Shipping an ecosystem that’s not quite right

← All field notes

The day a living world finally learned to hold itself together, it held in a shape we hadn’t drawn. Four kinds of creature settling and staying settled — but the balance leaned the wrong way from the one we’d sketched in advance. The interesting decision that day wasn’t which numbers came out. It was what to do about the gap between the world we’d pictured and the world we got.

The ecosystem itself — how it settled, what it’s made of, why crowding finally held it — is its own story, and the way we found the numbers behind it is another. This note is about something narrower and, to us, more reusable: the discipline of shipping a result that isn’t quite the shape you intended, and being exact about how.

The shape you sketch and the shape you get

Before any of the tuning, we’d drawn a rough pyramid in our heads — a fuller top of the food chain, fewer grazers crowding the base. That picture was honest aspiration. It was not a measurement. When the world finally settled on its own, it settled prey‑heavier and shorter on top hunters than the pyramid we’d hoped for: a wide field below, a top tier thinner than we’d wanted, the rarest creature rarer still.

The temptation in that moment is small and quiet and almost reasonable. You have a number that doesn’t match the picture. You also have knobs. One more tuning round, you tell yourself, and the numbers will move toward the sketch — and the shipped world will look more like the one you promised yourself it would be. That instinct is the thing this post is about refusing.

Two honest paths

There were two roads out, and both were honest. The first was to keep grinding: another round of tuning aimed squarely at the original sketch, pushing the top tier up and the base down until the populations matched the shape we’d drawn. The second was to accept the world as it had actually settled, write down precisely how it differed from the intent, and offer the further tightening as an optional follow‑up that nobody was obliged to take.

What made the second path legitimate — rather than a polite name for giving up — was that the bar for shipping had already changed, and changed deliberately. The day before, we’d let go of the aspiration to hit a precise population shape and replaced it with a looser, sturdier test: is the world bounded, does every kind of creature persist, does it settle smoothly instead of crashing or oscillating, do the predators sit in the right order above the prey, and does the pressure that keeps it bounded reach the off‑screen majority as well as the creatures near the camera? That last point — why a balance has to reach where you aren’t looking — is its own note. The point here is that once the bar is “clears what actually matters,” and not “matches the picture I drew first,” the as‑built world is allowed to ship.

It cleared that bar. Bounded, persisting, smooth, correctly ordered, and biting everywhere. So we took the second road.

Naming the gap, not burying it

Choosing to ship the not‑quite‑right thing only stays honest if the “not quite” is said out loud. The failure mode we were guarding against is subtle: it isn’t lying about the numbers, it’s massaging them — nudging a setting until the run photographs prettier than it really is, then presenting the prettier run as the truth. A number that looks calm can hide a world that isn’t the one you meant to build.

So the ways the shipped world falls short of the sketch weren’t tucked behind a reassuring figure. They were named in plain terms: the mix came out prey‑heavier and leaner at the top than intended; the creature at the very peak is scarcer than we’d hoped, though it persists. We examined that lean top tier closely and kept it on purpose — not because we couldn’t see it, but because we could, and judged the world that held it good enough to live in.

That distinction is the whole ethic of it. A deviation you looked hard at and chose to keep is a different object from a flaw you’re quietly hoping no one notices — and the reader deserves to be told which one they’re looking at. We would rather hand you a world with a named seam than a smoother one with a hidden one.

The follow‑up nobody has to take

The other half of shipping honestly is being precise about the remaining work — and about whether anyone is obliged to do it. We’d explored the first road far enough to cost it: closing the gap toward the original pyramid was a small thing, on the order of one more confirmation pass of about an hour. Not a rewrite. A nudge.

So we wrote it down as exactly that: a costed, optional follow‑up. Accept the world as it is, or ask for one more pass. The default, if nobody asked, was to keep the world as shipped — silence meant accept, not “quietly go do it anyway.” That matters more than it sounds. Leaving the tightening genuinely optional, with no‑answer defaulting to keep, is what stops “we could make this prettier” from sliding back into “we should,” one reasonable‑sounding round at a time, until you’ve spent a week chasing a sketch the world was never going to draw.

And the reason the gap was small enough to leave optional is itself worth saying: the four populations are tightly coupled, so chasing the sketch isn’t free. The kind of trap where reaching for fewer prey and more apex pulls against itself belongs to the tuning story. Here it’s enough that forcing the picture was costed, fragile, and elective — and we declined to force it.

---

An old call, applied to a result that shipped live

If this sounds familiar, it should. We’d made nearly the same call once before, from the other direction: a balance we couldn’t yet find led us to ship a finished mechanism switched off, with the gap named, rather than thread a needle to fake stability for a demo — that story is here. The shape of the judgment is identical. The difference is the verdict. There, the honest state was “the mechanism works, the balance doesn’t yet” — so the honest ship was off‑by‑default. Here, the honest state was “the world holds, just not in the shape we drew” — so the honest ship was live, with the deviation written down beside it.

Same ethic, opposite outcome, and that’s the point. The rule was never “ship” or “defer.” The rule is that the decision turns on the real bar plus an honest account of the gap — and whichever way that points, you say so. A world we switched off and a world we shipped live both passed through the same question: does it clear what matters, and have we told the truth about where it falls short?

Why we’d rather ship the honest seam

It would have been easy to spend the extra hour, land closer to the pyramid, and never mention that the first settled shape leaned the other way. The result would have read as cleaner. It would also have quietly taught us the wrong lesson — that the job is to make the world match the picture, when the job is to make the world hold and then be candid about what holding actually looks like.

A game about tending a world rather than commanding one earns nothing by pretending its world arrived exactly as imagined. The current numbers are a stepping stone, and we’ve said so; the fixed ceilings underneath them are a deliberate placeholder, and the better long‑term answer — ceilings that emerge from the world itself rather than from a number we typed — is written down as where this is going. None of that is a confession. It’s the same discipline that lets the forest be alive: trust the result you actually got, and don’t dress it up as the one you meant to get.