One world with every climate in it: cold at one end, warm at the other
For most of The Long Watch’s life you chose a world’s climate the way you chose its size: boreal, temperate, arid, or tropical, fixed at the moment the world was made. That one choice quietly decided everything the world could ever hold. In the middle of July it went away. A world now runs cold at one end and warm at the other, and every kind of country lives somewhere inside it.
There are eight kinds of country in the game. A world used to contain about two of them, and roughly a quarter of the plants and animals we had written. The roster was never the thin part. It was cut four ways, and each world only ever met its own quarter.
The world you chose was the world you got
Every patch of ground already works out what kind of country it is by reading its own warmth and dampness and taking the closest match, which is how a meadow comes to be sure it is a meadow. The per-world climate sat on top of that as two blunt instruments. It shifted every reading in the world by the same fixed amount, so an arid world was warm and dry everywhere at once. And before any matching began, it struck most of the eight countries off the list of what a place was allowed to become.
The result was a world that could only ever be one thing. A boreal world was cold pine and steppe from edge to edge, and the desert species, the wetland species and the scrub species we had spent months writing were all somewhere else, in a world you would have to start over to visit.
Making one number into a map
Diagnosing that took a day and left us slightly embarrassed. Climate was doing exactly two things, and both could be loosened rather than replaced. Take the fixed shift and let it vary across the map; delete the filter. Everything after that was consequence.
Temperature now slides along one axis of the world, cold at one end and warm at the other. Moisture slides along the other, a wet flank and a dry one — the reading the animals that need damp ground depend on. Both are nudged by the same seed-derived noise that shapes the hills, so the boundaries between regions wander the way a treeline wanders instead of running as ruled lines. The land chooses from all eight countries now: meadow, temperate forest, boreal forest, tropical forest, tropical wetland, scrubland, cold steppe, desert.
The first proof was a walk to the four corners of a world and a look at what stood in each. Four corners, four different countries. Then we flattened the gradient back to a single value and watched the coverage check go red, because a check that has never failed has never been tested.
The code a world grows from lost its climate part along the way, which broke every save ever written with it. We declined those files at the door instead of converting them, and said why in plain words.
Three countries where nothing could live
The same day turned up something worse. Desert, cold steppe and scrubland carry poor soil by design; they are meant to be hard ground. The plant side of the simulation squashed that soil reading into a sane positive range before using it. The creature side did not. It multiplied the raw negative value straight into how many grazing animals a patch of ground can hold, then clamped the result at zero.
So carrying capacity in those three countries was not low. It was exactly zero, at every point, at every setting, in every world ever grown. Plants grew there. Animals mathematically could not. They are the same three countries we had already given something green to grow; the emptiness this time was on the animal side.
We built a survey before we built the fix: something that walks all forty species in the roster against a real world and reports, one by one, whether each can make a living in the places it belongs. Before trusting a word of it, we made it fail. It named eleven grazing species with nowhere at all to live, then flipped all eleven to viable when we lifted the soil reading by hand. That settled where the sterility came from.
You cannot tune your way out of a multiplication by zero.
Which retires an earlier stretch of work in a way that is almost a relief. We spent a long while dropping new grazers into cold and dry worlds, watching them collapse within days, and treating it as a balance we kept failing to find. There was no balance to find.
The fix is a floor instead of a clamp. Poor ground now yields a low number rather than zero, so hard country holds a thin population instead of none.
Nobody’s hand on the scale
The safety net we could have built was ruled out before any of this was written, and it stays ruled out. Nothing in the game watches the populations and props up a species that is declining. There are no per-species target numbers anywhere in the world.
Balance has to fall out of the shape of the chain instead. Soil and climate decide what grows; what grows decides how many grazers a place can carry; the grazers decide how much room there is above them, and a big hunter needs a great deal of room. A species that cannot make a living in a world thins, retreats to the regions that suit it, and may go locally extinct there. That is the system working.
The reason has nothing to do with purity. A hidden hand keeping the world level would make every loss unearned, and loss is the one thing this game refuses to fake.
What only shows up once there are eight
Two faults had been sitting in the game for a long time, invisible because no world had ever contained the countries that would reveal them.
The first: the ground colouring was still being handed the old two-country list. Six of the eight biomes had no colour to draw and came out black — nobody had ever noticed, because until that week nobody had ever stood in one.
The second: every Large world had been running its climate gradient at twice its intended reach, because the gradient was measured against the wrong world width. The cold end and the warm end ran off the edges of their own map. That one surfaced out of the same thread that recently taught a Large world to actually be large.
Neither was caught by our reproducibility checks, and neither could have been. Those prove a world rebuilds identically every time; they say nothing about whether it rebuilds correctly — the gap we went after later by breaking our own checks on purpose.
A small world of many lands
The last piece is the one you will actually read. Under the slowly turning world on the screen where you choose a world, there is a line or two of soft text about the place. It used to be one of four blurbs, one per climate. It now describes one world: “A small world of many lands, from cold pine ridges to sun-warmed sands.” Beneath it sits a line that changes with the season the world starts in, and in winter that line reads “Winter hushes the cold reaches while the warm lands keep their green.”
That screen had been promising climate zones, plural, since long before anything underneath it could deliver one. The code has finally caught up to its own brochure.
There is a limit worth naming out loud. At the sizes a world comes in today, a climate band reads about seventy to a hundred and forty metres across, which is a short walk. Every country is in there, and the whole roster has somewhere to live, but you cannot yet set out from the cold end in the morning and arrive somewhere tropical. That is a matter of scale rather than climate: a bigger world spreads the same span of cold to warm across a longer journey. Which makes how large a world can grow the next thing to solve.
