The problem is that if both your interface and implementation are variables where AI can make arbitrary changes with full agency and little verification, maintenance becomes tricky as your understanding of the system as a whole dwindles.

It’s better to be on top of the interface than the implementation. But at the same time, brittle systems caused by overly rigid interfaces are a real thing. In that case, interface exploration can be done by a human with assistance from LLMs, but allowing an LLM to arbitrarily explore and make changes to the interface sounds like a recipe for disaster.