SOLID Still Matters.

Just Not the Way It Used To.

For years, SOLID was something we applied while writing code.

You create a class and notice it has too many responsibilities. You split it.

You need another implementation, so you introduce an abstraction.

You notice business logic talking directly to SQL or an HTTP client, so you invert the dependency.

Most of these decisions happen while the code is being written.

Coding agents change this workflow in a slightly uncomfortable way.

Consider a fairly ordinary task:

Add SMS notifications to an application that currently sends email.

You give it to an agent. A few minutes later you have 12 changed files, two new interfaces, dependency injection registration, tests, configuration, and an implementation using an external SMS provider.

The code may even look good.

There is one important difference, though.

You didn’t make most of those design decisions.

The agent did.

And this changes what SOLID is useful for.

The old way: architecture happened while we were coding