The Repository Deep Time Archives

The 550-Million-Year-Old Glass Sponge Had No Glass

October 13, 2026 · Runtime 9:23 · Watch on YouTube

A storm in a shallow sea, about 550 million years ago, buried a goblet-shaped animal whose wall was boxes inside boxes. In 2024 it was named a glass sponge. Not one grain of glass was found in it — and the case that there never was any rests on how it lies in the rock.

Chapters

  1. 0:00 A storm in a shallow sea
  2. 0:25 Glass sponges and their glass
  3. 0:52 Fossils on pig-pen roofs
  4. 1:18 Buried by storms
  5. 1:39 Dated above and below
  6. 2:04 Specimen NIGP-176531
  7. 2:31 Boxes inside boxes
  8. 2:58 A fractal that does not branch
  9. 3:21 In glass sponges, the grid is glass
  10. 3:40 So where is the glass?
  11. 3:55 Cloudina, in the same rock
  12. 4:13 It behaved like tissue
  13. 4:33 Bulging boxes, loose crosses
  14. 4:55 The phylogenetic test
  15. 5:17 The missing glass is an input
  16. 5:37 Never made, or lost?
  17. 5:56 Two kinds of absence
  18. 6:19 Where's the glass? (2010)
  19. 6:44 The glass came later
  20. 7:03 Sponges by the genes, 2026
  21. 7:30 Spicules leave moulds
  22. 7:55 Three lines, one window
  23. 8:20 Measured, argued, open
  24. 8:46 Shape first, glass later
  25. 9:05 Next: the animal at the end of its trail

Transcript

A storm in a shallow sea 0:00

About five hundred and fifty million years ago, a storm passed over a shallow sea in what is now southern China. It lifted lime mud and dropped it on whatever stood on the sea floor. One of the things it buried was shaped like a goblet, taller than a forearm, with a wall made of boxes inside boxes. In 2024 it was named a glass sponge. Not one grain of glass was found in it.

Glass sponges and their glass 0:25

Glass sponges still live on the deep sea floor today. They build their bodies from spicules: needles and crosses of silica, real glass, grown by the animal itself. This fossil has the pattern those crosses make, and none of the crosses. By the end, you will know what the rock actually recorded, what was inferred from what is missing, and why two studies published this year lean on that inference.

Fossils on pig-pen roofs 0:52

The rock is the Shibantan Member of the Dengying Formation, a dark limestone near the village of Wuhe, in the Yangtze Gorges. Local farmers quarried it for roof tiles. For years, part of the fieldwork was looking up, at the roofs of pig pens, where fossils lay half hidden under modern lichen and moss. Later the team quarried the beds themselves, so that every fossil came with its own layer.

Buried by storms 1:18

That limestone was laid down below the reach of everyday waves, but within the reach of storms. Its graded beds and broken fragments are storm deposits, and storms are probably what buried the things living there, quickly. That is why this rock keeps organisms that had no shell at all, right beside a few that did.

Dated above and below 1:39

The Shibantan itself has never been dated directly. Below it, ash beds at the top of the Doushantuo Formation give about five hundred and fifty-one and five hundred and fifty million years. Above it, an ash in the Baimatuo Member gives five hundred and forty-three, give or take three and a half. The fossils sit between the two, in a single layer about two metres above the base.

Specimen NIGP-176531 2:04

The specimen that carries the name is NIGP-176531, at the Nanjing Institute of Geology and Palaeontology. Its conical body is two hundred and ninety-one millimetres long and a hundred and eight wide, narrowing into a stem. A second specimen is complete: four hundred and forty-seven millimetres from a disc on the sea floor, up a stem, to the rim. Eight are known, and two are whole.

Boxes inside boxes 2:31

What makes it unusual is the wall. It is a lattice of raised rectangles, each about twelve millimetres by nine, outlined by grooves. Finer grooves split every rectangle into four. Finer ones again split each of those into four, and a fourth order is faintly visible. The species is named after Georg Cantor, the mathematician of sets that repeat themselves at every scale.

