What to build first in an open world

Almost everyone starts an open world by making the world. A big terrain, a coastline, some mountains, a river, and then the slow filling-in of forests and villages. It feels like the obvious order, because the world is the thing the genre is named after.

It is the wrong order, and the reason is simple. A large empty terrain answers no questions. You cannot tell from looking at it whether it is fun, because nothing about an open world is experienced from above. It is experienced at eye height, at walking pace, over a distance, and the only way to learn anything about it is to be inside it going somewhere.

Worse, terrain is expensive to change once anything is on it. Every road, building, spawn point and quest marker pins the landscape in place. People end up defending a map they made in week one for the rest of the project, because moving the mountain now means moving forty other things.

So the map is not first. What comes first is the thing that happens on it, and in an open world that thing is movement.

The single highest-value first task is a hundred metres of ground, however ugly, that you enjoy crossing. Not a level. A strip of terrain with some slope in it, a few rocks, a bit of water, and a character who moves across it.

Then spend a genuinely uncomfortable amount of time on how that movement feels. Acceleration and stopping. Whether the camera lags or sticks. Whether the character steps up a small ledge or is halted by it. How steep a slope has to be before you slide. Whether there is a run, and how the run relates to the walk. Whether jumping is a real verb or a formality.

The reason this comes first is arithmetic. In an open world, the player will spend more time moving than doing anything else. If movement is a chore, every single thing you build afterwards is separated from every other thing by a chore, and no amount of interesting content compensates. Games with unremarkable content and superb movement get played for hundreds of hours. The reverse is rare.

And the test is cheap and honest. Walk from one end of your hundred metres to the other, twenty times, while thinking about something else. If you find yourself taking the long way round for fun, or hopping up rocks because hopping is pleasant, you have something. If you are counting down the seconds, fix it now, because it will not get better when the distance is two kilometres.

When you do start on the world, build a small piece of it properly rather than a large piece of it roughly. One square kilometre is a good target and is larger than it sounds: at a normal walking speed it takes several minutes to cross, which is enough to contain a real journey.

Working small forces you to confront density, which is the actual design problem of the genre. How often should the player encounter something? What counts as something? How long is too long between them? Those questions have answers you can only find by walking, and you can only walk a piece that exists.

A small piece also stays cheap to throw away. If your density is wrong at one square kilometre, you rebuild an afternoon. If it is wrong at a hundred, you have a fortnight of work you cannot face redoing, so you keep it, and the game ships with the wrong density everywhere.

The other benefit is that a finished square kilometre is a standard. Once one region genuinely works, you know what "done" looks like, roughly how long it takes, and what has to be in it. Estimating the rest of the game becomes multiplication rather than optimism.

Inside that kilometre, build in order of how far away a thing can be seen.

Landmarks first: the things visible from across the region. A tower, a broken bridge, a strangely shaped peak, a column of smoke. These are what make an open world navigable, because they let the player answer "where am I" and "where do I want to go" without opening a menu. Breath of the Wild is the standard reference here for a reason: the world is organised so that from almost any high point you can see several things that make you want to walk towards them, and having walked, you can see several more.

Then paths. Roads, ridgelines, river banks, valley floors, anything that suggests a direction. Paths are how you communicate the shape of the region without instructing anybody. A road that curves around a hill tells the player there is a reason to go around, and the player who cuts over the hill instead is making a choice, which is the whole point.

Then buildings and detail, last. Detail is what everybody wants to do first because it is the satisfying part, and it is the part that matters least to whether the region works. A region with three good landmarks and clear paths and no detail at all is already playable and already tells you the truth. A region full of beautifully dressed buildings with no legible structure is a maze.

The core loop of an open world is small and worth writing down explicitly: the player sees something, decides to go to it, and finds something there. Everything else is an elaboration of those three beats.

Test all three separately, because they fail separately. Sight fails when there is nothing distinct on the horizon and everything looks the same in every direction. Going fails when the trip is boring, blocked, or so long that the intention wears off. Finding fails when the player arrives at the interesting-looking rock formation and there is nothing there, which is the failure that does the most damage, because it teaches the player that looking around is not worth doing.

