Lessons to learn how to make your code vibrate using AI like ChatGPT

What if you could automate your least favorite task at work?

Alphabet CEO Sundar Pichai and other technology CEOs are encouraging beginners to try their hand at vibecoding, which uses AI to generate code with natural language prompts and improve it through conversation.

Even though I have zero coding experience, I decided to try and create something that would be useful to me.

This project ended up having some twists and turns and took half a day to complete, but I learned some invaluable lessons about how LLM works and why there is a learning curve in understanding how to use AI.

task

My least favorite part of my job as a journalist is having to be on social media platforms that don’t bring me any joy.

Additionally, checking and monitoring the feeds of about a dozen tech CEOs’ individual X accounts is not an efficient use of time.

My goal for this project was to create a single web page where I could view people’s posts without having to log into the platform or access their individual feeds. First, I wanted to have three columns on the page, each representing President Donald Trump, Tesla CEO Elon Musk, and OpenAI CEO Sam Altman, and I wanted the columns to update automatically whenever something new is posted.

Let’s try it first

Around 2pm local time, I told ChatGPT that I had no coding experience and needed to start with the very basics. Next, we explained the task in detail.

Related articles

ChatGPT has created a simple plan that starts with downloading the software, creating a file, and launching a web page using Terminal, the MacBook’s command-line interface. It then generated a very long string of code and asked me to enter it into one of my code creation tools, TextEdit files.

The steps went smoothly until I actually started the program on my local web server and saw an error message in each feed column. Please follow ChatGPT’s advice to adjust the feed refresh rate and try the different links provided to read the feed.

After still getting the error I asked them to change the code. That’s when everything went into meltdown mode.

Every time you try to generate a new set of code to fix an error message, a new type of error message appears. And when nothing worked, ChatGPT asked me to perform a task that was way beyond my head.

The chatbot finally got tired of me and gave me a “Gateway Timeout” error message that completely prevented me from launching from the platform for the next two hours.

Please take two

At this point it was 7pm so I decided to ask for help. I messaged my partner, who majored in computer science and physics and has a Ph.D. And he said I was approaching this job the wrong way.

He suggested removing everything and simplifying the task to just getting one person’s social feed. He also said people should stop talking about ChatGPT as if it were a “knowledgeable toddler” and treating it as a search engine. And most importantly, instead of expecting instant results, you should try to understand what ChatGPT is doing and ask for an explanation of its solution, he added.

With his advice in mind, I described a new task: creating a page that essentially functions as a Trump Truth social feed, without actually requiring access to the platform. Instead of rushing into the task, I messaged ChatGPT and asked what was the best way to tackle this task.

It seemed like you were almost flattered by my question and said that you should first create a mock feed with fake posts to test if your code is correct before getting to the actual posts. I followed the steps to create a TextEdit file and named it. They then entered the provided code into the appropriate file and ended up with a mock feed containing three slides of fake posts.

ChatGPT then showed me exactly what code in the file needed to be changed to go from the mock version to the real thing.

At this point, I got the “Unable to retrieve feed” error again, but instead of opting for a more complicated solution, I was asked to try Safari or Firefox instead of the Chrome browser. I made the switch and after 30 seconds of loading, my first attempt at a vibe-coded web app was successful.

conclusion

Computer scientists are definitely right that even with the help of AI, if you don’t understand what you’re doing while coding, it’s obvious.

However, if you want to code a simple solution to a small but nagging problem, you may want to learn more about how ChatGPT communicates and research the best ways to facilitate it.

In my experience, I’ve found that coding with ChatGPT is most effective when you’re willing to do the prep work yourself and learn what ChatGPT is trying to accomplish, rather than expecting quick and easy results.

It’s also wise to scale up gradually, by running mock trials before switching to real data and gradually adding features once you get the basics working.

I hope my hours of vibe coding can save you some time every day.