How I accidentally built a multi-layer web architecture trying not to waste a hosting plan
So, this started with something that looked simple… and turned into a small infrastructure lesson I didn’t expect.
Most of my domains are registered on Cloudflare. That was intentional. I like having one place where everything is controlled, fast, and flexible. DNS, SSL, routing… all clean. At the time, I was building a lot of things in Figma Make, so pointing domains through Cloudflare made sense. It was quick, almost frictionless.Then I entered what I now call “the migration phase”.
I started thinking: maybe I need something more stable, more “real”. Something that feels like a proper system, not just prototypes. So I bought a hosting plan on Hostinger. Looking back… I didn’t fully need it at that moment. It was one of those decisions where you think you’re being strategic, but actually you’re just moving fast without clarity.
Still, I paid yearly. And it was a fair price. So the question became:How do I use this properly without wasting it?That’s when WordPress came back.I’ve used WordPress for years. Not in a trendy way, not as a “modern stack” decision. Just… it was always there. It grew with me. I had notes, drafts, fragments of thinking going back 13, maybe 14 years. That matters. There’s something about continuity that no new tool gives you. So instead of fighting it, I decided: let’s use WordPress again. But properly this time.
My domain was on Cloudflare. My hosting was on Hostinger. And I did not want to break anything by changing nameservers again. That part is important, because changing nameservers is like moving the foundation of your house. You can do it, but you better know why.
So I didn’t.Instead, I kept Cloudflare as the source of truth. Then I asked a very specific question:
How do I point a subdomain to a WordPress installation without touching nameservers? This is where something interesting happened. I used both ChatGPT and Gemini. Not one, both. And the answer converged.The logic is simple once you see it:You don’t move the domain.
You route it. So the setup became:
Cloudflare manages DNS.
Hostinger hosts the WordPress instance.
And I create a subdomain:blog.luana.systems
Inside Hostinger, I created a new website with that exact subdomain. Hostinger then gave me either an IP address (A record) or a target hostname (CNAME). That’s the “destination”. Then I went back to Cloudflare and added a DNS record:
Type: CNAME (or A, depending on what Hostinger gave me)
Name: blog
Target: the Hostinger address
Proxy: enabled if not mistaken
No nameserver changes. No downtime. No drama. That gave me blog that is:
- Connected to my long-term writing history
- Sitting inside my current domain structure
- Using infrastructure I already paid for
- Without breaking the rest of my ecosystem
There is a small lesson here that I didn’t expect.I thought I was solving a hosting problem.But actually I was learning how to think in systems.Most people treat domains, hosting, and platforms as one thing. They’re not. They are layers. And once you separate those layers, you get control.Would I make the same decisions again? Maybe not exactly. I probably wouldn’t rush into buying hosting before I needed it.But now that it exists, I’m using it properly.WordPress… still there, after all these years. Not perfect, not modern, but stable. Familiar. A place where writing accumulates instead of disappearing.That matters more than people admit.In parallel, I discovered something that actually changed how fast I could ship: deploying through Vercel.Using Claude Code, I could generate a site, push it to GitHub, and then connect that repository directly to Vercel. The flow became almost frictionless:
Write or generate code → push to GitHub → auto-deploy on Vercel.
WordPress is about continuity. Accumulation. History.
Vercel is about execution. Shipping. Iteration.