Prompt Engineer: Should I Keep Doing This Job?
Honestly, I see it negatively. I say this as someone who works in the field, but I don't think "people who are good at writing prompts" will last long as a profession.
There are three reasons.
1. The better models get, the less value prompt tricks have. I've seen many tricks that worked last year become default behavior this year.
2. It becomes fragile the moment it's deployed. Optimized prompts are tightly coupled to a specific model version. Even just changing the model means you have to rework them.
3. In the end, what remains isn't technique but domain understanding. The person who knows what data to put in, in what order, and what to validate wins.
So these days, what I do isn't polishing prompts but building evaluation sets. If you can compare results quantitatively, you can respond even when the model changes.
Of course, I'm curious about opposing views. How do those of you who are actually working in this field see it?