Overview
What was the project?
This was just meant to be a fun, quick experiment to see I could create a website that some fun retro games in my free time. The entire site was built in under 1.5hrs.
Where I Started
ChatGPT
From my previous experience using different prompt coding tools such as v0, Cursor, and Windsurf I've found that when starting off things go much quicker and smoother when you begin with a well thought out product requirements document.
​
I kicked this off using ChatGPT and asked it to:
"Generate a list of top 10 classic 8 or 16 bit video games from the 80s and 90s."
​
Naturally, it served up legends like The Legend of Zelda, awesome, but maybe a bit too ambitious for a quick project.
So I adjusted the scope and prompted:
"Update the list to games that could realistically be built using v0."
​
That got us into more manageable territory, and kept the build light, fun, and totally doable.
My next step was to get some help with game names to avoid any copyright infringement:
Lastly I asked ChatGPT to generate me a requirements document to be used in v0, and explained some of the styling and functionality that I wanted it to have. You can view the Requirements Document here.
Prompt Coding in v0
80/20
With all of the projects I've prompt coded, I've found one consistency; after entering a well thought out requirements document, the initial few prompts will get you 80% there, in the initial 20min. Then you'll spend the next few hours, days refining and adjusting either by prompting or adjusting code manually to get everything how you want it. One important note... anytime I build something in v0 or similar tools, I treat it as a prototype. These tools are amazing for speeding up early exploration, but they’re not a replacement for thoughtful design, human feedback, or proper code review. This project? Just a fun experiment to see how quickly I could spin up a retro-inspired site with a little help from AI.
The Prompting Begins
I started off simple, just by uploading my requirements doc and giving the same context around what I was looking to build. When building actual prototypes I would start off by adding rules, personas, setting up the type of application, fonts etc. I wanted to use. For this purpose the document was enough.
​
The output was a solid first attempt, but had only one working game.
​
In order to speed up the site creation. I decided to reduce the amount of games to 5 that I thought would be the easiest to build. Once I got to that point, it was just a matter of prompting it to build out each a game, giving it a max of 3 levels each so to not overcomplicate things. I worked with the tool a bit to get the styling I wanted, by referring to certain games and adding images.
Once it got it close to where I wanted it, I decided to jump into the code and fine tune it, adjusting buttons and styling. One thing to note here is that I have a decent understanding of html/css, but I am far from a developer. When prompting, the tool gives you feedback letting you know what it just changed. This actually can help you identify in the code what I should adjust further to get things where I want them. In this case I was trying to figure out why the screen looked washed out. The tool told me it adjusted the opacity of the "retro-grid". I was able to search for this in the code and then make manual adjustments to finish the styling.
Reflection
Prompt coding tools such as v0 are fun to use and it's amazing how quick you can turn an idea into a reality. It's important to remember that the output is only as good as the input. For realistic design applications, nothing can replace the in person interviews, mappings, and any other discovery that is done up front to define the actual problem needing to be solved. It's easy to just jump into these tools with an idea and get an amazing looking output without solving an actual problem. We should be using these tools as just that, tools, that help to accelerate our design process. And any outputs from these tools should be validated.
​
With that being said they are also fun to play around with in order to build quick sites and applications.