🌲 Opathorlokan University · Living Systems · OPA 4.3.15
A meat-eating Hawaiian caterpillar lives inside spider webs and builds its armor out of the corpses of the spider's prey. It selects body parts, measures them, trims them to fit, and refuses anything that isn't a corpse. That last fact is the whole lab. Selection means discrimination. Discrimination means something is doing the deciding. This lab asks what — and refuses to pretend we already know.
SPECIMEN: Hyposmocoma sp. "bone collector" · unnamed · monotypic lineage · Oʻahu, Waiʻanae range · n = 62 over 22 years
01 · What was actually found
In 2025, Daniel Rubinoff and colleagues at the University of Hawaiʻi at Mānoa described a caterpillar unlike anything else in the order Lepidoptera: carnivorous, living inside another animal's web, dressed in the remains of the dead.
Fewer than 0.13% of the ~200,000 known moth and butterfly species eat meat. This one does more than that. It crawls through cobwebs in rotted logs and rock cavities, eating insects the spider has already trapped or killed — even chewing through silk to reach a meal. Only one bone collector lives per web, because a bigger one will eat a smaller one. It is a carnivore, a cannibal, and a squatter, all at once.
But the behavior that names it is the decorating. After feeding, it takes the inedible leftovers — an ant's head, a weevil's head, a fly wing, a beetle abdomen, a shed piece of the spider's own skin — and weaves them onto a portable silk case it carries on its back. Cases have been found bearing parts from at least six different insect families.
The caterpillar is particular. Each new piece is rotated and probed with its mandibles several times, then measured, then chewed down to a size that fits the case. And the tell: in captivity, when denied real body parts, the caterpillars refuse other detritus entirely — bits of bark, generic debris, none of it. They will only use corpses.
That refusal is not decoration. It's a decision. The animal is drawing a line between this counts and this doesn't, and acting on the line. Every other lab in this college has to explain that line. So do we.
02 · The same behavior, read two ways
The bone collector sits exactly on the seam between two other labs in this suite. Before you look at any data, hold both readings honestly. They can't both be the whole story, and the interesting part is that the behavior fits uncomfortably into each.
Kelly's Colony teaches that complex structure needs no planner. Ten thousand ants with no boss build a city because each ant runs three or four dumb rules — drop food near food, never in the brood, dig toward the boundary — and the city falls out of the math. The intelligence lives nowhere. It lives in the interactions.
Read the bone collector the same way and the magic drains out fast. Maybe it's just: (1) eat the soft parts, (2) if a leftover is hard and corpse-shaped, trim it and silk it to the case, (3) if it's not corpse-shaped, ignore it. Three rules. No deciding, no inner life — a chemical signature on real cuticle triggers the attach routine, and bark simply never trips the switch. "Refuses detritus" becomes "detritus lacks the trigger molecule." Case closed.
The Workspace lab describes a narrow internal space where reasoning happens — where a system holds a concept, operates on it, and directs action from it. Not reflex. A place where this is a corpse, that is mere detritus is actually represented, compared, and used.
Read the bone collector this way and the sizing step is the tell. Rotating a part, probing it repeatedly, measuring it against the case, trimming to fit — that's not a trigger firing. That's an object being evaluated against a goal state and modified until it matches. And the caterpillar is solitary. There's no colony to smear the cognition across, no pheromone field doing the thinking. If anything is deciding here, it's one small brain, alone.
Kelly's Colony is the strongest case in the whole college that complex behavior needs no central intelligence. The bone collector is the hardest thing for that argument to swallow — because it's an individual, doing something goal-shaped, with no swarm to hide the cognition inside. If emergence explains the ant city, does it also explain the one caterpillar sizing a weevil head? Maybe. That's not a rhetorical question. That's the experiment on Tab 04.
NULL sets down two boxes. One is labeled RULES. One is labeled REASONS. NULL puts the caterpillar in neither. NULL puts the caterpillar on the lid, straddling both, and waits to see which way it leans. NULL has done this before. NULL is prepared to wait a long time.
03 · What the evidence actually supports
Here's where most "look how smart it is" stories quit — right before the part that could embarrass them. We don't. Two columns: what caterpillar cognition research already establishes, and what nobody has shown for the bone collector specifically.
Caterpillars have the learning organ. Insect larvae — moth caterpillars included — possess mushroom bodies, the associative-memory center of the insect brain, functional from the earliest larval stages onward. It isn't a reflex relay. It's described as a re-coding center that converts raw sensory input into value-based information — strikingly close to the "re-coding / broadcast" role your Workspace lab gives J-space.
Caterpillars demonstrably learn — and remember through metamorphosis. In a classic result, fifth-instar Manduca sexta caterpillars were trained with an odor paired to a shock; they learned to avoid the odor, and the aversion was still present in the adult moth — a definitive demonstration that associative memory survives metamorphosis in Lepidoptera. A caterpillar learns something and the moth still knows it after its brain was partly dissolved and rebuilt.
