Although much progress has been made in software verification, software testing remains by far the most widely used technique for improving software reliability. Among various types of testing, developer testing is a type of testing where developers test their code as they write it, as opposed to testing done by a separate quality assurance organization. Developer testing has been widely recognized as an important and valuable means of improving software reliability, partly due to its capabilities of exposing faults early in the development life cycle. In this position paper, we present our positions on future directions of developer testing along four dimensions (which of course we do not claim to be complete): correctness confidence, specifications, (dis)integration testing, and human factors. Our positions are originated from two recent promising technologies in developer testing: parameterized unit testing and dynamic symbolic execution, also called concolic testing.