Before the Enlightenment, what we now call science was alchemy. Productive, sometimes brilliant, but without a formalized process for distinguishing real understanding from convincing error. The scientific method changed that, and science became possible.

Software engineering today is closer to software alchemy. We build through craft and intuition, and when it works we often can't fully explain why.

AI is making the labor of producing software cheap while leaving the work of understanding it hard, and that combination is new. We've never had to build a shared mental model of a software solution deliberately. It has always come for free through the raw effort of writing code.