Doing things that could be in principle automated by AI is still fundamentally valuable, because they bring two massive benefits:
- *Understanding what happens under the hood*: if you want to be an effective software engineer, you need to understand the whole stack. This is true of any engineering discipline really. Civil engineers take classes in fluid dynamics and material science classes although they will mostly apply pre-defined recipes on the job. You wouldn't be comfortable if the engineer who signed off on the blueprints of dam upstream of your house had no idea about the physics of concrete, hydrodynamic scour, etc.
- *Having fun*: there is nothing like the joy of discovering how things work, even though a perfectly fine abstraction that hides these details underneath. It is a huge part of the motivation for becoming an engineer. Even by assuming that Vibe Coding could develop into something that works, it would be a very tedious job.
When students use AI to do the hard work on their behalf, they miss out on those. We need to be extremely careful with this, as we might hurt a whole generation of students, both in terms of their performance and their love of technology.