You can vibe code with your favorite “AI agent bro” all you want, but you need to be in charge and know what to ask it to type.
Are you going to ask it to add tests?
Not if you don’t know what tests are.
Are you going to ask it to do a TDD pair-programming session with you?
Not if you don’t know what TDD or pair-programming are.
Are you going to ask it to write blog posts for you?
Not if you don’y know what blogs are.
Unlike a car, where you only need to know how to drive (like that’s easy), in software engineering you need to be able to judge what “bro” is spitting out. You need to be able to know how you are going to test (and I don’t mean write unit tests) what the thing spit out. You need to be able to tell (somehow) that what the “malevolent genie” materialized is what you asked it to, or better yet what you need.
Anyway, maybe eventually AI will be able to churn out perfectly functional, secure, clean and shiny code.
But what do I know. Want to know the future? Hire a witch!