The goal of the ISIS projest is to provide a high-level support for fault-tolerant distributed computing by automatically replicating data and code. The extent to which information is replicated and the physical location of information are not specified directly by the programmer, but are instead inferred from a specification, which looks much like a conventional program in an object-oriented language. This novel approach to fault-tolerant software construction requires much less sophistication from programmers than current alternatives. Moreover, optimization techniques that would be too complex for implementation in general purpose applications can be supported by the ISIS system. This overview discusses the goals of the project, its current status, and some of the implications of our work.