This is a genetic programming (GP) system inspired by the grammatical evolution (GE) technique.
Unlike a classical GE system, this one uses a subtree crossover and a subtree-local mutation operator, thus eliminating the destructive effect of (naive) GE genetic operators, and always operates on valid pruned genotypes.
Unlike a classical GP system, this one operates on linear genotypes mapping them to executable phenotypes by applying a grammar-based transformation.
Santa Fe Trail problem is used as a test problem during the development of the system.
Here is a short video demonstrating the system in work! (and no, it does not always find a solution so fast; somewhere around six to eight out of ten times it doesn't find one at all)
$ swift run