Thinking Outside the Black Box

A worthwhile thought experiment:

If AI can one day quickly and reliably write new code and edit existing code to produce the intended new behavior of the software system, such that the system itself effectively becomes a black box… there's a whole slew of things that don't matter anymore.

Human readability, for instance. And if we're no longer optimizing code for the human who needs to read, understand, and change it later, what else goes with it? Code review. Design patterns. Modularity. Frameworks. Fights over syntax, or even languages. Who cares what the inside of that black box looks like?

There's a lot we always assumed was essential complexity of software engineering that was really just the goings on inside that box. And sure, it was the really expensive part of building and running a software product, but that doesn't mean it was the most valuable. The most valuable parts have always been what happens outside the black box. Time to double down on those skills.

Regardless what the future holds, this will always be true: high-quality software is both valuable to the business now and easy to change later.