My 2 cents: Go for (1), keep several potential future use cases in mind, and make damn sure you don't architect yourself out of those being natural directions f

My 2 cents: Go for (1), keep several potential future use cases in mind, and make damn sure you don't architect yourself out of those being natural directions for the code to go. Often this leads you to stub out an interface that looks like the beginning of a nice architecture, but not actually flesh out the code beyond what is necessary. This strategy generally requires more thinking than coding, but in my experience that is usually a good thing.