It is often the case that in game AI and other entertainment AI fields (as well as AI in general), we build prototypes in a toy world. The toy world is rich enough to emphasize the particular advances of our research, but simple enough not to involve more system building than necessary. Some ask whether we should continue to build prototypes, that are playable but not fully engaging in the way complete games are, or whether we should build more complete games that are demoable. This ties into business models.