[YeGoblynQueenne]

Yeah, it's confusing. The standard maze generation algorithms start with a fully-blocked grid where each cell has four walls, then proceed to punch holes through one wall at a time. This is basically the same procedure as solving a maze, except when solving a maze you only move between cells that already have holes punched between them. So a solver is really a generator run backwards.

Think of it as an automaton: it can both accept and generate a string. The string here is the path through the maze.

But I think the main, shall we say, contribution of the article is the method they use to find the farthest two points on a maze, that they then set as the entry and exit point of the maze. It's at the end of the article, after the discussion of their preferred method to generate a maze.