The clade next door already links case behavior to real cognitive machinery. The Heliconius butterfly — the same animal that runs "The Maintainer" in The Immortality Question (4.3.8) — has enhanced long-term memory and expanded mushroom-body plasticity, with mushroom-body size tied to foraging ecology and cognition. Same family of insect, cognition scaling with how hard the animal's life is. That's a bridge that fell out of the data, not one I bent to fit.
Everything in the green column proves capacity, not this behavior. The learning that's been demonstrated is associative — odor plus shock. That's a real cognitive event, but it's a far lower bar than holding a goal state and modifying an object until it matches. The clade clears the associative bar. Nobody has tested whether the bone collector clears the workspace bar.
And the sharpest experiments simply don't exist for this animal. The closest is a 2013 Mānoa dissertation on the carnivorous Hyposmocoma clade showing prey preference varies among species — but that's variation between species, not individuals getting better at building. With 62 specimens in 22 years, the fine-grained behavioral work is nearly impossible to run. The column isn't empty because the answer is "no." It's empty because the question was never asked.
04 · The one test that would settle it
There's a single experiment that separates a program from a workspace, and it's almost runnable even on an animal this rare. Take a bone collector mid-build. Knock a placed part loose, leaving a gap. Then watch.
A program runs its sequence regardless — it doesn't perceive the gap, so it keeps going. A workspace notices the mismatch between the case and the goal state, and fixes it: re-probes, sizes a replacement, patches the hole. Failure-correction is the tell.
You just watched two hypotheticals, each clearly labeled. Here is what the lab will not do: tell you which one is real. Because nobody has run this experiment on the living animal. The two animations above are equally consistent with everything currently known. The gap between them is exactly the size of the missing data — and that gap is the entire reason this lab exists. If you ever get to watch a real bone collector meet a broken case, you'll know something no one on Earth knows yet.
Not just "does it fix the hole" — that's the headline. The real signal is in the details: does it re-probe the gap before acting? Does it size a new part to the specific hole, or grab any part? Does correction get faster with practice (a learning curve)? Do some individuals do it and others not (variation)? Each of those pushes the needle from "looks intelligent" toward "here is the measurable signal" — which is the only move that ever actually counts.
05 · Put it on the wall
You've walked the discovery, both readings, the real data, and the experiment nobody's run. So — one commitment. What is the bone collector doing when it sizes a corpse and refuses a twig? Pick one. The lab will put it on the wall and show you the bill your position has to pay.
A few simple rules on real cuticle produce the whole thing. Discrimination is just a trigger molecule; sizing is just a fit-until-it-stops loop. No inner space required — same as the ant city.
Holding "corpse vs not," evaluating a part against the case, trimming until it matches — that's an object measured against a goal state. A small solitary brain is reasoning about its armor.
The hardware is real, the capacity is proven, the specific proof is missing. Holding the two readings side by side without collapsing them is the honest state until the data arrives.
Your ledger is yours alone. Nothing leaves this page — no backend, no tracking, no cookies. It lives in browser memory only and is gone the moment you close the tab.
However you committed, run it through the gauges before you leave — the same test The Socratic Mirror teaches:
The Bone Collector · v0.1 · a Living Systems lab on one carnivorous caterpillar and the limit of what we can honestly claim about it.
The animal, its ecology, and every number: 62 specimens over 22 years, 15 km² range, the 6-million-year-old monotypic lineage, carnivory in <0.13% of Lepidoptera, the selection–sizing–refusal behavior, and "decorate or die." All from Rubinoff, San Jose & Doorenweerd (2025). The cognition data — mushroom bodies in larvae, Manduca memory surviving metamorphosis, Heliconius mushroom-body plasticity — is real published work, cited below.
Opathorlokan University, the "emergence vs workspace" framing, the scorecard, the break-the-case experiment as a teaching device, and NULL the Penguin. The failure-correction animation is a hypothetical, clearly labeled — it depicts two possibilities, not a recorded result. No claim is made that the real animal has been tested this way. It has not.
The animal. Rubinoff, D., San Jose, M., & Doorenweerd, C. (2025). Hawaiian caterpillar patrols spiderwebs camouflaged in insect prey's body parts. Science 388(6745), 428–430. — doi.org/10.1126/science.ads4243
Behavioral variation in the clade. Behavioral Ecology and Evolution of Hawaii's Endemic Carnivorous Caterpillars (Hyposmocoma), Ph.D. dissertation, University of Hawaiʻi at Mānoa, 2013.
Memory through metamorphosis. Blackiston, D. et al. (2008). Retention of Memory through Metamorphosis: Can a Moth Remember What It Learned As a Caterpillar? — associative aversion learned by Manduca sexta larvae persists into the adult.
The insect learning center. Reviews of the mushroom body as the associative-memory / re-coding center in insect brains, functional from the earliest larval stages; e.g. Parnas, Manoim & Lin (2024), Learning & Memory. — doi.org/10.1101/lm.053825.123
The Heliconius bridge. Enhanced long-term memory and increased mushroom-body plasticity in Heliconius butterflies (2024), linking mushroom-body size to foraging ecology and cognition.
The science is the hook; the honesty is the point. Nothing here is a claim that the bone collector is conscious, intelligent, or has been tested for failure-correction. It is a map of exactly how far the evidence reaches — and where it stops.