Don't outrun your headlights
Design only for as far ahead as you can see
When driving at night through a completely dark place, it is not wise to go so fast that you cannot break to the point of fully stopping the car before anything becomes illuminated by the headlights; an animal, a tree, a broken vehicle – whatever –, you are safe when know that you can safely stop.
Something similar can be said about the craft of software engineering. It is not uncommon for us to get excited when thinking of ways to organise the folder structures of our project, of how to improve scalability early on, of how to optimise a database query that is unlikely to live longer than a bee, and so on.
We are engineers and we want to save us from trouble because we can see ahead. I understand. And I also understand that sometimes we are actually optimising for a hypothetical scenario that may never become real. If we need fortune telling, we can be sure that we are going beyond our headlights.