Documentation without accurate and descriptive method/member names is much more harmful than the inverse. If an abstraction is sufficiently complex to warrant a lengthy description of why it exists, then it should have a design doc. In practice, most code within a repo is pretty simple in what it accomplishes and if it's confusing to a reader, then it is most likely because they don't understand the design of the larger component or system or simply because the implementation is poor. There are of course cases where comments are really useful or even necessary (e.g. if going against best practices for a good reason or introducing an optimization that is for all intents and purposes unreadable without explanation), but they are exceptions.