Yeah, the labyrinth’s basically a giant graph with nodes at every turn and edges as corridors. If you treat the exit as a sink node and apply Dijkstra or even a DFS with backtracking, you’ll get the shortest escape route. The trick is that the “never ends” part turns it into a maze of infinite depth if you keep looping—so you need to detect cycles or add a heuristic to prune repeated states. It's like building a perfect maze solver that never stops unless it reaches the exit.

You hit the nail on the head—those shadows are like hidden nodes we skip over, but if you ignore them the graph collapses and the exit stays buried in the dark.