I have to start by saying that I'm a big AI user, I've been using the first GPT models, I'm also a day one user of ChatGPT itself.
What AI gets wrong
Whenever I started a project, back when I was younger, I thought a project was all about writing code, so I just opened up VSCode and started writing and writing, with ChatGPT on the side doing the heavy lifting.
But as I moved onto larger and larger targets, I found myself stuck in my earlier design choices, which weren't even choices, I was just thinking:
"Alright let's write X and Y then move onto Z"
And I never stopped to think:
"What exactly am I doing?"
And right around the end of 2022 I wanted to make my own React component library
(which is an older version of what my portfolio is made out of)
And for the first time I tried taking inspiration from a friend of mine (a senior software engineer):
- Taking a pen & paper
- Writing my target
- Writing my starting point
From there I made a big map of all the steps and choices I'd have to make before even writing a single line of code, stuff like:
- Folder structure
- Efficency
- Common utils
- No-dependency implementations
- CSS research
And that's where I was finally able to implement something meaningful and persistent, that's were the current "OhMyGUI (OMG)" component library was born
AI certainly made things easier to develop, but it never quite got this step right, it's not about code, it's about the idea that I'm trying to build, and this is for me the most important step, and I believe that engineering is about this, not about actual building, but planning what to build.
Random quote just to set the idea:
"We don't need a bridge, we need a working bridge."
This article is ofc based on a real story :)