I think it's really limiting to assume that procedurally generated worlds need to be sandboxes as I don't think that's necessarily true. Sandbox games tend to be the more popular and obvious use of procedurally generated content, but it's not like the technique is bound to that genre. Many Roguelikes use procedurally generated content of some kind, and they're not really sandbox games. On the contrary, they're games with goals, challenges, and balance. They generally have set goals that are set by the developer and are not very dynamic, but the stuff between the where you start and your goal can often involve a lot or random content and procudural generation. Some genre's that aren't really sandbox games have been using some form of lightweight procedural generation for a long time, even if we don't often call it such. This is fairly common in ARPGs or "Diablo-alikes". Some strategy games also apply random or procedurally generated content as well X-Com: UFO Defense being a classic example. As I mentioned above, you don't necessarily have to pair procedurally generated worlds with procedurally generated goals. It's not really a black and white thing, a lot of games use some kind of random or procedurally generated content to varying degrees, and the games we call "procedurally generated" are just games that use more of it. On the topic of goals, there are a variety of different ways to make goals dynamic while still keeping them as goals. Maybe you don't let them emerge purely out of a simulation, but you can definitely do better than writing everything by hand. Generally a game is going to have a single, very general, and somewhat vague overarching goal that defines the overall "plot" the game. Unless you're making some kind of pure sandbox game or something like that you're always going to have that. It's probably going to be something like "save the world" or maybe "win the war" or something really general like that. However, the sub-goals that lead you to that ultimate goal don't really have to be as set in stone. I've already mentioned Roguelikes as doing this quite a bit, but there are some other specific examples outside that genre as well. One example is how Soldak's Din's Curse and Drox Operative handle goals. They have "quests" like any other RPG and these quests have a lot of elements that are written by hand, but the way they use the hand written elements is a bit more dynamic. For example, Din's Curse has the idea of mobs grouping together occasionally gaining higher status sometimes by killing each other. When mobs gather on a given floor and become organized, that is where "kill this group of mobs" come from. When mobs kill or overthrow each other and gain status, that's where "kill this boss" eventually comes from. When they eventually gather enough to come to the surface and attack the town, that's where "protect the town" comes from. This by itself is relatively simple as you have a set of pre-made events that may occur randomly, and that by itself that isn't all that interesting. However, these games do something else to take this idea even further, they arrange these events in chains of "cause and effect" that may affected or interrupted by the player or each other. This means that what starts as a series of pre-made events that get put into the world randomly can eventually mix with each other and what the player does to create even more situations. For example, maybe the RNG decides that a mob kills another mob and "levels" up. This is more or less arbitrary to start with, but this sets off a series of events that may eventually turn out different depending on how other events turn out. If the player chooses to ignore this mob, or simply doesn't notice this mob, this mob may eventually become more powerful as long as it doesn't get killed off by the player or another mob. If this happens, and it might not, this mob may eventually become a boss. A boss might make a doomsday device, they might gather an army, they might send some of their minions to the surface to cause trouble, or they might all of those as well as a few other things. Maybe the player will pay attention to the new boss and kill him off, or maybe they'll be too pre-occupied with other things. Maybe the boss may be overthrown by another new boss, and the army changes mob type or dissolve completely. Now when you have many of these "cause-and-effect" chains going on and branching off at the same time, with things like mobs organizing and causing uprisings, food poisoning raising goods prices and killing the populace, merchants getting killed or setting up shop, and random storms/curses, you can get a lot of mileage out of what originally starts out as a relatively small number of well understood pre-made events. It's up to the player to attack these problems and fulfilling a set of "victory conditions" to achieve the large overarching goal. Another somewhat less structured way of doing things is to not try to name "goals" or "quests" all that specifically, but only explicitly name a large overarching goal and dynamically put things in the player's way that create interesting situations. This way, short term sub-goals emerge as the player plans his way towards his ultimate goal. This is more or less how some roguelikes and some strategy games work. To name a specific example. AI War: Fleet Command does this a lot. In that game you always have the goal of defeating the evil AI, but the stuff that gets in your way is subject to a lot of random number generator as well as elements that interact with each other to create even more unusual combinations. You may encounter an enemy that is powerful against a certain unit type, or might do something very bad if you attempt a certain action carelessly, and a variety of these might be found in the same system where their effects overlap. Maybe the AI might decide to attack or do something else on top of what you are already encountering and cause problems that overlap with those you already have. Again, you still keep an large and very general overarching goal. and you have relatively simple pre-made objects and behaviors that can interact with each other to multiply the amount of variety you can provide with them. AI War: Fleet Command combines this with a variety of options for starting conditions, available units, AI behaviors, and in-game events. This allows the player to set up a variety of games that have a tendency to behave differently on top of the variety the game provides already.