Optimistic Algorithms make assumptions about likely outcomes of future events. Instead of waiting for an optimistically predicted computation to complete, the program continues to execute using the predicted value. Information about the correctness of the prediction is piggybacked on messages, and in the event of a misprediction the distributed computation is rolled back to a consistent state before the misprediction, and forward execution resumes with the correct value.

The most basic optimistic assumption is the absence of failure. Optimistic Recovery is the application of this optimistic assumption to the creation of fault-tolerant systems, and provides an alternative to traditional transaction-based fault recovery techniques.

The project produced a variety of results ranging from theoretical foundations of optimistic algorithms, to studies of the applicability of optimistic techniques to the Unix filesystem, to an implementation of optimistic fault recovery on the Mach operating system.