged deployment pipelines. How the code is written has changed; what gets shipped is conventional, documented software.
Project fit
Twenty focused minutes: what you need built, in which languages, and what a prototype-first timeline looks like for it. If the project is not a fit for this practice, you will hear that plainly.
Book a project fit callInitial qualification only. Not a free audit or commitment to purchase.