Skip to main content
  1. Blog/
  2. AI/

How a Non-Technical Person Built a Website with Codex + AI in Two Hours

·827 words·4 mins
Fawei
Author
Fawei

In recent years I have gradually developed a habit of reading ebooks, and I have collected many ebook resource sites.

A long time ago I had an idea: if these scattered sites could be gathered into one platform, it would be much easier for people who need them.

But the reality is that I am not from a technical background and I knew almost nothing about building websites. Even though I looked up tutorials and solutions, there was still a barrier for me, so I shelved the idea for years.

Until recently, the progress of AI reignited the idea.


AI lets ordinary people “write code”
#

Over the past two years, AI has been evolving at a leap.

Now we can use natural language to let AI generate, modify, and optimize code. In the industry this is called vibe coding: people provide the intent and direction, and AI handles the implementation.

This not only greatly improves developer efficiency, it also gives non-technical people like me a chance to turn ideas into real products.

Before this, I had already used AI to add pages to the official site of an open-source project and experienced what “zero-basics development” feels like. This time I decided to go further and build a site from scratch with AI.


Why Codex?
#

There are more and more AI coding tools on the market, such as Claude Code and OpenCode.

In the end I chose Codex from OpenAI.

Simply put, Codex is an AI programming assistant, like a “24/7 developer partner” you can talk to.

It can help you:

  • Generate website code
  • Adjust page styles
  • Fix bugs
  • Deploy the site
  • Analyze project structure

And the whole process is basically done through conversation.

Website: https://openai.com/index/introducing-codex/


Preparation: installation and setup
#

Codex can be installed in many ways, and the official site covers them.

I prefer a chat-first workflow, so I installed the official desktop app and did not bother with the command line.

Using Codex has a cost:

  • Buy a ChatGPT subscription
  • Or use an OpenAI API key

For cost performance, I chose the API key route and bought a one-month package on Xianyu to test. Sellers usually provide detailed instructions, so you can configure it by following the steps.

Reminder: third-party channels carry risk. For important projects, use official channels.


Building a site from one sentence
#

After the preparation, I could officially start “writing” the website.

My first sentence was roughly:

I want to build an ebook resource directory, with categories and cards that link out.

Then I added:

  • The page structure I wanted
  • Reference sites
  • Style preferences

Codex quickly generated a local webpage that I could preview in the browser.

Next came iterative conversation-based tuning:

  • Change this area to a card layout
  • Make the categories clearer
  • Increase the font size
  • Add a search box to the home page

The whole process felt like collaborating with a real developer.

The site structure I built is simple. It is essentially a directory navigation site with no complex interactions, so the total development time was under two hours.


Deploy so others can access it
#

After writing the code, the site was only accessible on my computer.

To make it public, it had to be deployed to a platform such as:

  • GitHub Pages
  • Netlify
  • Vercel

Because I had already used GitHub Pages for my blog, I chose it again.

If you are not familiar with deployment, you can ask Codex to handle it. After authorization, it can complete the commit and publish steps for you.


Buy a domain to make it more “official”
#

After deployment, the site was accessible, but the default URL was long and hard to remember, which is not good for SEO.

So I bought a custom domain and set up DNS and CDN acceleration.

I bought the domain on Alibaba Cloud. It costs only a few dozen yuan per year.

At that point, the site was fully live.

Live site: zhaoshuba.top

Zhaoshuba English interface


A few lessons from the process
#

Based on this experience, here are a few takeaways:

1. Third-party APIs have risk
#

The Xianyu option is cheap, but stability is not great. Be careful for important projects.

2. Ask AI when you do not know
#

When you run into problems, just ask Codex. It solves most of them.

3. Clearer prompts lead to better results
#

The more specific your requirements are, the closer the AI output is to your expectations.


Closing
#

I used to think that building websites and writing code were “exclusive skills” for technical people.

But now, AI has flattened that barrier.

With Codex, I not only finished this ebook directory site, I also refactored the style of my old blog along the way.

Many tasks that used to take days or weeks can now be done in a few hours.

For ordinary people, this may be one of the most important dividends of our time.

Thanks to AI, an idea finally became real.