The look that didn’t touch the simulation: new bodies over a world that never felt them

← All field notes

For most of its life, The Long Watch drew its animals as plain boxes — stand-in shapes while the ecology underneath grew up. This week about forty of them finally got real, sculpted bodies, and the part worth telling is the part that never moved: not one byte of the running world, or of any world you’d already saved, changed. The creatures look like themselves now, and live exactly as they did.

The animals were boxes

Underneath, the world had been alive for a long time. Foxes hunted, herds grazed and bred and aged, the dead fell and fed the ground. But when you looked at any of it, an animal was a box: a rough block standing in for a creature while everything that made it a creature got built out around it. About forty species wore that same box.

Now they have bodies — each one grown from its own species and its own genes. Each animal went from a crude twelve-triangle block to roughly two thousand triangles of real shape — a muzzle, a back, four legs, the particular silhouette of that animal and no other. A deer reads as a deer, a fox as a fox; the birds of prey at the top of the food chain finally look like birds. None of them behaves any differently. They only, at last, look like what they always were.

An aerial golden-hour view of a voxel meadow dotted with distinct animals — a deer, a fox, grazers, and small birds — each a recognizable sculpted shape.Concept art · pre‑alpha
The same herds that had grazed for months, finally shaped like themselves.

A costume, never the creature

The way to do this without disturbing anything is to treat the new body as a costume. The animal underneath stays the same handful of facts it always was — how heavy it is, what it eats, where it sits in the food chain — and the model does nothing but read those facts and decide what to draw. An animal being nothing more than its body and its place in the chain is a principle we built on early, and the sculpted shape is one more thing it pays for. The body invents no new fact for the world to track, and hands a save nothing new to remember.

That is why it was safe to ship in an afternoon. The change lives entirely on the side that shows the world, never the side that runs it, so every world you saved before keeps loading and playing precisely as it did, only better to look at. We migrated no saves; there was nothing to migrate, because the file never learned the bodies exist. It is the same frozen-format discipline that once let us build a whole week of features without the save changing once.

The one real cost

A sculpted body costs more to draw than a box; that much was never in doubt. The question was how much, and where it would bite. It bit in the busiest scene we had, a growing population with 385 animals on screen at once, where drawing bodies instead of boxes nudged the frame rate a hair below the smoothness we had always held it to.

The reflex, when a scene gets heavy, is to cap it: fix a ceiling on how many bodies you will draw at once and protect the frame rate by simply refusing to render the rest of the crowd. We did not reach for that. A cap on what is on screen is the kind of decision that looks prudent and quietly makes the world feel smaller, and it was not one we were willing to make on a hunch.

Measure, don’t presume the cap

So we measured before deciding anything, the same instinct an earlier pass followed when it went looking for a slowdown instead of guessing at it. We built a deliberate stress scene and leaned on it. Five hundred creatures on screen held a smooth sixty frames a second, comfortably more than any real world would ever put there. Then we shoved it far past reason: four thousand-odd creatures, roughly eight times the population a world is ever meant to hold. There it collapsed to about eleven frames a second, a slideshow.

An aerial golden-hour view of a voxel valley densely populated with many animals — grazing herds, scattered hunters, and birds in the air — all on screen at once.Concept art · pre‑alpha
A whole world’s worth of creatures at once, and no cap to thin them.

That told us what we needed. The slowdown wasn’t a wall waiting at some particular number of bodies; it was the entire scene growing richer, cost climbing smoothly with everything drawn in it. There was no single limit worth imposing, no clean line where rendering one more animal was the mistake. So the game ships with no cap at all on how many bodies can be on screen.

Measure, don’t presume the cap.

What we accepted instead was smaller and plainer: in that one busiest scene, the frame rate now settles a frame or two under the floor it used to clear. We watched it run, judged the richer world worth the sliver it cost, and lowered the floor to meet it rather than shrink the world to fit the floor. A slightly softer worst case, in exchange for animals that look like animals everywhere else.


The day after, we did the same for the plants: every species carved into its own voxel form, drawn the same read-only way, with nothing about how anything grows or dies touched at all. It is why leafy trees can now go bare through the winter while the pines keep their green, a change in how the world is seen and none of it a change in how the world lives. The forest had already learned to draw itself this way; now the animals grazing under it have too.

The animals didn’t change. We only stopped drawing them as boxes.

Keep reading

Concept art · pre‑alpha