Early development: Entities randomly placed and moving around randomly.
Later development. All test entities. Green people are "ghosts" and hunt normal people (white.) Yellow people are "ghost busters" and kill ghosts. The yellow circles surrounding each ghost buster represents his neighborhood region in which he can detect ghosts. Ghosts with small yellow rings around them are potential targets; those with red rings around them are targets selected by a ghost buster.
Flocking steering behavior has been added to normal people; you can see them beginning to travel in groups. Triggers added, here in the form of "bombs." The ghost busters have organized and are now controlled by a central GhostBustersHQ class that directs and coordinates their attack. As a result, a smaller number is now much more effective against a larger group of ghosts (though as you can see here, this time the ghosts appear to have the upper hand.) Game events and updates from the GBHQ now appear in the Comm window. Optimizations allow for many more game entities than before.
Here is a version of an "automatically generated" navigation mesh that I did by hand as I was designing an algorithm to do this.
Here is the final result of the algorithm.
Primitive navigation using navigation meshes and the A* algorithm with some path smoothing (path shown):
Map editor created. Building a basic map here:
Navigation mesh generated for this map:
The editor allows test entities to be generated and a live simulation of the world to be run without opening the actual game.