When making it faster moved the whole world
We meant only to make the world run faster. It was a tidy, unglamorous performance pass — find the slow part, do it less often, free up some frames. What we did not expect was that the same valley would quietly begin feeding more animals. Making the simulation faster, it turned out, moved the whole living world.
This is the story of how that happened, and of the choice it forced on us: when a tweak meant to save time changes the balance of life itself, do you reach for the dial and put things back — or do you trust the world to know better than you do?
The valley got heavy
Once The Long Watch had a whole living-creature loop running — animals being born, grazing, growing hungry, hunters moving among them — the world became far heavier to run than the quiet, empty meadow ever was. A world that breathes costs more than a world that only sits there. That is fair enough; we wanted it to breathe.
So we measured before we cut, the way a hotspot is almost never where you’re sure it is — and the slow part wasn’t the animals at all. It was a single piece of bookkeeping the world redid dozens of times a second: walking the entire field of plants, on the order of five thousand of them, just to refresh how much food had grown back. That whole-field refresh cost exactly the same on every pass, no matter how few plants were alive or how few animals were there to eat them. Grass does not grow back dozens of times a second. We were just asking it to, out of habit.
Two clocks, and the right one
The world keeps two heartbeats. A fast one drives the animals — the moment-to-moment business of moving, eating, deciding — and a slower one drives the ecology, the unhurried work of plants growing and giving back. The regrowth chore had been riding the fast clock, refreshing the food dozens of times a second. We moved it onto the slow one, where it now runs only a few times a second instead.
The trick that keeps it honest is the same one that lets the grass grow on a slow beat in the first place — slowing the work down without changing the answer: each refresh is handed the exact span of time since the last one and grows that much food back, so the total returned over an hour is the same as before. Nothing was made stingier. The grass grows back at exactly the same long-run rate — we simply stopped checking on it so anxiously.
In the long, settled stretch of a game — the part a slow game like this spends most of its life in — that made the simulation roughly one and a half times faster per step, with an even larger gain in the sparse, early days. (The headroom that made any of this possible came earlier, when we decided the world needn’t be simulated in full detail everywhere at once — simulating what you’re looking at and abstracting the rest.)
The same land, more life
And then the surprise. The more interesting consequence of the change was not technical at all. Moving the chore onto the slower beat quietly lifted the valley’s settled population by about fifteen percent. We hadn’t touched a single number about food or fertility or appetite. The same land was now feeding noticeably more animals.
It makes a kind of sense once you sit with it. Food now returns in larger, less frequent pulses rather than a constant thin trickle, and that change in rhythm ripples straight up the food chain — into when there is something to graze, into which animals make it, into how many the valley can carry. Change the cadence of the grass, and you have, without meaning to, changed the shape of every life that depends on it.
We set out to save a few frames, and instead changed how much life a valley could hold. The food chain doesn’t care why you touched it.
The temptation to put it back
Our first instinct was to undo it — not the speedup, the side effect. Dial the regrowth rate down a touch, restore the old population, keep the lovely new speed and the old familiar balance both. It felt like the responsible thing.
So we tried. We swept the regrowth rate across a range of values and watched where each one settled, and the old number simply would not come back: the population floored at the new, higher level no matter how we set the rate. Stranger still, pushing the rate higher settled the valley lower, not fuller — the same backwards lever turned up again later when we balanced the whole ecosystem by running a hundred worlds. It told us plainly that the higher population was not a knob we’d nudged. It was intrinsic to how the bursts of new growth and the grazing now interacted. No single dial was going to undo it, because no single dial had done it.
Read the plateau, not the growth
The sweep taught us one more thing worth keeping. Early on, while a world is still filling up, the population numbers tell you almost nothing about where it will end. Two regrowth rates that looked all but identical in mid-growth settled, in the end, at meaningfully different levels. The only honest reading is the plateau — the settled count a world arrives at after the plants have had their first great collapse and the dust has cleared — never the hopeful figure on the way up.
So we stopped reading the climb and started reading the rest. And what the rest told us was that the new, higher valley was a perfectly healthy one. The world stays limited by crowding rather than by starvation — animals run out of room far more than they run out of food. The plants still collapse before winter as they should, the founders still survive to the spring thaw, and a second generation is still born. Every test of a living world the old balance passed, the new one passed too.
A number nobody chose
That left the real decision, and it came down to a single question we had to ask ourselves honestly: was the old population a number we had ever deliberately chosen? It wasn’t. It was simply where the world happened to settle before, under the old rhythm. It had never been a target. It was just a value we’d grown used to seeing.
And once we said that plainly, there was nothing to restore. You cannot put back a goal you never set, and forcing the regrowth rate into knots to reproduce a figure the world could no longer reach would only have chased a ghost. So we left the dial exactly where it was. (The proper way to make winter dangerous, we’d separately decided, was to let the cold itself raise the risk of dying — not to starve the valley through its food, which collapses long before the season turns anyway.)
The truest balance is the one the world arrives at on its own. Our job was to let it, not to back-fit a number nobody had ever meant.
We accepted the new equilibrium as the truer answer — the one the valley reaches when you let its own rhythms play out — and the long question of how a world full of creatures finally learns to hold itself together is its own story. What this one is about is humility. We came to make the simulation faster, and the world handed back something we hadn’t asked for: a little more life, on the same ground. The right thing was not to argue with it — only to watch where it settled, see that it was well, and let it stand.
