I built this website at the start of 2026 on WordPress and Elementor. It was a portfolio: somewhere for people who were thinking about hiring me to see who I am and the kind of work I produce. It did that job.
By the autumn of the same year I had rebuilt it from scratch with Claude Code, as plain files, with no WordPress and no page builder underneath it. This is the honest version of why, including the parts that were harder than I expected.
Screenshot to add
PageSpeed Insights, before and after, side by side
Screenshot to add
securityheaders.com, before and after, side by side
What building with Elementor was actually like
I have been building websites since I was sixteen, so I am not someone who was defeated by the tool. I knew what I wanted and I knew the words for it. The problem was how long it took to get there.
The look I was after needed CSS and JavaScript that Elementor could not produce on its own. Small things. The way panels open, the way the menu behaves. Even back then, I was using Claude to write that custom code and then pasting it into a WordPress site. So the site was already part custom. It just had a page builder wrapped around it.
Then there is responsive design. You build a section for desktop, and then you build it again for tablet, and again for phone. Sometimes that is three versions of the same section living in the same page. Every version adds weight, and weight is speed. The part people do not expect is that this starts before you add anything at all: install WordPress, install Elementor, publish an empty page, and run it through a speed test. You already cannot get a great score.
That was the slow drift. The thing that actually made me walk away was realising that everything I wanted to do to my website was making my website worse.
The two things that made the decision for me
Payments. I wanted to take payments on my own site. In Armenia you cannot simply plug in Stripe or PayPal, so taking cards means integrating with a bank's vPOS. Doing that inside WordPress means more plugins, more connections, more things between me and the bank. Owning the files meant I could build the payment flow the way the bank needs it and nothing else.
Security. A client's WordPress site was hacked. Nothing dangerous came of it, but the reason was ordinary: the WordPress version had not been updated. That is always the shape of it. WordPress and Elementor both get vulnerabilities, and you are safe for exactly as long as you keep up with every update, every plugin, forever. You can try to stay ahead of it and still not succeed. When you upload your own files there is no admin login to attack and no plugin to leave out of date, and you can decide the security posture at the start instead of patching it later.
Why I did not just hire someone, or move to Framer
I looked at Framer. The sites people build with it look genuinely good. But the learning curve is steep, and at the end of it you are locked into a subscription you pay forever, which is exactly what I did not like about Elementor: your site exists at someone else's pleasure, and if you stop paying you lose the tools that built it.
Hiring was the other option, and I have done it. Before Claude Code existed I worked with a developer. I would tell him what I wanted, wait a few days, then we would test it. We were productive, but the loop was slow. Something that took us a month back then is something I can now do in a day.
And the thing is, if I hired someone today I would still be writing the same instructions. I am the one who knows what I need. The only difference is how long I wait for the answer. With Claude Code there is no wait, and it will sometimes raise a problem I had not thought of, or make a change on its own initiative that turns out to be the right one.
I know what experienced developers say about this, because I have heard it: AI is not there yet, the code is not clean. They are not wrong. I see the mistakes it makes. Sometimes it complicates something instead of simplifying it. But in business, speed matters. How quickly can you test an offer? How quickly can you put something in front of someone? That is worth more to me than perfect code, and it is worth more to my clients too.
If you want the best of both, there is an obvious order: use Claude Code to build the smallest real version of what you want, see it working, and then hand it to a developer to refine or rebuild. You will communicate far better with something real in your hands, and it will cost you less time and less money than describing it in the abstract.
How the move actually went
Simpler than it sounds. I had a clean, organised folder with all the files and images for my website, and I had the WordPress site downloaded. I gave Claude Code the folder, the download, and the address of the live site, and asked it to rebuild it.
It did. The front end came back looking like my website. Not approximately: accurately. I think the folder and the WordPress export are why. When you only give an address, you do not get this.
It was a big project, so it took a while, and some single prompts took a few hours to work through. But once the copy existed, I could finally make all the changes that would have been painful in WordPress. The speed scores came out better than sites I admire.
What surprised me
The good surprise was how close the rebuild was on the first pass. In my head this was an experiment: can this even be done? And it was done.
The bad surprise came later, with time. The longer you work on a site, the more it starts acting up. A change in one place quietly undoes something somewhere else. My menu's behaviour changed more than once without my asking, and I had to notice and say so. If you make changes and simply trust that they were made correctly, you can end up with a version of your own site you do not recognise. You still need a human looking at it.
The hardest single thing was absurd. There are round dots on my menu, and for a long time it kept drawing them not quite identically, no matter how many times I asked. I had to find the right way to say it. Then it was right. That is the real texture of this work: not big failures, small stubborn ones.
Nothing broke after launch. Things broke during the build, which is where you want them to break.
The part I did not expect to build
WordPress gives you a back end. You log in, and there is a dashboard with your pages and your form entries and your settings already in it. When you build from files, nobody hands you that.
So I built it. My site now has its own back office where I see the things only the owner should see. That is the part I surprised myself with. When people imagine building a website with Claude Code, they picture the pages. They do not picture building the admin view behind them. I did not plan it either. It came out of asking the right next question: fine, the visitor can pay, but where do I see that it happened?
This is also the honest answer to what I lost. Nothing was taken away from me, but nothing comes pre-made any more. Every time I add something, an offer, a checkout, a post, I have to think about both sides of it: what the visitor sees, and what I see afterwards. WordPress thought about the second half for me. Now that is my job. I can build it, and I would rather have it exactly how I want it, but it is work that used to be free.
What is different day to day
I do not drag and drop sections any more. I describe what I want and it gets built, the same way I would brief a developer, and the better my terminology the better the result.
Publishing changed too. I used to upload files to the host by hand every single time I changed something. Now it publishes itself over SSH. I no longer touch the host.
And before anything goes live, the site gets checked: the pages themselves, then the pages in a real browser at phone, tablet and desktop width. It adds a few minutes. It also catches things I would never have found on my own, which is exactly what I want standing between a change and the public.
What it costs
My bill is higher than an Elementor licence, because I pay for the top plan rather than the twenty dollars a month that would also work. You can build a site on the cheaper plan, you just run into usage limits and wait.
But I do not use this only for my own website. I use it for client work and for everything else in my business. It behaves like having an employee, and my business moves faster because of it. Measured that way it is the easiest cost I carry. And unlike a builder subscription, if I stopped paying tomorrow my website would still be mine and still be online. That is the difference that mattered to me.
When I would tell you not to do this
If you have a system that works, do not break it. There is no prize for destroying something functional.
What I would do instead is build alongside it. Put the new thing on a subdomain, or build the moving parts, the offers, the landing pages, the checkout, with Claude Code and point to them from the site you already have. Keep your main infrastructure where it is and let the new way prove itself in the places where it is cheap to be wrong.
Why I did it anyway
The world is changing and adapting is not optional. Adapting also means taking a risk, and the risk is real: AI is not perfect right now, and I have shown you the places where I felt that. But I think staying too conservative is the more expensive choice over time. It is not the strongest or the cleverest who come through a change like this. It is the ones willing to adapt fastest.
If you want a website built this way, that is the work I do. Tell me about your project.
