Saying I got into technology because I’m lazy and I like to do things the smart way, isn’t exactly wrong.
There is definitley a dopamine kick in pressing a button and watching an hour worth of work be handled on its own through a good system design.
But dopamine is addictive and the fastest way to waste time?
Automating something you don’t know how to do manually.
I remember being a junior dev looking at a mountain of pictures to annotate for a vision model I was training, and I told my senior dev
“There has to be a smarter way to do this?” and he said “yes - by just doing it”
And as annoying as it was he was right. Because I had to go through it to learn the process myself through lived experience once. Because when we later on built a pipeline to preannotate and outsource the annotation, I first hand knew which buttons to push.
Sometimes I feel awkward for saying the bluntly unsexy thing, because compared to do things manually until you have learnt what you need to learn which is the boring right answer, the guy who can offer you a 60-agent framework sounds like a drink at a beach.
Yet it’s the thing that has helped me the most building robust system - don’t take impatient shortcuts. Shortcuts are successful when you can lean on exisiting tested knowledge. When you are building a off a new project from an existing codebase - that’s a great shortcut.
But premature shortcuts in automation often happen because:
You don’t know what you need, and you try to conflate learning while automating.
And you can’t know for sure what you need without building it the first time without and MVP test of it - which is often fastest and cheapest done manually.