The Discovery at 1,000 Meters
In 2024, the Jiaolong crewed deep-submergence vehicle descended to a seamount slope in the Northwest Pacific. At 1,000 meters — no sunlight, pressure 100 times that at the surface, temperature near 4°C — scientists spotted a strange "jungle": translucent, glass-like vases standing on the rock. These were glass sponges (class Hexactinellida), among Earth's oldest multicellular animals, whose skeletons are made of hydrated silica (SiO₂·nH₂O) — the same chemistry as window glass. Their needle-like spicules, some over ten centimeters long, protrude like glass thorns sharp enough to deter any predator.
But what caught the scientists' attention was inside. When specimens were opened, an 18-centimeter worm lay coiled in the central cavity (atrium) — equipped with a pair of spring-knife-like jaws and bristled flanks, like a blade sheathed in a scabbard.
The worm was formally described in March 2025 by Zhou Yadong's team at the Institute of Oceanology, Chinese Academy of Sciences, in *ZooKeys*: Eunice siphoninsidiator — "the ambusher in the tube."
A Distant Cousin of the Bobbit Worm
To appreciate why this worm is interesting, meet its famous relative, the Bobbit worm (*Eunice aphroditois*): up to 3 meters long, it buries itself in tropical shallow-sea sand and strikes with lightning speed, shearing fish in half with scissor-like jaws — divers call it the "chainsaw of the sea floor." Its strategy: build your own burrow, ambush with brute force.
*Eunice siphoninsidiator* took another path. At 18 centimeters — less than 1/15 of its cousin's length — it doesn't dig. Instead, it occupies the glass sponge's central cavity, a space woven from silica spicules like a thorn thicket. It relies not on speed or strength, but on position and camouflage. Hence the name: *siphon* (tube/cavity) + *insidiator* (ambusher).
The Logic of Living in a Thorn Bush
Silica spicules are a "no entry" sign to most animals: sharp, brittle, and hard to remove once embedded. But the worm enters anyway — because inside the thorns is safe. The deep-sea slope hosts plenty of predators (deep-sea fish, octopuses, large crustaceans), and an exposed 18-cm worm is easy prey. The sponge's cavity is a reverse fortress: predators would get a mouthful of glass shards trying to reach in, while the cavity itself is a roomy conduit for the worm to coil, move, and ambush.
The sponge's defense becomes the worm's refuge.
Nor is it one-sided freeloading. Dissection found barnacle and brittle-star fragments in the worm's gut — these are exactly the epibionts that try to attach to the sponge's surface and compete for food. Glass sponges are filter feeders; if their surface gets covered, their filtration efficiency drops. The worm eating these epibionts amounts to cleaning service.
The worm provides security; the sponge pays rent and shelter. The paper calls this a "potentially mutualistic relationship."
Obligate Symbiosis: Every Sponge Had a Worm
The most striking data point: every examined glass sponge specimen contained a worm in its central cavity. Specimens were collected across multiple Jiaolong dives (JL230, JL233, etc.) at the DD, O-Hakucho, and Albo seamounts in the Northwest Pacific. Whichever site or dive, any Farreidae glass sponge sampled harbored *Eunice siphoninsidiator*.
Such a 100% symbiosis rate is rare in marine biology — it implies the relationship has evolved to an obligate stage: the worm cannot live without the sponge, and the sponge is habituated to the worm.
An instructive comparison: *E. siphoninsidiator* is morphologically and molecularly very close to the Atlantic *Eunice norvegica* (COI genetic distance around 13%), suggesting they are sister species that diverged only millions of years ago. But *E. norvegica* lives in ordinary rock crevices, not sponges. Same body plan, different habitat strategy — letting scientists ask precisely: what made one worm choose the sponge? The answer may lie in seamount ecology: seamounts are deep-sea oases with complex currents and abundant food but scarce hard substrate already occupied by corals and sponges. The glass sponge's cavity is a ready-made, defended, vacant room.
An Original Concept: Defense as Habitat
The author proposes a generalizable concept: "defense as habitat" — one organism's defensive structure can become another's habitat.
The key is functional conversion. Silica spicules are defense for the sponge but habitat for the worm: shelter, ambush position, camouflage. Function is not inherent to a structure; it is defined by relationships.
Nature offers many examples of "defense as habitat":
- Sea urchin spines shelter small fish, shrimp, and worms
- *Echinothrix* urchin spines host specialized symbiotic crabs
- Coral calcium-carbonate skeletons shelter countless reef-dwelling species
- Pufferfish spines carry small attached crustaceans
- Paper: Zhou, Zhang, Shen & Zhang (2025). "Ambusher in sponge: a new species of Eunice (Annelida, Eunicidae) commensal within deep-sea Farreidae on northwest Pacific seamounts." *ZooKeys* 1230: 25-36. https://zookeys.pensoft.net/article/140329
- WoRMS Top-Ten 2025: https://www.marinespecies.org/worms-top-ten/2025
- NOAA Glass Sponges: https://oceanexplorer.noaa.gov/ocean-fact/glass-sponges
- Schmidt Ocean Institute: Glass Sponges, the Living Ornaments of the Deep Sea (2020)
But *Eunice siphoninsidiator* takes this to the extreme: it lives not just beside the defense, but inside it — and turns the defensive structure into an ambush position, upgrading "defense as habitat" to "defense as ambush post."
AI Parallels: Leveraged Strategies Beat Brute Force
The author draws three design principles for AI systems:
1. "Borrowed ambush" = tool calling. The Bobbit worm digs its own burrow; *E. siphoninsidiator* borrows the sponge's. In AI engineering, this maps to "build infrastructure yourself vs. call external tools." The Euclid-MCP project's core idea: don't train an LLM to do Prolog-style rule reasoning — outsource it to a Prolog engine. LLM provides semantic understanding, Prolog provides precise reasoning; mutual benefit, like worm and sponge.
2. "Security for rent" = small-model + large-model division of labor. The Rebucca video surveillance project works this way: a small model does initial screening (the guard), a large model does authoritative review (the landlord). Together, false-positive rates drop by an order of magnitude. Neither replaces the other — each does what it does best and trades services.
3. "Living in the thorns" = finding niches within constraints. The spicule thicket is an obstacle to most animals but a niche for the worm. The colibrì project — 1300 lines of C code running 744 billion parameters on a 25GB GPU-less laptop — is the engineering version: MoE decouples total parameter count from memory requirements by 15×, turning "not enough memory" from an obstacle into "sparse activation" as a design space. Don't eliminate constraints — turn them into design space.
Closing: The Overlooked Insight
Glass sponges have existed for hundreds of millions of years. So have the worms. Their relationship may have persisted for millions of years — but only in 2024, via the Jiaolong dive, and 2025, via the paper, did we "see" it. Major breakthroughs are often not new discoveries but re-seeing existing things — a theme echoed by the 2025 resolution of the Ten Martini Problem using tools available since 1976.
The sponge ambusher worm reminds us: nature is full of relationships that already exist but go unseen. An 18-centimeter worm doesn't need more speed, size, or strength — just a different position. In today's AI arms race of bigger models and more compute, the worm suggests an alternative: in ecological niches, leveraged strategies may outlast brute force. The Bobbit worm is 3 meters of apex ambush; the sponge ambusher is 18 centimeters in glass thorns — and may have lived there for millions of years.
Strength is not the only path to success. Sometimes, living inside the thorns is.
---
References: