A Sphere at 3,601 Meters
In March 2025, near the South Sandwich Islands in the South Atlantic, the remotely operated vehicle SuBastian was moving slowly between volcanic craters and trenches. At a dive site marked "Trench North," east of Montagu Island, the depth gauge read 3,601 meters.
There, the camera caught a small, nearly spherical object—resembling a golf ball or a spiky chestnut. It was not a rock or coral fragment. It was a living sponge, a new species of the genus *Chondrocladia*, and it was a hunter.
Its spherical surface is covered with microscopic barbs. When small crustaceans or copepods drift close, the barbs hook them like Velcro; the sponge's cells then slowly envelop and digest the prey over several days. It is a creature with no mouth, no stomach, no eyes, no legs, and no brain—yet it hunts.
This is the species nicknamed the "death-ball sponge." In October 2025, the Nekton-led Ocean Census and the Schmidt Ocean Institute confirmed it as one of 30 new species found during the Southern Ocean expedition.
Why Would a Sponge Eat Meat?
Sponges (phylum Porifera) are among Earth's oldest multicellular animals, existing for at least 600 million years. Their basic body plan is simple: a porous sac; water flows in, choanocyte cells filter out food particles, and water exits through an opening at the top. This mechanism is so successful that sponges live everywhere from shallow hydrothermal vents to under polar ice—no brain, muscles, or nervous system required, just water flow and patience.
But the deep sea has a problem: food is scarce. Nutrients on deep-sea plains come mainly from "marine snow"—dead plankton, fecal matter, and debris sinking from the surface. By 3,000 meters, little energy remains. A filter-feeding sponge must process enormous volumes of water for a trickle of food.
*Chondrocladia* took a different path. Roughly 200 million years ago, its ancestors abandoned—or rather modified—traditional filter feeding. Using specialized spicules evolved into barbed structures, they hook passing small animals directly. The sponge does not move; its body surface simply changed from a "filter mesh" into "flypaper."
A 2008 study in *Nature* framed this carnivorous strategy as an evolutionary adaptation to oligotrophic deep-sea environments: where food is extremely scarce, waiting for dinner to arrive is inferior to making dinner bump into you. The energy in a single passing crustacean equals the organic matter from filtering thousands of liters of seawater.
*Chondrocladia* is not alone. Over 30 carnivorous sponge species across the family Cladorhizidae have independently evolved similar strategies. The 2012-discovered "harp sponge" (*Chondrocladia lyra*) uses branching fronds with terminal bulbs to expand its hook coverage; the 2025 "death ball" curls its whole body into a sphere to maximize surface area. Same problem, different solutions—exactly how evolution works.
The Worm Without a Mouth
Even stranger is a creature found alongside it: *Osedax*, the "zombie worm," which lives on whale falls. When a whale dies and sinks, its skeleton becomes a deep-sea oasis that can sustain an ecosystem for decades.
*Osedax* has no mouth, no digestive tract, no eyes. Its "roots" penetrate whale bone, secreting acid to dissolve it, while symbiotic bacteria inside its body break down bone lipids into compounds the worm absorbs. In effect, *Osedax* turns itself into a tree planted in whale bone, drawing nutrients through a fungal-root-like bacterial system—except its soil is a whale dead for decades.
*Osedax* is not a new species, but its presence is a signal: whale falls exist here, and a complete deep-sea food chain is operating—an ecosystem we know almost nothing about.
The Hidden Garden Under the Ice
On January 13, 2025, iceberg A-84, roughly 510 km² (about the size of Chicago), calved from the George VI Ice Shelf on the Antarctic Peninsula, exposing seafloor that had been ice-covered for centuries or longer. The research vessel R/V Falkor (too) changed course, and on January 25 became the first humans to explore the area.
Over 8 days, the ROV SuBastian dove as deep as 1,300 meters and found large sponges, coral gardens, icefish, giant sea spiders, and octopuses—biomass and diversity far beyond expectations.
