Personally, I both agree and disagree with this.

1. I do think changes should be more carefully weighed than they sometimes are
2. I definitely don’t think that’s why. The biggest (but not only) reason we don’t handle UI in this same way is that if changes happen the computer isn’t capable of relearning how to do things in that moment, and people 100% are. We should provide affordances, communicate how things can still be accomplished effectively, and more carefully weigh the cost of making changes, AND ALSO people are capable of learning how to do a thing again when presented with changes

I’d also add, my perspective as a more-design-than-technical person is there’s way more material benefit to making those changes with people than with APIs. If the api provides the needed functionality, a computer can access that functionality. But people aren’t robots, they have to learn, rather than be taught to interact with the interface that we give them, and making it more organized, more visually clear, creating better affordances, etc, has much greater immediate benefit than I’d assume refactoring or updating an api in some way

(I also think there may be an interesting conversation to be had about whether UIs are better conceptualized as being akin to tools, or storefronts. Both are interactive, but some interfaces are obviously tools one masters (creative/productive software), and other interfaces are more like places we visit that we might reasonably expect to change sinage, reorganize, optimize how things work, etc (which would be true of many websites) or any mix of the two!)