Ask HN: How to avoid over-engineering software design for future use cases?

I have worked at Microsoft and Google in multiple different teams. One thing I realized that sometimes engineers go extreme in designing things/code for future cases which are not yet known. Many times these features don't even see any future use case and just keep making the system complex. At what point we should stop designing for future use cases ? How far should we go in making things generic ? Are there good resources for this ?