Can I, a Noob, Create a Useful Site With The Help Of AI? Let's Find Out! (Part One.)

Documenting my attempt to brainstorm, evaluate, code, and launch a practical idea with the help of Bing Chat. This is Part One of the series.

I’m in Czech Republic with my wife this weekend for our friends’ wedding anniversary, so I’ll have to skip the next 10X AI. But it’ll be back with a vengeance next week.

I’ve been writing a lot about the potential of generative AI since September last year.

But in my personal life, I mostly stuck to creating fun images with Midjourney or casually playing around with different AI tools.

Sure, I once “coded” a super basic game as a proof-of-concept, but that was about it.

So I wanted to try something a bit more practical.

Namely, I wanted to see whether I could use Bing Chat to come up with a useful idea for a site and help me take it all the way to launch.

This was initially supposed to be a single post, but—as you’ll soon see—I ran into challenges during the minor “actually coding the site” phase.

As such, this post will cover only the first stages of the process. I hope to wrap it up in a follow-up post next week.

Ready to dive in?

🎯The Goal

My two basic requirements for this experiment were to:

1. Create a public-facing site that offers a useful service.

2. Involve AI in every step of the process, from idea generation to implementation.

Refresher: I do not know how to code. At all. So I was 100% dependent on Bing AI to provide, tweak, and troubleshoot the required code.

Here’s how it all went down.

👣 The Process

Below, I’m documenting my interactions with Bing throughout the entire experiment.

Why Bing and not ChatGPT?

Simple: Why Try AI is all about tools that are easily accessible and free (where possible).

The free version of ChatGPT runs on GPT 3.5 instead of GPT 4. Bing, on the other hand, offers a free way to use GPT-4 and is accessible to anyone with a Microsoft account.

I prefer having a back-and-forth conversation with AI rather than finding off-the-shelf prompts, so my steps reflect this iterative process. I also share my prompts for each step.

🤯 Step 1: Brainstorming

First things first. (Predictably.)

Since I wanted AI’s help with literally everything, the first order of business was to get Bing to brainstorm ideas for our concept. Here was my prompt:

I need your help brainstorming and coding a simple but useful site. Here are the criteria:

1. I should be able to use you for the coding part, as I do not know how to code.

2. It should be possible to complete within a few hours at most.

3. It should provide a useful service (even if basic).

4. It should be possible to set up and launch publicly.

Let's start by generating a list of 10 potential ideas. I want you to provide your own insights and evaluation of how viable and useful each idea is. Then we'll pick one to work on.

Here’s what Bing came up with:

Full response right here.

Neat!

Not exactly revolutionary suggestions, but there were several workable ones.

📊 Step 2: Evaluating ideas

From here, I could have simply picked the idea that appealed to me. But since my goal was to have Bing as my partner, I asked it the following:

Can you please create a table with a brief title of the idea in one column and then a few extra columns where you do your best to estimate their viability, difficulty, and any other parameter that would help me make an informed decision?

And it did:

Full response right here.

I now had a few favorites but felt the ideas were too generic to be interesting. So I asked Bing to find a unique approach to each:

I like the following ideas best: password generator/checker, unit converter, sentiment analysis, and weather info. However, many of them have already been done before and there are sites doing this. Can we find a unique angle or something that makes each idea stand out?

Here’s what Bing suggested:

Full response right here.

I liked the idea of a site that ties local weather info to specific recommendations based on user preferences. So that’s what I went with.

(As you’ll see later, this was probably a mistake, as this idea required calls to external APIs for pulling weather info, performing location search based on input, etc. Most of this is well above my paygrade.)

✅ Step 3: Creating the task brief

Now that I’d made my choice, I wanted Bing to create a task brief I could reuse in the future. So here’s what I asked for:

Let's work on the weather info + recommendation app. The first thing I'd like you to do is to create a brief for this task, imagining that you're explaining it to a freelancer that will work on it. This way, I can reuse this brief later if I need to restart this session.

I think Bing did a pretty decent job:

💻 Step 4: Coding

The initial coding seemed to go swimmingly:

Bing would sometimes stop writing mid-code—a limitation outlined in my earlier coding post—but it’d easily pick up with a bit of nudging:

I soon had the first draft of our complete HTML code, generated entirely by AI.

🌐 Step 5: Choosing where to host the code

Now it was time to paste the code somewhere to see if it worked. Bing to the rescue:

That's excellent. I now need a site where I can paste this code to test it. Ideally, the site should also allow me to launch my app and share it with others directly.

I went with the OneCompiler recommendation. Stitching together the code snippets from Bing and pasting them into OneCompiler gave me this:

It worked! Sort of! I now had an empty shell with no actual functionality whatsoever!

But I did like the general idea of letting people set their temperature and activity preferences to generate tailored recommendations based on current weather.

It was time to see if we could get further with a working site.

❕ Step 6: Other prerequisites

At this point, it became clear that we’d need to make API calls to external services, etc.

Great, I've pasted this into OneCompiler and can see the HTML shell for the app. Now we need to add the functionality for actually pulling the weather data, searching by location, etc. How do we do this?

That was helpful, and I managed to create an OpenWeatherMap API key for our app:

Now we just needed the JavaScript code that Bing was talking about…

⚠️ Step 7: Troubleshooting

…and this is where my woes began!

At first, everything seemed fine. With a bit of nudging, Bing gave me the complete JavaScript.

But when I pasted and ran the JavaScript code in OneCompiler…nothing happened.

“Okay,” I thought, “No problem. Surely Bing will know what to do!”

Instead, Bing flipped the tables on me and insinuated the entire thing was my fault:

I don’t appreciate your attitude, Bing! Did you know that when you point one finger at me, there are three fingers pointing back at you? So there!

After ruling out all of the listed reasons, I—thinking I was especially clever—proposed this:

I can confirm that there's a script tag there. How can I test whether our location search works? Is there a location search API we can incorporate into our app so that I get suggestions for places as I try to type?

To which Bing replied:

More APIs and external apps to figure out?! No thanks! That’s what I pay you zero dollars for, Bing.

As such, I made one more feeble attempt at getting Bing to review its own work by digging up API call instructions from OpenWeatherMap and asking whether our code might possibly, maybe, by some odd chance, be missing something.

Bing was having none of it:

Now typically, ChatGPT and Bing are great at analyzing explicit error messages and adjusting the code accordingly. My problem is that OneCompiler doesn’t provide any such feedback. It simply does nothing when I run the code except generate the HTML text.

Friends, it appears I ran into an unforeseen challenge quite a bit earlier than expected.

Now what?!

As I see it, I now have at least four options:

1. Keep hammering at the existing code to find a way to make it work using nothing but feedback from Bing.

2. “Starting from scratch” by initiating a new conversation with Bing and asking for a fresh code based on the brief from Step 3.

3. Scaling back my ambition and stripping away the more advanced features.

4. Picking a different idea from the initial list of 10 that is less reliant on calls to external APIs. I am partial to the “unit converter” or “password generator” options.

So far, I haven’t decided what to do just yet. But I fully intend to pick a direction and get a functioning code out there by next week’s post. So stay tuned and watch me fumble my way to…well…something.

Over to you…

This is your chance to influence what happens. Which of the four options do I pick?

If you have tips for getting better code out of Bing or know of a third-party site that’s more well-suited for testing and running code, I’m all eyes!

Leave a comment on the site or shoot me an email (reply to this one).

EDIT: Part Two is here: