Raw LLM Responses

i think that "if you don't learn it you'll get left behind" and "it's advancing at such a rate that soon, incredibly large-scale tasks can be done with a single prompt" are fundamentally incompatible viewpoints. if it really does become this powerful (and i'm not saying it will or won't), part of this power will be its ability to enable productivity without a learning curve. sure there's a productivity difference NOW as people learn about worktrees and parallel agents and all that. but in this theoretical future, any productivity gains stood to be enjoyed by a skilled dev would already be accounted for by the ai. if you tell me you can accomplish X task optimally by managing N agents configured to Y setup and teaching them Z skill, shouldn't this theoretically sophisticated future AI already be able to reason that this is the fastest way to accomplish this goal and bake that into its proceedings without you needing to do it yourself?