I remember when the term prompt engineering first came out. It was kind of a joke, right? The idea that prompting could be a skill?
It was a joke for two reasons. One was the idea that one person could be significantly better than someone else at something like prompting seemed kind of silly. And 2.
It just seemed kind of obvious that eventually it wouldn't matter how good you are at prompting, the models would just become good enough to understand what you needed.
However, it actually seems like the opposite is happening. I think anyone who spends a sufficient amount of time with chat, T, or other large language models has realised that prompting really is a skill. And the smarter these models get, the more it matters how good your prompts are now.
Sure, if you just have a simple question or request, it doesn't matter too much how good your prompt is. In fact, it might matter less than it used to.
However, a lot of people don't realise that these models can now produce the kind of output that people used to pay good money for. And the reason they don't realise it's possible is because it's difficult. Writing prompts that reliably create high quality output is a skill, and the skill cap goes up every time the models capabilities improve.
As it stands, this trend doesn't show any signs of slowing down. The better the models get, the more it's gonna matter how good you are at prompting. Prompt engineering is a skill.
Context engineering is a skill. And they're skills that I think most people should be developing, at least to some degree.