AI has changed how we ship software.
As software engineers our discipline is barely a lifetime old, keep the lessons we’ve already learned.
Before we celebrate code that no one writes or reads anymore, it’s worth examining how many of us got to write code for a living.
The majority of us veteran engineers fell in love with code alone at a screen, creating new realities with our minds and fingers. I fell in love with the process of learning and building new games and systems that I couldn’t imagine being possible days, months, or years earlier.
We now build at a pace our younger selves couldn’t have imagined.
I believe reality will land somewhere in the middle, between pessimistic resistance and blind overoptimism.
Overoptimists believe unlimited tokens and the latest models will let us stop reading and reviewing code. They picture a world where agents review production code and nobody understands the codebase, all for speed.
The pessimistic engineer is still handcrafting all their code and rarely using the latest models to “optimize” their workflows.
Both camps agree our profession is shifting.
Both believe they produce quality production code that drives real value to their companies.
The business doesn’t care about who is technically right. Whoever costs less wins, as long as throughput and quality hold.
Our goal as software engineers is dependable code that keeps the business running.
As an engineering collective we’ve realized our business counterparts can’t use the models to build real-world software. They’re accepting that reality as well but now scrutinizing their teams’ AI budgets: what did the spend deliver?
Challenge and reshape everything we know instead of discarding it.