I just gave up

I thought building a website with AI would be straightforward. It took me 9 months until I was satisfied.

At first, my wife helped me with the design, and she started by hard-coding it. I figured Cursor would help her code it faster.

It did help, but we just didn't know how to use the tool properly. We got stuck more than once.

The site looked alright, but it wasn't the quality we wanted.

We switched to Framer and she redesigned it. She really enjoyed the tool and it looked amazing.

I was planning on using what she built as a template for 10 experiments across different domain names. Then I discovered Framer's weird pricing model. You pay a subscription per website, not per account.

That meant 500 or 400 bucks per month for 10 websites. Plus it's hard to integrate Framer to the rest of your workflows.

I wanted to be able to do my work with an AI companion to brainstorm, get some research done, get some data on keywords, etc. But all in the same place. I figured it needed to be hard-coded to achieve this. Framer is a closed system, so you can’t really achieve this (no solid API).

When GPT-5 came out, Cursor offered free credits for a week. Glory! I decided to export all the Framer components and push it live myself.

I worked on it for a couple of days when I had the time. I researched the best frameworks. I created a PRD for the migration. I did a task list based on the PRD. I worked on each section diligently, creating components and sections. When it was done, I looked at the preview.

And oh my days. It was so shit. Worse than a website from the 90's.

I worked for so long. I just gave up.

I didn’t touch this disappointment for a few weeks.

Then I realized something. I decided to take all that stuff (the PRD, the task list, and my new found knowledge of components) to go at it again.

This time, I'd thoroughly test out each component and each section before moving to the next.

Results were night and day.

The problem with AI is it would just fill in gaps and make assumptions that might not necessarily be the best for the project. I really had to be strategic and ask it questions about why this tech stack, if it would be problem for performance, why a certain approach and not another, etc.

Even though I don't know the ins and outs of the specific technology, I’m able to make an informed decision about the direction I want to take. Otherwise, I’m taking decisions blind without understanding their consequences.

What worked was taking screenshots from the Framer sit and being systematic. "Okay, let's do the nav. Let's make sure each of the components work. Let's make sure the nav looks good. Let's preview everything at each step. Let's refine along the way."

I would screenshot each section on mobile and on desktop just to show how to make it responsive. There's an interactive process you need to have to make sure you get what you want in the end.

Each time I got stuck, I went back to the drawing board and asked it to give me more information about specific situations so I could be more strategic.

That’s the big lesson here for me. Whenever I feel like I’m stuck and this stupid computer doesn’t understand what I want, it’s generally the time to dig deeper. That means I’m bumping into the limits of my knowledge. I can’t ask better questions or give better instructions until I understand what’s going on.

You need to be systematic and strategic with AI rather than letting it fill gaps.

The more context you give and the more you understand the overall process (even if it's high-level), the less frustrating, easier, and more rewarding it becomes.

You can't expect AI to guess what's in your mind. You still need to be the architect, the guiding force that has all the context.

Anyway, it still took me a bit of time to make sure everything worked perfectly. But I got there.

Now I can directly grab content from my Google Drive, search keyword volume, brainstorm articles with AI directly in Cursor and add new content to my site. If I need to optimize copy based on keyword volumes, I can do that directly and get the AI to guide changes on the website.

That's freaking awesome! Great performance, SEO is optimized, accessibility is through the roof. Everything looks amazing.

The whole goal is to have everything centralized for my business. I’m making big strides and this is so exciting!