There’s been no shortage of hand-wringing over how AI will forever change, if not completely consume, the software development profession. This isn’t an obituary for the profession I love, but as a practitioner and AI-first convert, I think the writing is on the wall for something else: the concept of taste as it applies to the craft of coding.

For folks not directly involved in building complex software systems, this may be hard to understand, but by my estimation at least one-third of the expense in software development and maintenance comes down to differences in individual taste.

As a software developer, I’ve spent thousands of hours acquiring, refining and expressing my own tastes. It’s a normal part of the mastery process. Tastes are how we compress the cognitive complexity of the job—a way to feel productive and in control. Shared tastes let you find your tribe, develop shorthand with your in-group and “other” the out-groups with conflicting preferences.

Nerd taste is 100% about the HOW, 0% about the WHAT.
Let me give you examples. I’m talking about preferences for one programming language over another, frameworks, methodologies, conventions, architectures and paradigms. Node.js/GraphQL/NoSQL vs. REST/Spring Boot/MySQL. Monoliths vs. microservices. Mono-repos vs. multi-repos. Containers vs. serverless. Emacs vs. VI. VS Code vs. IntelliJ. Azure vs. AWS vs. GCP vs. the legacy data center hardware you’ll have to pry from someone’s cold dead hands when the CFO finally forces the shutdown.

It comes down to: “ABC will get the job done, but I really prefer XYZ.”
Don’t get me wrong; there are often legitimate technical reasons for one decision over another. But if we nerds are honest? It’s mostly about what we like better. And if we’re really, really honest, we can point to countless projects that were late or over budget due to differences in individual taste.

So why is it time to mourn the death of code nerd taste?
The simple answer: quality results rarely depend on one specific set of tastes, and LLMs really don’t care about yours. More to the point, AI coding assistants are so powerful that it’s more practical to surrender to THEIR tastes than to coach them into “enjoying” coding “your way.”

I can now use #AI tools to produce amazing results in programming languages I have minimal experience in. I can have them hammer out excellent user experiences using what I previously would have dismissed as “spaghetti code.” The HOW is becoming less relevant to me every day.