A Beginner's Website-Building Diary: From Asking Gemini Questions to a Website in 30 Minutes

A note before you read: this article is not a technical tutorial, and it won’t even teach you how to write a single line of code. It’s the real story of a complete beginner with no technical background seriously wanting a website of their own for the first time — including the confusion I ran into, the pitfalls I stepped in, and the discovery that surprised even me at the end: building a website really isn’t as hard as I imagined.

---

It all started with a question

A few days ago, I knew absolutely nothing about “building a website.” Servers, domains, deployment, frontend, backend — it was all a blur in my head. My only tool was an AI chatbot everyone uses: Gemini.

Yes, that Gemini — free, with a clean interface, happy to answer whatever you ask. Back then I didn’t even know the difference between a “static website” and a “dynamic website,” let alone anything called Astro.

So the first thing I did was wonderfully simple — I treated myself as someone who knows nothing, and asked Gemini: if I want to make a personal website, where do I start?

---

Step one: Let Gemini help me pick a theme and a name

The hardest part of building a website is usually not the technology, but “what website do I actually want to make?”

I had three or four completely different ideas floating around in my head at once: I wanted to write about photography, record the ordinary bits of everyday life, and occasionally dig into personal finance and trading. They were too far apart — I couldn’t decide which one to pick, let alone figure out how to squeeze all those stray thoughts into a single site.

So I dumped every fragment of an idea into Gemini and asked whether they could all live together. Gemini didn’t rush to hand me a plan. Instead, it helped me sort things out and came back with a surprisingly smart structure: rather than “one big mixed bag of everything,” make it “a knowledge garden made of several clearly defined columns.”

It helped me organize those scattered interests into four sections:

• AI and local computing power
• Photography and optics
• Quantitative trading
• Everyday saving and life’s little wins

The moment those four names appeared, I got it instantly — this is what my website should look like. Each stands on its own, yet they share a common spirit of “hands-on tinkering, doing it yourself.” Riding that same spirit, Gemini also helped me come up with a name I was quite happy with: CyanLabs (Qingyan Labs).

“Qing” carries the color cyan — evoking AI, calmness, and a techy feel; “yan” means to grow and spread. A cyan-toned, lab-flavored place for recording all my tinkering — the name was settled. Throughout the whole process, I never wrote a single line of code. I just kept talking to Gemini.

---

Step two: Figuring out what the “website-building environment” actually is

With the theme and name settled, the next question was very practical: with this ordinary Windows laptop of mine, can I actually build a website without spending any money or buying a server?

So I went back to Gemini: in which cases is DIY a real hassle, and in which cases is there actually a shortcut? Gemini taught me one crucial concept — building a website doesn’t necessarily mean buying a server. There’s something called a “static website”: it’s just a set of fixed web page files that you can drop onto a free hosting platform, and people can visit it — no server costs at any point.

It also explained a lot of “basic environment knowledge” to me. I was a bit dizzy at first, but I grabbed the few points most useful to a beginner like me:

• I need to install some things on my computer that “run website-building tools” (like Node.js) — install them and follow the steps
• Once the website files are done, throw them onto a free platform called Cloudflare Pages, and it goes live
• The whole thing can be previewed locally first, so I can see how it looks and only publish once I’m happy — no need to fear breaking anything

Looking back, these are all basic points now, but for me back then, it was like a window suddenly opening: I didn’t need to understand all that complicated jargon from the start — I just needed to follow the pattern and take the first step.

---

Step three: Getting to know a new species called “Agent”

I understood the environment concepts now, but the biggest problem remained — even if I knew I should build a website, I can’t write code. Those config files and page codes — I didn’t recognize a single one.

That’s when Gemini gave me another lesson and introduced me to something completely new: the AI Agent.

It explained that the biggest difference between an Agent and an ordinary chatbot AI is this: with a chatbot, you ask and it answers; with an Agent, you can actually have it roll up its sleeves and do things. You can tell it “help me build a website in this folder,” and it will really go and create files, write code, and run commands on its own — like hiring an assistant who actually gets things done, not just a consultant who talks.

This concept hit me hard. I realized building a website might really be about to get simpler — I don’t need to become a programmer myself; I just need to learn how to give an Agent clear, sensible instructions.

So the focus of my learning shifted from “how to write code” to “how to be a proper client and explain things clearly.”

---

Step four: Why I ended up with a “dual Agent” setup

You might think one AI assistant would be enough, right? I thought so too at first. But once I actually got started, I found a very real pain point:

An ordinary conversational AI is great at brainstorming and reasoning, but when you ask it to actually sit down in the file system, dutifully create dozens of files, and run commands over and over until they succeed, it often feels a bit “helpless.”

A dedicated Agent that can genuinely operate your computer and working directory, on the other hand, is great at burying its head in the work — but sometimes it gets so focused on the details that it loses sight of the big picture.

That’s when Gemini gave me a very clever suggestion — a decision I’d still give it a thumbs-up for today: why not try a “dual Agent” division of labor?

