What I Realized While Looking at the Traits of People Who Are Good at Prompting
I've been experimenting with all sorts of prompts at work, and I've noticed that people who are good at it have some things in common.
First, they decide the format of the desired output up front. They set constraints like "give it to me as a table" or "summarize it in three lines."
Second, they don't try to get it perfect in one shot. They get a draft and then refine it, iterating as they go.
Third, they kindly explain the background knowledge to the model, as if explaining it to a person.
In the end, being good at prompting isn't about handling AI well—it's about the ability to clearly organize requirements. People who are just good at communication end up being good at it.
6 answers
Agreed. People who can't organize requirements can't write prompts either lol. In the end, good writers win.
But isn't this a bit of an old take? These days, models can produce good results even if you throw together a rough prompt. Of course, it'd be different for complex tasks, but generalizing it as 'communication skills' seems like a stretch. I feel like as tool performance improves, prompt engineering itself is heading toward disappearing.