Comment by interstice

There are dragons here. The hope, presumably, is to be able to describe a solution at a high level and have an AI figure out the details instead of a developer (or detail oriented) person. The problem I can see with this is ownership over those details, AI can't currently explain its reasoning or react too well to issues that crop up with its own previous decisions, since it's usually a 'fresh' instance for each prompt. In other words it might know how, but it can't know why. Or at least not without retracing every step every time.