• One Agent acts as the “commander-in-chief,” handling the overall strategy, assigning tasks, running the pipeline, and keeping an eye on quality — like a project manager.
• The other Agent acts as the “senior welder,” doing fine, detail-oriented work whenever the interface code demands precision and craftsmanship — like a senior engineer.

The two roles work together, each doing what it does best. It made total sense to me — one person acting as both project manager and welder easily drops the ball on one or the other; split them up, and both stay more focused.

So this became the combo: Hermes takes the lead as commander-in-chief, handling the overall orchestration and operations of the entire build; whenever a complex UI component comes up, Codex — the “welder” — is brought in to handle it.

---

Step five: Hooking the Agent up to DeepSeek V4

With the tools chosen, the Agent still needed a “brain.” Here, Gemini helped me sort out the difference between a “model” and an “Agent”:

• The Agent is the hands and the skeleton — it operates files, runs commands, and works through the process.
• The Model is the “brain” — it handles understanding and thinking; it’s what actually does the reasoning behind the Agent.

The two work together: only with the right model attached does an Agent become smart enough.

In the end, I equipped this dual-Agent setup with DeepSeek models, and I even made a fairly careful division of labor:

• Hermes (the commander-in-chief) mostly runs on DeepSeek V4 Flash, with thinking mode off — all for one reason: speed. Orchestration and most routine operations don’t need that deep reasoning; faster and cheaper is better.
• Only when handling those complex component codes is Codex brought in, letting Codex use DeepSeek V4 Pro to think carefully and craft the fine details.

A combo of “fast grunt work + slow fine work.” At the time I didn’t really understand the logic behind it, but I just thought — oh, so you can hand tasks of different difficulty to “brains” of different intelligence, and get both speed and savings.

---

The climax: in the final half hour, a website prototype actually appeared

And then came the part that amazed even me.

After I had laid out all the pieces — theme, name, environment, Agent, dual-Agent division of labor, model configuration — the actual “building” itself took only about the final half hour.

In that half hour, the Hermes Agent was off on its own, grinding away creating files, tweaking configs, running commands, and fixing errors; when it hit a fairly complex UI component, I handed it to Codex for fine work, and once Codex finished and came back, Hermes stitched the whole site back together.

I basically just watched it work, occasionally answering a question or two — like “should this button be emphasized?” and “what copy should go in this section?” The whole time, I never wrote a single line of code myself.

When it showed me the homepage in the preview, I stared at the screen for a good long while: the cyan logo, the cards for the four columns, the entry point for an article — all there. It was a real prototype of a website: accessible, and ready to be filled with more content.

I kept confirming one thing in my mind: this is me — a beginner who, an hour ago, was completely lost when it came to code — standing in front of a website built by hand.

---

The ledger: 8M tokens, RMB 1.36

You might be curious how much this whole thing cost.

I kept a careful record, because the number was so small it surprised me. From the two Agents doing their work and running commands to handling the components, the entire build consumed roughly 8 million (8M) tokens — for AI, that’s already a considerable amount, since the whole site had to be built from scratch with plenty of trial and error.

But what does that translate to in money?

About RMB 1.36.

Yes, you read that right — not RMB 136, not RMB 13.6, but RMB 1.36. For comparison, that’s probably cheaper than my breakfast. That number genuinely made me question reality for a moment: I’d always assumed building a website costs a fair amount of money, and yet a live-ready website prototype costs less than the price of a single breakfast.

Of course, I’m also aware: partly it’s because DeepSeek’s models are priced very low to begin with, and partly it’s because this “fast model for grunt work, slow model for fine work” division of labor cut out the unnecessary expensive reasoning. But either way, from that day on, “even one person can own a website at low cost” became real for me.

---

Ending: from a single question mark to a Qingyan Labs of one’s own

Looking back, all I did was something very small and very ordinary: I opened a chat window and asked a question.

But from that moment on, I used Gemini to settle the theme and name, understood the basic website-building environment, met the Agent, learned the dual-Agent division of labor, gave the Agent the fast-and-cheap DeepSeek V4, and then watched it build a website from scratch in half an hour — the whole thing cost no more than one breakfast.

I didn’t become a programmer, and I still can’t explain what those complicated terms mean. But I now own my very own “Qingyan Labs” — and I know that I can keep filling it with whatever I want, whenever I want.

If you’re also someone who knows nothing about code but has always wanted a website of your own, then I have just one piece of advice from someone who’s been there:

Don’t be scared off by the word “technology.” What you need isn’t the ability to write code — it’s the willingness to spend an afternoon asking the right questions, and then trusting those Agents to get the job done for you.

The website is built. As for the stories inside it — like that trade, or that time the air conditioner got fixed — well, that’s what comes next.

---

Dictated by a beginner, written with the help of Hermes Agent (DeepSeek V4 Flash), and recorded on the day CyanLabs was built.