I Thought AI Would Build My Website. Instead, It Taught Me How.

Before we get into this, I need to be honest about where I started.: I had no coding experience. None. I'd never built a website from scratch, never deployed one, never opened GitHub, never connected an API, and I couldn't have told you the difference between a domain and hosting if you'd asked me to guess.

I say that because I think a lot of people picture "building a website with AI" as something close to magic — type a sentence, press a button, done. I didn't believe it would be effortless, but I still underestimated how much of the work would fall on me — and how much I'd need to learn from scratch before I could actually build something I was proud of.

Why I Didn't Just Use What I Already Had

I already owned hosting through Hostinger, and I'd used Elementor before — the drag-and-drop builder a lot of small businesses lean on because it's simple. I could have taken that route. I know how to do that.

But I didn't want basic. I didn't want boring. Every time I'd tried Elementor before, I could never quite get it to look like what was actually in my head. So instead, I opened Replit, with no roadmap and no idea what I was walking into.

My First Time Doing Any of This

TishTips.com is not a second attempt at something I'd already figured out. It's my first live website. Before this build, I'd never touched Replit, GitHub, Netlify, Sanity CMS, DNS, SSL, or a custom domain. I'd never taken something from an idea to a real live website that other people could actually visit.

Every term, every mistake, every fix — it all happened right here, on this build, over about two weeks. I want to be clear that I wasn't confused for weeks on end. I was learning one piece at a time, solving problems as they showed up, and moving on to the next one.

The Actual Workflow

Building this site was never one AI tool doing everything. It was me, connecting several of them, each playing a different role.

ChatGPT is the AI I've used the longest. It knows how I think and how I communicate, so it became the translator between my vision and the technical world underneath it. I'd describe what I wanted in plain language, and ChatGPT would turn that into something I could actually hand off.

Replit became the AI coding agent partner that generated and updated most of the code. I'd take what ChatGPT helped me put together and prompt Replit with it, and Replit would write and execute the code. When something broke, I'd screenshot the error or copy exactly what Replit said, hand it to ChatGPT, and ask what it meant. Sometimes that solved it in one exchange. More often it took three or four rounds before I understood the problem well enough to fix it. Occasionally I'd bring Gemini in too, mostly to compare a solution ChatGPT gave me against a second opinion when something still wasn't adding up.

Claude wasn't part of building the site. Claude came in afterward, to help me write about it.

None of these tools talked to each other on their own. I was the one carrying information between them.

What Kept Breaking

Enough things broke along the way that I started keeping a mental list:

• Deployments failed, with error messages I couldn't make sense of until I handed them to ChatGPT.
• GitHub was completely new to me — I didn't understand what a commit was or why it mattered until I skipped one and broke something because of it.
• Netlify builds failed for reasons that often traced back to something small in the code, not Netlify itself.
• Environment variables and API keys were something I'd never dealt with before and getting them set correctly between platforms took real trial and error.
• DNS and custom domains were new to me too — getting a domain to actually point to the live site took extra research I didn't expect to need.
• Images wouldn't load, showing up as blank boxes for reasons I couldn't see.
• Sanity wouldn't sync, so a published article wouldn't show up the way it was supposed to.
• Featured articles disappeared from the homepage entirely.
• The homepage cached old content, so I'd update something and still see the old version until I learned that was even a thing.
• Routes broke and slugs mismatched, so a link would 404, or the homepage would reference an article I'd already deleted or renamed.
• Beehiiv newsletter integration, which looked simply until it wasn't.

I want to be clear — it wasn't chaos every single day. But it was often enough that I had to learn something most tutorials never mention figuring out which system owned the problem is half the actual work.

The Mindset Shift

At some point, something changed in how I reacted to all of this. Early on, every error felt like a small disaster — like I'd broken the website and maybe couldn't fix it. Eventually I stopped thinking that way. Instead of "I broke the website," I started asking, "Which service is actually responsible for this?" Is it Replit, Netlify, Sanity, GitHub, or something I changed myself three days ago?

That one question completely changed how I approached every new problem after that. It's a small shift, but it might be the biggest thing this whole process taught me — not a piece of code, but a way of thinking through a problem before panicking about it.

The Night I Almost Stopped

There was a night, about a week into building this site, where I sat back from my laptop and genuinely asked myself if I should keep going. I don't even remember which bug it was anymore. I was tired, frustrated, and wondering if I'd taken on something that was way over my head.

What kept me at it wasn't a burst of motivation. It was that I didn't want this to become another thing I started and never finished. So, I stayed with it, one problem at a time, until I actually understood what I was doing.

What AI Actually Did — and Didn't Do

AI did not build my website for me. I built it, with AI beside me — several different AIs, each doing a different job, with me as the one connecting all of it.

ChatGPT translated my ideas into technical language. Replit generated and implemented the actual code. Gemini occasionally helped me compare answers. Claude helped me write about it afterward. None of them could tell me, with certainty, what was actually wrong on my specific site in that specific moment. Every decision, every test, every judgment call was still mine.

What AI did do was compress time. Concepts that would've taken me weeks to learn on my own, I learned in real time, attached to a problem I actually needed solved that night.

Where It Stands Now

TishTips.com is live — multiple AI hubs, real articles, images that load, a working connection between Sanity and Netlify. It came together in about two weeks, not because any of it was simple, but because I kept working through one new piece at a time until it worked.

The Real Lesson

I don't want anyone to read this and think AI replaced the need to actually learn something. It didn't remove the frustration or the hours of not knowing what was wrong. What it did was make the learning faster, and made sure that when I chose to keep going, I wasn't doing it alone.

AI didn't replace the learning.

It accelerated it.

Two weeks before starting this project, I didn't know what GitHub, Netlify, Sanity, or DNS even were. Today, I have a live website that I built myself—with AI as my teacher, not my replacement.