prompt engineering is a temporary adaptation, evaluation engineering imo is the future (for now). trends of companies like prism eval, rolific, telus, outlier AI, mechanize, other frontier AI labs (anthropic/open AI etc) be more leaning to people outside technical backgrounds is quite interesting. eval testing and human decision making is extremely difficult and requires good logic ofc, but it's funny to know the original meme is pretty much ironic now. This isn't to say my reasoning is brand new, but it was initially difficult to put a finger on the analytics; yet now with the stats and resources it's becoming more obvious to where things are shifting :)