A fractal that does not branch 2:58

Living things are full of fractals, but almost always as branching tubes: lungs, leaf veins, roots. A pattern that divides boxes into boxes is, in the authors' words, exceedingly rare. They know it from one other place: the glass sponges of the Palaeozoic, like Hydnoceras, from the Devonian rocks of New York.

In glass sponges, the grid is glass 3:21

In those sponges, the grid is made of glass. It is built from spicules shaped like crosses, laid at right angles and bundled at several scales. Where the spicules dissolved after burial, the grooves they left behind look, in the authors' words, virtually identical to the Chinese fossil.

So where is the glass? 3:40

So where is the glass? Nothing mineral survives in Helicolocellus. Even the body wall itself is gone. What remains are casts and moulds: the shape the sediment took around something that was no longer there.

Cloudina, in the same rock 3:55

The case that there never was any glass rests on a comparison inside the same rock. The Shibantan also holds Cloudina, small tubes with mineral walls. Buried by the same storms in the same lime mud, Cloudina keeps a three-dimensional tube.

It behaved like tissue 4:13

Helicolocellus does not. It lies flattened and partly compacted, with sparry calcite filling the space where its central cavity was, exactly like the soft frond Arborea and the soft tube Wutubus that lie beside it. A mineral skeleton should have held its shape. This one behaved like tissue.

Bulging boxes, loose crosses 4:33

There is one more clue. On the holotype, every rectangle bulges outward, as if a flexible wall had been pressed by sediment against a stiffer frame. And near the base of several specimens, the boxes sit out of line, knocked loose, the way the cross-shaped spicules of Cambrian glass sponges scatter when the animal decays.

The phylogenetic test 4:55

Then a test that does not depend on the eye. The team placed the fossil in a Bayesian analysis of seventy-nine animals and two hundred and thirty-five characters, sixty-seven of them scored for Helicolocellus. It landed among the sponges, on the branch that leads to glass sponges, with a Bayes factor of almost sixteen over the alternative.

The missing glass is an input 5:17

But look at what that placement is made of. The tree puts Helicolocellus just outside the living glass sponges, and the authors explain why: every living glass sponge has mineral spicules, and this fossil has none. So the missing glass is not a result of the analysis. It is one of its inputs.

Never made, or lost? 5:37

And the paper does not claim to know which history it records. Its placement, the authors write, is consistent with either one: spicules invented later and separately, or spicules once made and then lost. The fossil shows a grid without glass. It does not show why.

Two kinds of absence 5:56

Two features of living glass sponges are missing. One is the inhalant pores, four to thirty thousandths of a millimetre across. The finest detail this limestone records is about ten times coarser, so that absence says nothing. The missing glass is a different kind of absence: the grid itself is preserved, down to grooves a third of a millimetre wide.

Where's the glass? (2010) 6:19

That matters because of a gap. In 2010, a paper led by Erik Sperling asked, in its title: where's the glass? Biomarkers and molecular clocks put siliceous sponges deep in the Precambrian, yet the oldest spicules appear only near the very end of the Ediacaran, about five hundred and forty-three million years ago. They called it two hundred million years of missing record.

The glass came later 6:44

One explanation was chemistry: the water in Precambrian sediments may have dissolved the glass before it could fossilise. Helicolocellus points the other way. A sponge could have a body, a shape and a lattice, all in organic material, and the glass could come later.

Sponges by the genes, 2026 7:03

In January this year, a team led by Eleonora Rossi tested that with genes. From a hundred and thirty-three protein-coding genes, they date the last common ancestor of living sponges to about six hundred and eight million years ago, in the early Ediacaran. With fossils included, the chance it had no spicules is at least ninety-seven and a half percent. And silica spicules appear to have evolved four separate times.

Spicules leave moulds 7:30

A second study asks the opposite question: could the spicules have been there, and simply lost? It is a preprint from Colorado, not yet peer reviewed. In the Cambrian, spicules turn up again and again, even in intervals hotter than the Ediacaran, when silica should dissolve faster. And about eighty-six percent of them survive not as glass, but as moulds, casts and replacements.

