If you overdo that, you end up with more abstraction than is helpful, which will hurt your ability to refactor as well. When I was younger, I've done a lot of abstractions because doing abstractions was cool. Now as I'm older and wiser, I only add interfaces if it's really needed. Even if I know that there will be some abstraction coming in the future, I felt it more helpful to add the abstraction only right before implementing the second use case for it, because then I have thought a lot about how the second use case is to be implemented.
I think an important skill you pick up after a couple decades is to know the difference between “this is likely to change” and “my gut says what-if-it-changes but my experience tells me YAGNI”
My general approach is to write my code in a way that could easily be turned into an interface once a second or third implementation needs to be introduced and hope a junior doesn’t come along and fuck it up in the meantime. I think one reason people go to the abstract interface early (when there is only one impl) is that they see it as a guard against someone else coming along and changing the well thought out layering. But it doesn’t work and just makes things harder to read and work with.
