How much bigger a world can grow: the lever we built and switched off

← All field notes

The Long Watch has always let you choose how big a world to tend — Small, Medium, or Large, picked the moment you create one. Until recently, though, choosing Large changed how the land looked without changing how much of it there was. This is the story of building the lever that could finally make a big world big — and shipping it switched off, which is how we found the wall that caps how large any world can grow.

A size that changed the look, not the land

World size is one of the game’s load-bearing choices. It isn’t a quantity dropped into a menu; it’s meant to set a whole envelope of trade-offs — enough room for creatures to migrate, for biomes to feel regionally distinct, for a local ecological collapse to stay local rather than end the world, and for there always to be a frontier left to tend. The size you pick even rides along in a world’s shareable code, so two people who enter the same one settle the same world, at the same scale.

So it was a little uncomfortable to admit that, for all that weight, size barely touched the ground. A bigger world was generated with taller, broader-featured terrain — the hills at Large stand noticeably higher and wider than at Medium — so a large world looked grander. But the actual populated footprint, the area where creatures and plants get scattered, was a fixed span a couple of hundred metres across — about a twenty-fifth of a square kilometre — hard-coded and copied into several separate places, identical for every size. There was no single spot in the game that answered the question ‘how big is this world?’, and no lever that could make Large physically larger.

The gap between that and the intent is almost comic. The design targets a Small world at roughly five square kilometres; a Medium one — the recommended default — at fifteen to twenty, about half the size of Skyrim’s map; and a Large one at forty to fifty: genuine wilderness scale, with biomes a player might not visit for hours. Against today’s twenty-fifth of a square kilometre, that is a hundred- to thousand-fold more ground.

An aerial golden-hour view of a vast voxel wilderness of rounded hills, a winding river, and forests receding into haze toward a distant horizon.Concept art · pre‑alpha
Wilderness scale — the size a Large world reaches for, where whole regions lie hours apart.

Building the lever, and leaving it at zero

The recent work built the missing lever. Three things landed together: a single source of truth for how far a world extends — one value every placement and coverage calculation now reads, in place of the same number copied into several spots; placement that scales from it, so a bigger world seeds its creatures and plants across a bigger area; and the far-off, unseen layer of the simulation stretching its reach to match, so a large world’s distant life still fills it rather than thinning out past the old fixed box.

And then we shipped the whole thing turned off. Every size multiplier ships at one — neutral — so a Small, Medium, or Large world all reproduce today’s exact footprint, unchanged. The mechanism is installed and moves nothing yet. The real per-size numbers, the ones that would actually make Large large, would need to be somewhere between eleven-fold (to reach five square kilometres) and thirty-five-fold (to reach fifty), and we deliberately left every one of them for a later pass, to be dialed in slowly by feel.

We land the mechanism, not the tuning.

That is the whole discipline of it. Building a scaling mechanism and tuning a scaling mechanism are two different jobs, and doing them at once means you can never tell a bug in the machinery from a number that simply wants adjusting. So we built the machinery first, proved it inert — every existing and every new world identical to before — and left the door to bigger worlds open without yet walking through it. It also let us be honest about something we’d otherwise have hit the hard way: an older planning note still on the books says no world should exceed a few square kilometres, which flatly contradicts the forty-to-fifty target. That contradiction has to be reconciled before anyone turns the dial up — better to see it now, with the lever safely at zero.

The wall baked into every save

Building toward bigger worlds is also how we found the first thing that says how big they can ever get. The soil in The Long Watch keeps a memory — a record, laid over the world as a grid, of where the land has been enriched and made fertile. That grid is a fixed size, a little over two hundred and fifty metres across, and here is the catch: it is written permanently into a world the first moment that world is saved, and it cannot be resized afterward. An existing save’s fertility map is the shape it was born with, for good.

So even with the new lever installed, a world that already exists can only grow about twenty-eight percent larger — a bit over a quarter again its size — before its placements start spilling past the edge of that grid. Past the edge, the soil has no memory to enrich, and fertility quietly stops working at the world’s rim. Not a crash; just a world that, out at its new frontier, has forgotten how to grow things. For a game whose whole premise is a world you keep across many evenings, silently breaking old saves is the one thing we won’t do — so that grid is a hard ceiling until we redesign it.

It isn’t the only wall, only the tallest. Two more stand between here and a genuinely large world: turning distant, unseen regions back into real creatures as you approach gets expensive at scale, and the records the game keeps for far-off populations are only cleared away once a lineage goes fully extinct, so a big world would slowly bloat its own save. Both are solvable; both have to be solved before the multiplier is worth turning up. The lever is in; the road past it is mapped, not yet paved.


Why the big machinery is a large-world thing

There is a satisfying way the pieces confirm one another. The far-off, folded-down layer of the simulation — the one that lets a world be bigger than any machine could ever render at once — earns its keep only when there is a genuine off-screen to fold. We learned that bluntly: switching that aggressive abstraction on at Medium size removed something like ninety-three percent of a world’s founding population, because at Medium the whole place already fits inside the closer, fully-simulated tiers — there is no deep distance to abstract, so the abstraction just deletes life that was meant to be right there in front of you. We backed it out and re-aimed it to stay dormant until Large, exactly where the design always said it belonged.

That failure is really a proof: the machinery for a huge world only makes sense once the world is huge. Which is the same thing the extent lever is saying from the other side — the reach of the far world scales with the world’s size, not with where the camera happens to point, so the abstraction and the extent grow up together. Turn the world up, and the machine that carries its unseen distance turns up with it. Leave the world small, and both stay quietly out of the way.

So where does that leave ‘how much bigger can a world grow?’ Today, honestly: not at all — on purpose. Every world, at every size, is still the same twenty-fifth of a square kilometre it always was, because the lever ships at zero. But the machinery to grow one is now real and proven safe, the far-world abstraction a big world needs is aimed at exactly the size that needs it, and the first hard ceiling — soil memory locked into every save — is no longer a surprise waiting in the tuning pass. The wilderness the design keeps promising isn’t here yet. But for the first time, the thing that would build it is.

Keep reading

Concept art · pre‑alpha