A system and method of distributed computation based upon a novel partitioning of computer networks. In the distributed computing system of the present invention, instructions are disseminated to select agents on a computer network, each of which is a member of at least one “path,” a path being an ordered set of one or more connected agents within the network. The computation is then performed by the agents along the path, and any results then returned. A computation to be performed is broken down into instructions executed on a unique ordered sequence of agents making up a path. This vastly increases the expressive power of a single network to be not only the sum of its parts, but the sum of all possible configurations of those parts on a graph. The basic components of the computing system are agents, probes and paths. The agents are interconnected forming a network and a path is an ordered sequence of agents. Information is exchanged between agents by the sending and receiving of probes between the agents in the network.