Doing the opposite works for me. I always start out defining the ideal API I want to solve a problem and then solve the problem using the not-yet-implemented A

Doing the opposite works for me. I always start out defining the ideal API I want to solve a problem and then solve the problem using the not-yet-implemented API. I then implement the API recursively by adding lower level API’s that makes it easy to solve the higher level API problem. And most of the time I already have a lower level API from a previous project I can easily reuse and perhaps expand. So I don’t start with the database moving towards the biz logic. I start with the biz logic moving towards the DB (if needed). It works really well. I think this is similar to how experienced Lisp developers work. Starting out defining a new “language” (a bunch of functions) that makes it easy to solve the problem and then recursively implementing it. However I work in C++ and Typescript not Lisp.