Founded too thin to breed
A new Long Watch world seeds dozens of species onto the map the moment it finishes growing. A dedicated count found that on every world checked, somewhere between a third and nearly half of those species had already lost their only chance to breed — not to bad luck later in their lives, but to where they’d been set down at the very start.
Every fix reached for first missed it, because each one aimed at behavior instead of geography.
Turning a dial that couldn’t reach the problem
For a long stretch, keeping a world’s species alive meant adjusting how eager or how far-reaching its creatures were about finding a mate: how often two grown animals would try, how wide a stretch of ground they’d search before giving up. A batch of side-by-side runs, each identical except for one of those settings, showed the dials doing exactly what they were built to do, and almost nothing for the thing that mattered. Individual head-counts moved. Whether a species survived at all barely did. A species that was already thriving got a little more thriving. One that was already gone stayed gone.
Where the map put them
So the question turned from behavior to placement: not how hard a creature looks for company, but whether there was ever anyone to find. Each species in a freshly built world is scattered across the map on its own, independently of every other species, with roughly ten individuals typically founding a new one. That scattering is random, and nothing about it guarantees any two of those ten end up close enough to notice each other.
A count built specifically to answer that question checked, for every single creature at the exact moment a world came into being, whether another member of its own species stood anywhere within reach. On every world tested, somewhere between a third and nearly half of all species had zero mate-able pairs the instant that world existed. Counted animal by animal rather than species by species, something like seven in ten individual creatures had no partner anywhere close enough to find.
Because creatures in The Long Watch barely wander on their own, that isn’t a problem a season or two of luck can fix. It’s permanent from the first moment the world exists.
A capacity of exactly zero can never be recovered from.
That’s a different failure than a bloodline slowly losing its vigor from breeding only within itself, a longer, quieter danger the world also tracks — this is a species with no bloodline to speak of yet, because it was never given the chance to start one.
Two shapes of missing
The failing species split cleanly into two different pictures. Most were spread too thin: two or more individuals of the same kind, placed on opposite sides of the map with no way to close that distance on their own. A smaller share were worse than spread: only one individual of that species had been placed at all, and no width of search range can conjure a second one out of nothing.
That split mattered, because it pointed the fix at the right target. Nothing about how a creature behaves after landing could touch either problem. Only where creatures land in the first place could.
A closer start, and a floor under every species
Two changes followed, and they work together rather than competing for credit. The first clusters each species’ founding members closer together right after the world places them, without changing how many of each species exist or roughly where they end up — the placement gets tightened, not redrawn. The second sets a floor: every species now gets at least three founding individuals, enough that losing one still leaves a breeding pair behind. It’s the same minimum distance the game already uses when you place a founding pair yourself, so a world seeding itself and a player seeding a species now follow one rule instead of two.
An earlier version of the clustering idea didn’t survive contact with the math. Scattering a handful of small clumps around random points on the map sounded like the same fix, but for the species already closest to the edge, it would have raised the odds of landing nowhere on the map at all: trading a species that’s present but unreachable for one that’s gone outright. That’s a worse outcome for a measure that only cares whether a species survives, so the idea was set aside before a line of it was built.
What changed when it shipped
Both pieces were built and proven behind a switch left off by default, so they could be checked against everything already shipped before either was allowed to touch a live world. Once turned on, the same count that had found the gap was run again. Every test world came back with zero species sealed out of breeding at the start, down from as many as twenty-six species on a single world before the fix. Species counts and total populations both rose on every world tested, and the extra population didn’t pile up in the predators at the top of the chain — a worry worth checking, given how much closer together the prey now stood.
The idea that didn’t make the cut
One more idea followed close behind: instead of moving founders together, let them wander farther on their own once they’d landed, so a scattered pair might eventually find each other by walking. Tried and measured the same way, it cost more species than it saved in every run — the wandering undid the very closeness the placement fix had just built, so species that had finally been given a chance to meet drifted back out of each other’s reach. It was set aside rather than shipped.
None of this makes the population count the measure of success — that was the mistake the rate dials made in the first place. What changed is narrower and more basic: every species a world places now has at least the chance to become a species that keeps itself going, instead of a headcount that was doomed before the world had finished loading.
