The engine has stopped saying no.

Scarcity used to do the designing

Put a wolf in the forest. The engine gives you the fur, the gait, the eyes that follow the player. It asks nothing in return — not even the obvious question: what does it eat?

The answer would cost something. Deer, which means undergrowth to graze, which means the shepherds two valleys over post a bounty within the year. The engine asks for none of it. The wolf is just a costume, and the forest does not know it is there.

It was not always this generous. A world used to offer a handful of actions — shoot, drive, pick up — and those few actions were the game.

That poverty was a gift. It made worlds easy to read and possible to master. Tetris is four rotations. Chess is six pieces. Grand Theft Auto was a gun, a car, and a cardboard box — and every story it told, it told with those. For forty years, the engine set the limits for the designer, quietly, for free.

That is ending, and it should. Simulation is going deep: engines that model a world's actual laws — its water, its weather, its economies, its appetites — rather than scripting their appearances. Every layer of depth mints new actions: the gun that could only be fired can now be thrown, dropped, tripped over. It is the most exciting thing to happen to worldbuilding in a generation. But the scaffolding goes with it.

If today's games look unchanged, that is production lag, not evidence. A big world takes years to build, and the stories are written only after the world is done — everything on the shelves was begun the old way. Those years are becoming weeks.

When anything is possible, the scarce thing is no longer capability. It is constraint.

Not refusal. A bill

Now say the most expensive sentence in worldbuilding: and also, there is a dragon.

A dragon is never just a dragon. It is fire, and fire wants fuel. It is flight, and flight rewrites the price of every wall in the kingdom. It is an appetite the size of a county.

A tool that says yes makes the world thinner. A tool that says no loses the designer. Both are wrong. There is a third answer.

Five questions. None of them invented — each one follows from one rule: nothing here is free. Answer them, and the dragon has somewhere to stand. Refuse, and it is a poster on a wall.

Everything came from somewhere

If iron exists, someone mined it. If someone mined it, they ate. If they ate, someone farmed.

If someone farmed, there is water. And land. And a reason they stayed rather than left. Someone protects them. That protection is paid for. Someone resents the price.

One rule, and a whole society falls out of it — including, if you follow the chain far enough, the grievance that starts your story.

This is the Tolkien mechanic, and it is usually misread. Middle-earth does not hold together because it is detailed. It holds together because a few deep commitments came first, and the rest follows. The differences from our world make sense as you learn more, because none of it was made up on the spot.

The commitment comes first, and it is small. Everything else follows.

Every action has a cost, and a consequence

The same rule binds the player. A choice means something only when its options cost different things — up front, or in what follows. The best games already work like this: a few routes through the locked door, each priced differently — picking the lock costs time, breaking it down makes noise, the key costs whatever its owner wants. Choosing is expressive because the costs differ.

Remove the prices and choice stops working. A hundred ways to do anything, none of them priced, is not a hundred choices — every route is the same route, and nothing the player picks costs them anything. The result is everything possible and nothing at stake: a toybox, not a game.

So the rule is not fewer actions. It is: no unpriced actions, and no two actions at the same price. An unpriced action — one the world charges nothing for — gives the player more to do without making any of it matter. Two actions at the same price are duplicates: merge them. Hold both rules and the list of actions stays short on its own.

Deep simulation supplies the consequences. Pricing them is what makes it a game.

Worlds become finishable

There is a plainer way to say all of this. When you can add anything, you are never done — there is always more to add. Scope creep is what happens when adding things is free.

A world built on commitments has something no sandbox has ever had: a definition of done. A world is finished when its bills are paid and nothing is missing.

The practical test is QA. Most testing on a big game is a hunt for holes nobody meant to leave: one mission quietly breaking another, an economy that fails when a player leans on it. When stories are built inside stated rules, those holes cannot be written in the first place. What remains to test is the game, not the world.

The designer is the author

This is not an argument for a machine that invents your world.

A world assembled from a machine's preferences is exactly the thin world described above. Nothing contradicts anything, and nothing means anything either.

The commitments are yours: what magic costs, why the empire rots, what the sea did to these people.

There is a name for what chooses between two worlds that both hold: taste. Whether a world adds up can be checked. Taste cannot. Among all the worlds that add up, someone still has to choose the one worth building. That choice cannot be delegated.

The commitments also repay their cost. Decide what the wolf eats, and the deer, the forest, and the shepherds are all settled by it — decided once, felt everywhere. A few decisions do most of the work.

Every simulation that goes deep enough will meet this problem: not as a quirk of one engine, but as the next layer of the craft. The machinery it calls for — the thing that reads your world back to you, finds the place where it does not add up, and hands you the bill — does not exist yet.