The third beat does not require much. A chest is fine. A view is fine. A small piece of story told through the environment is often better than either. What matters is that arriving is not nothing, consistently enough that the player keeps forming the intention.

And once the loop works in one region, protect it. The most common way open worlds go wrong later is that the map fills up with markers, so the player stops looking at the horizon and starts looking at the compass. The loop is still technically present, but the player is now following instructions rather than making decisions, and it is a completely different experience.

Once movement and one region work, the temptation is to make more region. Resist it a little longer and build one instance of every kind of content the game will have.

One enemy camp. One dungeon or interior. One settlement with a shopkeeper. One traversal puzzle. One collectable. Whatever your list is, one of each, in the region you already have.

This is the fastest way to find out which of your content types is much harder than you assumed. Interiors usually are. Anything with a human speaking to the player usually is. Anything that requires a save state to persist usually is. Discovering that when you have one of each is a scheduling adjustment. Discovering it when you have twenty of the easy kind and none of the hard kind is a crisis.

It also lets you see them together, which is where the real problems live. Two content types that are individually fine can be redundant, or can want the same space, or can pull the player in directions that conflict. You cannot see any of that when they exist in separate test scenes.

Open worlds are unusually dependent on their opening, because the player has to be taught how to want things. In a linear game the next objective is given. In an open world the player has to generate their own, and nothing in their previous experience of games guarantees they will.

So the opening has a specific job: it has to demonstrate the loop. Put the player somewhere with a clear view, make one thing on the horizon obviously interesting, and make sure that walking to it is rewarded. Do that once or twice and the player has learned the rule of your world. Skip it and you get the common failure where someone stands at a crossroads for a minute, picks a direction at random, finds nothing, and concludes the game is empty.

Building this early also protects you from a subtle mistake. If the opening is designed last, it gets designed by someone who already knows the world intimately and can no longer see what a stranger does not know. Every assumption is invisible from the inside.

Fifteen minutes is enough. You are not building the tutorial or the first act. You are building the demonstration, and it can live in the same square kilometre you have already made.

Only when the walk is good, one region works, one of each content type exists, and the opening demonstrates the loop, is it time to make the world large. Then be ready for the things that only appear at size.

Distance changes pacing. What is a pleasant two-minute walk becomes a tedious eight-minute walk, and the answer is usually not a faster vehicle but shorter distances or more on the way. Adding fast travel is the reflex, and it is worth being careful: fast travel makes the world convenient and simultaneously tells the player the world is something to skip. Games that gate it, earn it, or make it one-way are all making the same trade deliberately.

Repetition changes meaning. The camp that was a small pleasure the first time is wallpaper the ninth. Variation helps a little; what helps more is having fewer of them and letting the space between be genuinely empty. Emptiness is not a failure. It is the contrast that makes the next landmark matter, and open worlds that fill every hectare produce a curious flatness where nothing stands out because everything is a something.

And technical things break at size in ways they never do in one region: streaming, loading, save size, how many things can be alive at once. Finding those with a working small game behind you is very different from finding them with nothing playable at all.

Movement across ground you enjoy. One square kilometre with landmarks, then paths, then detail. The see-go-find loop working reliably inside it. One instance of every content type. Fifteen minutes that teach the player how to want things. Then size.

The thread running through it is that every step produces something you can play and judge. That matters more than the specific order, because an open world is a genre where your intuitions are unusually unreliable. Distances feel different in play than on a plan. Density that reads as sparse from above is often correct at eye height. A landmark you thought was dominant disappears behind a hill you forgot was there.

None of that is knowable in advance and all of it is obvious within thirty seconds of walking around. So the discipline is to keep the gap between building something and walking through it as short as you can stand, and to trust what you feel while walking over what you believed while planning.

That judgement is the part of the work that cannot be delegated to anything. Someone has to walk the world and say whether it is worth walking, and that someone is you.