The assumption in software engineering circles has been that if AI lets us write code faster, we will build more things and move faster.
That assumption misses a step.
Writing code is not the same as integrating it, maintaining it, or trusting it.
We are starting to discover that verification does not scale the way generation does.
I argue that the engineer who thrives in this environment will not be the one who can prompt the largest code output. It will be the one who can look at what comes out and reliably decide what survives.
This post is about why that skill — of refutation — is about to become central, and what it means for how we build software.
1. Volume is no longer a signal of value.
We are in the middle of a real GenAI explosion.
Code, images, video, documents — all of it can now be generated at a cost that is rapidly approaching zero compared to the historical human-labor baseline.
In software engineering especially, the ability to write a lot of code has traditionally been associated with productivity and even creativity.
... continue reading