In this paper we introduce and investigate a new type of automata which turns out to be rich in deep and complex phenomena. For our model, a maze is a countable strongly connected digraph called the board together with a proper colouring of its edges (the edges leaving a vertex have distinct colours) and two special vertices: the origin and the destination. A pointer or robot starts at the origin of a maze and moves naturally between its vertices, according to a finite or infinite sequence of specific instructions from the set of all colours called an algorithm; if the robot is at a vertex for which there is no out-edge of the colour indicated by the instruction, it remains at that vertex and proceeds to execute the next instruction in the sequence. The central object of study is the existence of algorithms that simultaneously solve, that is guide the robot to visit the destination in, certain large sets of mazes. One of the most natural and interesting sets of mazes arises from the square lattice $\Z^2$ viewed as a graph with arbitrarily many edges removed (each edge corresponds to a pair of opposite directed edges), together with the suggestive colouring that assigns to each directed edge the corresponding cardinal direction. In this set-up, a research question of Leader and Spink from 2011, which proved to be very profound, asks whether there exists an algorithm which solves this set of mazes. In this paper we make progress towards this question. We consider the subset of all such mazes which have arbitrarily many horizontal edges removed but only finitely many vertical edges removed in consecutive columns, and construct an algorithm which solves this subset of mazes.