I Rebuilt My Entire Web Presence in 48 Hours Using AI. Here's What I Learned.

I've been on WordPress for more than 15 years. I know the platform well. And for the last few of those years, it was costing me money and time I no longer wanted to invest — in hosting, in maintenance, in the slow accumulation of technical debt that comes with a WordPress site nobody is actively tending.

I'm not a developer. I'm a digital strategist with 20+ years of experience helping B2B companies grow online. I know enough to be dangerous with code, but not enough to ship a production React site from scratch without significant pain. WordPress was the path of least resistance. So there it stayed.

Then in late March 2026, a hosting renewal deadline forced my hand. Siteground — where my blog and my consulting site had lived for years — was coming up for renewal. The cost wasn't the issue. The inertia was. And this time, with AI tooling available that simply didn't exist two years ago, I decided to do it properly instead of paying for another year of debt.

Forty-eight hours later, everything was done. Both sites redesigned and rebuilt from scratch. Old hosting cancelled.

Here's what happened — and what it means for anyone still paying for WordPress.

The Stack

I'll get the technical part out of the way quickly, because the tools matter less than the approach.

React + TypeScript + Tailwind CSS + React Router 7 — the main consulting site (kiprin.com), built as a static site and deployed to Vercel
Simply Static — used to export the old WordPress blog as plain HTML, then redesigned from that base
Google Stitch — for design ideation and visual direction; used to explore palettes, typography, and layout before writing a line of code
Cursor Pro — the AI-powered code editor that did the actual building
Claude — for planning, PRDs, decision-making, and writing
Vercel Pro — hosting, already in use for another project, so no new cost

The consulting site was a ground-up build. The blog was a migration first, redesign second. Both are now fast, version-controlled, cheap to run, and — critically — mine to control completely.

What AI Actually Did

This is the part worth being precise about, because "I used AI to build a website" covers a wide range of things.

I did not prompt an AI and get a finished website. What happened was closer to having a very capable technical co-founder available at 11pm who never got tired, never lost context, and never needed me to explain the same thing twice.

Claude helped write the PRDs — the detailed specs for each site before a single line of code was written. Design decisions, content structure, redirect mappings, DNS configuration. All documented, all thought through in advance.

Cursor's agent mode did the implementation. I described what I wanted, it wrote the code, I reviewed it in the browser, we iterated. The things that would have taken me days to figure out alone — MDX routing, favicon generation, sitemap creation, 301 redirect logic — took minutes.

The judgment calls were still mine. What to keep, what to cut, what tone the site should have, which blog posts were worth migrating. AI doesn't make those decisions. But it executes on them immediately, without the usual friction between decision and result.

The WordPress Question

Here's the part that sat with me after it was done.

I've recommended WordPress to clients for years. It's the sensible default: enormous ecosystem, easy to manage, developers everywhere, hosting widely available. For a marketing team that needs to publish content without touching code, it still makes sense.

But for a professional site that a single person owns and controls? The calculus has changed.

WordPress, Wix, and Squarespace all built their moats on the same assumption: that non-developers needed a managed, visual layer between them and the web. That assumption was correct for a long time. Building a production-quality website without that layer required either significant technical skill or significant budget.

That assumption is now being dismantled — not slowly, but quickly.

Cursor and Claude together make it possible for someone with my profile (strategist, not engineer; comfortable with concepts, not syntax) to ship a React site with proper SEO, correct redirect logic, and a clean design in a weekend. Not a toy site. A real one, with MDX blog posts, JSON-LD schema, PostHog analytics, and a 404 page that actually looks like it belongs there.

And then there are tools like Lovable and Replit that push the accessible end even further. With Lovable, you describe what you want in plain English and it generates a working React app in your browser — no local development environment, no terminal, no Git setup required. Replit does something similar for a broader range of projects. These tools are still maturing, but the direction is clear: the barrier to building real things is collapsing for people who couldn't previously build them.

For the platforms, this is a genuine strategic problem. The value proposition of "we make it easy to build a website without coding" weakens every month. Not because the platforms are getting worse — they're not — but because the alternative is getting dramatically better.

What This Means in Practice

A few observations from the other side of this:

Speed compounds. The reason this took 48 hours rather than a weekend per site is that every decision fed the next one. Once the consulting site structure was clear, the blog followed a similar pattern. Once the DNS logic was understood once, it applied everywhere. AI tooling doesn't just speed up individual tasks — it compresses the learning curve between tasks.

The boring parts are now fast. Generating PNGs from an SVG favicon. Writing a sitemap. Setting up 301 redirects for 17 URLs with different destination mappings. These are the tasks that used to make a project feel heavy. They're now genuinely quick — which means you actually do them properly instead of skipping them.

Version control changes your relationship with risk. Everything in Git means nothing is permanent. I redesigned the blog the day after finishing the consulting site because the cost of getting it wrong was low. On WordPress, a redesign feels like surgery. On a static site in Cursor, it feels like editing a document.

The judgment still matters. The 20+ years of experience didn't become irrelevant — it became more leveraged. Knowing what a good site structure looks like, what copy actually converts, what SEO fundamentals matter, which tools are worth using: that knowledge is what turned AI capability into a finished result rather than a pile of generated output.

Should You Do This?

If you're a solo operator, consultant, or founder with a WordPress site you've been meaning to fix: yes, probably.

The prerequisite is comfort with ambiguity and some tolerance for things breaking temporarily. This is not a no-code experience in the Squarespace sense. You will hit errors. You will need to read them and respond to them. AI helps significantly, but it doesn't eliminate that entirely.

If you're running a content-heavy site with multiple contributors, or you need a non-technical team to publish without developer involvement, WordPress still makes sense. The platform isn't dead — it's just no longer the only sensible choice for the people it used to have locked in by default.

For everyone else: the tools are here. The window where this required either a developer or a page builder has closed. What it requires now is a clear sense of what you want and a willingness to work iteratively toward it.

The hardest part, honestly, was deciding to start.

Borislav Kiprin is a digital strategy and growth consultant based in Munich, working with B2B SaaS companies at Series A–C. He also builds things — including ContactSheet.ai.