Solving tasks written in ePL, as of now, boils down to exploring a large and unknown search space. Computation nodes randomly pick parts of the search space and scan them for solutions. Since, for the sake of fast speeds, there is no task locking or synchronization and no nasty bottlenecks resulting from slow nodes holding up everything. With a sufficiently large search space, the likelihood of multiple nodes working on the same subspace is negligible.
The search space is spanned by XEL’s pseudorandom input generator providing 320 bit of entropy for the formation of a solution candidate. This random data can be used to “walk through the search space” – for example, try out different paths through a network or different assignments to a CNF formula. When solutions are found, they are collected and stored on the blockchain. An iteration support allows tasks to perform a round-based exploration as found in Genetic Algorithms.