Three lines, one window 7:55

Dissolving the glass does not erase its shape. Put that together with the genes, and the lines meet. The 2026 sponge tree begins in the early Ediacaran, which is where the 2025 sponge-molecule paper from our last episode now starts its own timeline. Fossil, genes and molecules all point to sponges that lived for tens of millions of years without glass.

Measured, argued, open 8:20

So here is what the rock gives, and how far to trust each part. The grid is measured: three orders of rectangles, counted and sized. That the skeleton was organic is argued, from how the fossil lies next to Cloudina. Whether glass was never made or was lost is open. A 2025 review by Joseph Botting and colleagues still calls the Precambrian candidates ambiguous.

Shape first, glass later 8:46

If the reading holds, the Cambrian explosion looks a little different. Some animals that seem to appear suddenly in the early Cambrian may not be new animals so much as old ones newly building with minerals. The sponge would be the clearest case: the shape first, the glass later.

Next: the animal at the end of its trail 9:05

Next time, another animal from the same limestone: Yilingia, a segmented creature found lying at the end of its own trail. Ediacaran tracks are common. The bodies that made them almost never are. This one died where its tracks stopped.

Description and sources

Helicolocellus cantori (Wang et al. 2024, Nature), from a single layer of the Shibantan Member, Dengying Formation, Yangtze Gorges, China — undated itself, between ashes of 551.1/550.1 Ma below and 543.4 ± 3.5 Ma above. Holotype NIGP-176531: a conical body 291 × 108 mm on a stem; paratype 447 mm complete. The wall is a lattice of 12 × 8.9 mm rectangles, each divided into four (5.3 × 3.8 mm), each again into four (2.55 × 1.8 mm), with a faint fourth order — a Cantor-dust fractal known otherwise only in Palaeozoic glass sponges, where the grid is built of cross-shaped silica spicules. No mineral survives: the fossil is a flattened cast, like the soft-bodied Arborea and Wutubus in the same beds, while mineral-walled Cloudina keeps a 3-D tube. A Bayesian analysis (79 taxa, 235 characters) places it as a stem-group glass sponge — outside the living ones precisely because it has no spicules — and the authors say the placement fits either spicules invented later or spicules lost. In 2026, Rossi et al. dated the last common sponge ancestor to about 608 Ma with no spicules (P ≥ 0.975) and four separate origins of silica spicules; a preprint by Cui et al. found ~86% of Cambrian spicules preserved as moulds, casts or replacements.

PRINT-READY, FROM THIS CHANNEL

The Deep Time Record, Vol. 01 — the cores, the curves and the clocks

https://therepository.gumroad.com/l/deep-time-record

PRIMARY SOURCES

ARCHIVE IMAGES — author, date, archive and licence.

  • NOAA Okeanos Explorer Program, Gulf of Mexico 2012 Expedition — 2012 | Wikimedia Commons | Public domain
  • NASA Earth Observatory, ISS astronaut photograph — 2009 | Wikimedia Commons | Public domain

... and 5 more, each credited on screen in the video and listed in this project's manifest.

The 2 generated plates were made for this channel with FLUX.1-schnell. No person is depicted in any of them.

Deep Time Archives rebuilds the pre-human Earth from the physical record — outcrops, drill cores, golden spikes, catalogued specimens. Every episode starts from an object you can go and check, and the sources are listed above.

About the images. The fossil itself is not shown in a photograph: no image of it is openly licensed for reuse here. Its shape and the boxes of its wall are drawn in code from the measurements published in 2024. The Yangtze Gorges appear from orbit (ISS astronaut photograph, NASA, 2009). Every sponge shown is not the fossil: living Venus flower baskets (NOAA, public domain; Owen Anderson, NIWA, CC BY 4.0), a Devonian glass sponge from New York (James St. John, CC BY 2.0) and Walcott's 1920 plate of the Cambrian sponge Diagoniella. The Cloudina shown is in 545-million-year-old limestone from Canada, not from the Shibantan. The storm sea and the limestone shelf are AI-generated reconstructions of places; no organism is generated. This video is labelled as containing altered or synthetic content. The diagrams are drawn in code from the published numbers.

#DeepTime #EarthHistory #Ediacaran #Paleontology #Evolution