Deep-sea ecosystems are usually powered by two pathways: marine snow from surface photosynthesis, or chemosynthesis at hydrothermal vents. But this seafloor sits near no vents and was shielded by a ~150-meter-thick ice shelf, blocking direct settling of surface production. What keeps it alive?
The team's hypothesis: deep currents carry dissolved organic matter and nutrients from afar. The mechanism remains unclear. What is more puzzling is the organisms' size. Sponges grow extremely slowly—sometimes under two centimeters a year. Specimens of the size filmed may have lived for decades or even centuries.
This ecosystem is not newly established under the ice. It has been there all along—we simply could not see it.
30 New Species and 1% of the Ocean
Putting these findings together:
- A spherical, barbed carnivorous sponge hunting at 3,601 meters
- A mouthless worm relying on symbiotic bacteria to digest whale lipids
- A flourishing under-ice seafloor garden with giant sponges, corals, octopuses, and sea spiders
- The first footage of a live juvenile colossal squid (*Mesonychoteuthis hamiltoni*), 30 cm long, filmed at 600 meters—the first time the species was ever recorded alive in its natural habitat
- 30 new species from 14 animal phyla, from polychaete worms to crustaceans, sea stars, and gastropods
- And these findings are based on less than 30% of the samples collected
- Ocean Census press release, "Carnivorous 'Death-Ball' Sponge Among 30 New Deep-Sea Species from the Southern Ocean," 2025-10-29
- Schmidt Ocean Institute, "Thriving Antarctic Ecosystems Found in Wake of Recently Detached Iceberg," 2025-03-20
- Schmidt Ocean Institute, "First Confirmed Footage of a Colossal Squid," 2025-04-15
- Vacelet, J. et al. "Carnivorous sponges" (systematics review), 2016
- MBARI, "Harp Sponge" discovery, 2012
- Smithsonian Ocean, "Zombie Worms Crave Bone"
"The Southern Ocean remains severely undersampled," said Dr. Michelle Taylor, Head of Science at Ocean Census. "We have assessed less than 30% of the samples from this expedition and already confirmed 30 new species—which suggests how much biodiversity remains unrecorded."
Humanity may know less about the deep ocean than about the lunar surface. Estimates suggest less than 5%—possibly as low as 1%—of the deep sea has been explored.
Why It Matters
Evolutionary possibility. Life has experimented for ~3.8 billion years, and the deep sea is its largest, oldest, most stable laboratory. Hunting sponges, mouthless worms, and sunless ecosystems expand our understanding of what life can do.
The boundary of our knowledge. Our认知 of Earth is limited by our sampling. The A-84 discovery is stunning not because it violates physics, but because it exposes the limits of our models. Our model said "the deep sea needs surface nutrition"—that ecosystem clearly does not. The model was wrong, not the ecosystem.
Analogies and inspiration. Our understanding of large language models is also based on "sampling": we study frontier models and derive scaling laws, emergent abilities, and alignment challenges. But the design space of models is far vaster than the regions we have explored. Sparse-activation models (like MoE) being rediscovered in 2024 showed that "total parameters" and "activated parameters" are different things—just as deep-sea ecologists realized "surface settling" and "deep-current transport" are different things. A 744-billion-parameter model running on a GPU-less 25GB laptop via MoE routing is structurally isomorphic to the death-ball sponge: solving problems in resource-constrained environments not by doing the same thing harder, but by switching to a different level of solution. And under-ice ecosystems resemble "overlooked insights" in AI—old methods and abandoned architectures that may still work somewhere we have never dived to look.
Descending
The rules you believe in are only the rules of the region you have sampled. Sponges can eat meat. Worms can lack mouths. Ecosystems can live without sunlight.
These discoveries overturned no physical laws. They merely remind us that many of our biological "laws" are regularities that hold only where we have looked. A-84 exposed 510 km² of seafloor; the Southern Ocean covers about 20.3 million km². The remaining 70% of collected samples still await identification in laboratories.
The deep sea is not in a hurry—it has waited hundreds of millions of years. We are the ones in a hurry, because before we understand it, it is being changed by deep-sea mining, climate change, and acidification. We are burning a library we have not yet read.
---
Sources: