The Many Origins

Chapter 6 — The Bush, Revisited

STATUS: v0 (all five parts drafted) Documentary parallel: chapters/06-the-bush-revisited.md (Beats 6.1–6.8 + closing card) Last updated: 2026-05-22

What this chapter does

The explainer began with a picture: the tree of life, with the origins of multicellularity scattered across all three of its branches, none of them connected by descent. Across five chapters we have walked through what the field knows about how those origins worked. The problems each lineage had to solve, the molecular tools each one reached for, the traps that locked some of them in, the doors that swung back open in others. The picture is now fully populated. Every dot the chapter is about to gesture at has, by now, been a chapter or part of one.

This chapter does no new science. It is an audit of the picture we have drawn. It looks at what the corpus defends: six claims, plainly stated, each a reminder of where in the explainer the reader met it in full. And at equal length, what the corpus does not settle: six things the field has framed but not answered, several of which it may never answer. A reader-facing account that hides the second list is lying about the science.

By the end of this chapter you should:

  1. Be able to state, in plain terms, roughly six claims the explainer’s corpus defends, and roughly six it does not.
  2. Understand that the symmetry between those lists is the project’s structural honesty, not a cop-out.
  3. Know what Auroralumina attenboroughii is, when it lived, and why it is the closing image.
  4. Have absorbed the distinction between the first time we can see it and the first time it happened, and accept that the second is permanently out of reach.
  5. Recognise the closing line as the corpus’s epistemic stance, compressed into one sentence.
  6. Leave with the bush, and not the ladder, as the residual mental image.

The chapter is shorter than the others by design. The reader has come a long way; the work of this chapter is to slow down.

The chapter is in five parts.

PartTitleApprox. length
1Return to the bush~800 words
2What the evidence supports~1,300 words
3What the evidence does not settle~1,600 words
4A recent gain, and a closing image~1,500 words
5We will probably never know what was~700 words

Part 1 — Return to the bush

The tree, one more time.

Three deep branches: bacteria, archaea, eukaryotes. On each of them, lights. Every light is a separate lineage that, somewhere along its own line of descent, learned how to be more than one cell. Bacteria: at least four lights. Archaea: one, with a strong second alongside. Eukaryotes: somewhere between sixteen and a hundred, depending on which question you ask of the same tree [Grosberg & Strathmann 2007; Knoll 2011; Lamża 2023].

Chapter 1 drew that picture. The chapters after it added texture rather than lineages, things you now see when you look at the same image again.

You see why the cells in each lineage stayed together at all. The same family of problems coming around again on different parts of the tree, asked of cells with completely different molecular toolkits: how to stick, how to talk, how to divide the work, how to stop one cell from cheating on the rest. That was Chapter 2. Behind the dots, you see the molecules each lineage reached for, and you notice that almost none of the parts were invented when the multicellular lineage came along. They were lying around in the single-celled ancestors already, doing other jobs, waiting to be rewired. That was Chapter 3.

On the lineages that crossed the second threshold, the lineages that became complex, you see tiny marks on each one. Mutations that made going back hard. Genetic decisions that, once made, were difficult to unmake. That was Chapter 4: the ratchets. Snowflake yeast, MuLTEE, Type 1 and Type 2 [Libby et al. 2016; Bozdag et al. 2021; Tong et al. 2025]. The reason complex multicellularity, once acquired, tends to stay acquired.

And on some lineages (yeasts inside the fungi, Helicosporidium inside the green algae, certain cyanobacterial lineages) you see the lights dimming. Or going out. Reversions. Chapter 5: the door swings both ways. The same chapter showed that the door swings inside the body, too. Cancer is what it looks like when individual cells, inside a complex multicellular animal, fall back to programs their ancestors used before the multicellular bargain was struck [Aktipis et al. 2015].

That is what you now see when you look at the bush. The dots are the same dots. You can read them now.

This chapter does not redraw the picture. Everything that follows has been set up. Its job is to audit. We look at the picture we drew, and we say honestly which parts of it the evidence actually defends, and which parts the evidence does not yet, or never will, settle.

The audit comes in two lists, as the documentary does it. Six claims the evidence supports, one paragraph each, each a callback to the chapter where the reader met the full mechanism. Six things the evidence does not settle. They sit side by side. The second list is at least as long as the first, and at least as substantial. That is what intellectual honesty about a young field looks like in print.

One through-line worth noticing before we go in. Many simple origins of multicellularity. Few complex ones. The complex ones almost all routed through the same mechanism: cells dividing and not separating, staying clonal. That contrast, more than any single lineage, is what Chapters 1 through 5 leave a reader with. It is the residue of the whole book.

After the two lists, one short part on a recent gain (a place where the picture has sharpened in the last few years, the Ediacaran) and a closing image, a small fossil from a rock outcrop in central England that has become, in the last three years, the earliest visible member of the lineage that contains modern jellyfish [Dunn et al. 2022]. And then one final sentence.

→ Continue with Part 2 — What the evidence supports.

What this part draws on:

  • The polyphyly bush, fully populated: content/01-polyphyly/independent-origins.md; cumulative across Chapters 1–5.
  • The count framings (16–25; 45; ~100): [Grosberg & Strathmann 2007]; [Knoll 2011]; [Lamża 2023].
  • The chapter’s structural commitment to the symmetry between known and unknown: BIG-PICTURE.md §“What we are confident about” and §“What we don’t know”; chapters/06-the-bush-revisited.md Beat 6.1.

Part 2 — What the evidence supports

Here is the audit. Six claims the corpus defends. One paragraph each. Each one points back to the chapter where the reader met it in full.

1. Polyphyly is real, and extends to all three domains.

Multicellularity has been invented independently in many lineages, across all three domains of life. The conservative count sits between sixteen and twenty-five origins; the recent careful inventory using explicit criteria puts the eukaryote-only count at forty-five, with a projection toward roughly a hundred once still-unsampled microbial-eukaryote groups are properly looked at [Grosberg & Strathmann 2007; Knoll 2011; Lamża 2023]. Until 2023 the archaeal branch had a question mark on it; two studies have now removed that question mark [Tang et al. 2023; Rados et al. 2025]. The exact count is contested. The fact of independent origin is not. This is Chapter 1’s central claim, demonstrated lineage by lineage in Parts 2 through 5 of that chapter and given its closing image in Part 7.

2. The same problems recur.

Independent lineages, with no shared multicellular ancestor and no common molecular toolkit, kept running into the same set of problems on the way to being more than one cell. Cells had to stick together. Cells had to talk to each other. Cells had to divide the work, some cells doing one job, others doing another, and they had to find a way to suppress the cells that would otherwise cheat: divide too fast, take the benefits of being in a group without paying the costs. The molecules each lineage reached for to solve these problems differ in almost every case. The shape of the problem is the same. This was Chapter 2, and the convergence at the level of problem, with divergence at the level of molecule, is one of the most striking patterns in the data.

3. The molecular toolkit predates the trait.

When biologists started sequencing the genomes of single-celled organisms closely related to multicellular lineages, they expected to find the multicellular toolkit absent. They found it almost entirely present. The adhesion proteins that hold animal cells together, the signalling proteins that animal cells use to talk to each other, the transcription factors that switch animal cell types on and off: all of them were already in the single-celled relatives of animals, doing other jobs [King et al. 2008; Sebé-Pedrós et al. 2017]. The same pattern shows up everywhere we have looked. The master regulator gene that holds the sterile cells of Volvox in their permanently somatic state, regA, was repurposed from a gene that, in Volvox’s single-celled ancestors, was part of the cell’s day-night machinery [Nedelcu & Michod 2006]. The lineages that became multicellular rewired old parts rather than inventing new ones. This was Chapter 3.

4. Ratchets explain why complex multicellularity sticks.

The complex multicellular lineages, the ones with internal plumbing, multiple cell types, the kind of body that needs cells specialised to feed other cells, are lineages in which a particular kind of evolutionary change has accumulated: mutations that, once they happened, were costly to undo. Either because they were good in the multicellular context and bad on their own (Type 1), or because they sealed off the cellular machinery you would need to live as a single cell again (Type 2). Once enough of those changes have accumulated, the lineage cannot easily go back. It is locked in. The MuLTEE experiment at Georgia Tech has watched this kind of lock-in happen in real time in snowflake yeast, across six hundred daily transfers. Macroscopic, mechanically tough multicellular organisms emerged from a microscopic yeast ancestor in the lab, and stayed multicellular even when the selection that built them was relaxed [Libby et al. 2016; Bozdag et al. 2021; Tong et al. 2025; Pentz et al. 2022]. That was Chapter 4.

5. Reversion to unicellularity is real.

The locking-in is a bias, not a one-way ticket. Multicellular lineages have gone back to being single-celled. Some yeasts descend from filamentous fungal ancestors and have lost the body plan their ancestors had [Naranjo-Ortiz & Gabaldón 2020]. The same has been documented in green algae, in cyanobacteria, and experimentally in snowflake yeast itself; within hundreds of generations, the right selection pressure pulls cells back apart [Kuzdzal-Fick et al. 2019]. The count of origins is, for this reason, a net number: forward transitions minus the ones that reverted and erased themselves. Reversion is rare enough to leave most complex lineages still multicellular hundreds of millions of years on, and common enough that the count we have is, in principle, an undercount. This was Chapter 5’s first half.

6. Cancer is cooperation breakdown.

Inside a complex multicellular animal, the cells are still cells. They still have, in their genomes, the parts you would need to live and divide on your own. Most of the time those parts are switched off, regulated, kept in line by the cooperation programs that built the body. When the regulation fails, when a lineage of cells inside the body starts dividing too fast, ignoring stop signals, taking resources without contributing, what you have is cancer. The framework that treats cancer as a failure of multicellular cooperation, rather than as a foreign invader, is by now well established. The cells that go cancerous are doing something cells have always been able to do; the multicellular bargain just normally prevents it [Aktipis et al. 2015]. This was Chapter 5’s second half.

That is the audit list. Six claims. None of them rests on a single paper; each holds up across multiple independent lines of evidence. There is more in BIG-PICTURE.md and the corpus, but in the chapter where the reader meets it, this is what the explainer has earned the right to say plainly.

The list deliberately leaves something out. It does not say cancer is a return to an ancient unicellular program. That is the atavism hypothesis, and it lives, fairly, on the other list. The framework that cancer is cooperation failure is settled; the mechanism that says cancer is specifically a re-expression of pre-multicellular biology is not (Chapter 5 walked through where that debate sits). The line between item 6 above and what we don’t yet know is exactly the kind of line this audit draws.

→ Continue with Part 3 — What the evidence does not settle.

What this part draws on:

Part 3 — What the evidence does not settle

Most popular accounts of biology do not give the what we don’t know list this much room, and most do not give it equal weight. The genre’s failure mode is to keep the camera on the things the field has answered. The corpus’s choice, and this explainer’s, is to keep the camera on both, in equal time, and let the reader see that the second list is not shorter, not less important, and not closing soon.

Six items. Each one a question the field has framed cleanly but cannot yet answer. Some of them may close in the next decade. Some of them, as the closing line will say, will probably never close at all.

1. We do not know why multicellularity starts in any given lineage.

There is no unified answer to the question what triggered the move from one cell to many? The cyanobacterial answer is a metabolic trick: the nitrogen-fixation enzyme cannot tolerate the oxygen the photosynthetic cells of the same filament produce, so the lineage specialises in space. The Volvox answer is a co-opted regulatory gene reused from a circadian role to lock somatic cells out of reproduction. The Dictyostelium answer is an aggregative response to starvation, with cAMP repurposed from an even older job in single-celled amoebozoans. Each lineage seems to have its own story about what selection pressure pushed it past the threshold. There may not be a single answer to find. The triggers differ. The toolkits differ. So do the geological backdrops each lineage made its move in. Why life keeps inventing multicellularity is much better characterised than why a particular lineage invented it when it did. The second question may be irreducibly lineage-by-lineage.

2. We do not know what closed the door on the animal stem.

The closest single-celled cousins of animals alive today are the choanoflagellates, the lineage we met in Chapter 3, sister-group to animals, descendants of the same single-celled ancestor that gave rise to us. Some choanoflagellates form transient multicellular colonies. Some do not. None of them have made the transition that the animal lineage made: from facultative multicellularity (sometimes a colony, sometimes single cells, depending on conditions) to obligate multicellularity (always a colony, no going back to a free-living single cell as an independent organism). The animal lineage crossed that threshold and stayed across. The choanoflagellate lineage, in its modern members, did not. We don’t know the genetic basis of that step. We don’t know what specifically closed the door for the stem animal but left the choanoflagellates with the door propped open. Recent work on facultative choanoflagellates [Larson et al. 2023; Ros-Rocher et al. 2025] is starting to map what cells do when they make the transition into and out of colonial form, but the kind of evidence that would settle the question (cell-by-cell readouts of which genes are switched on at the exact moment a population commits to being many cells rather than one) does not yet exist in any of these systems. The mechanism that locked our ancestors into multicellularity, while their cousins stayed facultative, is one of the central unsolved questions in the field.

3. We will never know how many ancient reversions left no trace.

Chapter 5 documented several reverters in the modern world. Lineages that were multicellular and lost it. Yeasts, Helicosporidium, certain cyanobacteria. What those cases tell you is that reversion happens. What they cannot tell you is how often it has happened in the past. A lineage that reverted to single-celled life and then went extinct, or that reverted long enough ago that all signs of its previous body plan have been erased from its genome, is invisible to us. The “count of independent origins” is by definition a net count: the forward transitions, minus the reversions that left no record. The historically true number is strictly greater than the number we can recover from living organisms and the fossil record combined. We can show that the gap is non-zero. We cannot say how large it is. This is a permanent uncertainty, not a temporary one.

4. We don’t know if the Volvox mechanism tells us much about the animal one.

Volvox, the famous green algal ball with a clean division of labour between somatic cells and reproductive ones, is the best-characterised molecular case of a complex multicellular transition that biology has. We can identify the gene that holds the somatic cells in place. We can compare Volvox to its unicellular cousins. We know where in the Volvox genome the change happened, and where it came from [Nedelcu & Michod 2006; Hanschen et al. 2016]. So a reasonable question is: does the Volvox story generalise? When animals made their transition, did something analogous happen? Did a gene that was doing one job in the single-celled ancestor get repurposed to keep a new cell type in line?

We do not know. Volvox is young. Its germ-soma split is something like two hundred million years old, far younger than the animal origin, and the lineage sits within a single algal family. Extrapolating from Volvox to the deep transitions in larger lineages is a guess, not a finding. The general lesson, that regulatory rewiring of pre-existing parts does the work, is well supported. The specific lesson, here is what the rewiring looked like in animals, is not what Volvox gives us. Volvox gives us a multicellular transition. There is no general transition.

5. We do not know why some clonal lineages get locked in fast, and others never do.

Chapter 4 set up the contrast. Some clonal lineages, given enough time, accumulate the kinds of mutations that lock multicellularity in: the ratchet works on them, and they end up unable to revert. Volvox locked in within a couple of hundred million years. Animals locked in too, on a longer timescale. Other lineages, with similar molecular toolkits available to them, have not. Snowflake yeast in the lab accumulates Type 1 ratchet changes over hundreds of generations. In nature, plenty of clonal lineages have stayed facultative for hundreds of millions of years and never crossed the threshold the snowflake-yeast experiment routinely crosses in months. Why is the ratchet so much stronger, or so much faster, in some lineages than in others? The field does not have a quantitative answer. The MuLTEE work, and the broader Ratcliff-lab programme that built it, has shown that the framework is real and measurable [Libby et al. 2016; Bozdag et al. 2021; Tong et al. 2025]. Whether you can predict, from looking at a lineage’s molecular features, how strong its ratchet will be is still open work.

6. We do not know what the first obligate multicellular animal ancestor was actually like.

This is the largest unknown on the list, and the most pointed. The molecular-clock estimates for the origin of animals place it somewhere around seven to eight hundred million years ago, with wide credible intervals [dos Reis et al. 2015]. The first body fossils that look convincingly like animals do not appear in the rock record until perhaps a hundred and fifty million years later. Whatever made the transition from facultative to obligate multicellularity on the animal stem, whatever lineage actually crossed that threshold in whatever ocean, is for our purposes gone. It left no descendants that can be unambiguously identified. The molecules its cells used to stick together, the signals they sent to each other, the way they decided which cells would do which jobs: all of this we are trying to reconstruct, indirectly, from what its surviving distant cousins still do and from what its eventual descendants ended up doing.

The deepest branches of the animal tree, the sponges and the comb jellies, disagree in the current literature about which of them branched off first, and that disagreement (which we walked through in Chapter 1) changes what we should imagine the very first animal was like. Sponge-first would mean the first animal had no neurons and no muscles. Ctenophore-first would mean the first animal had both, and the sponges lost them. The field has not converged [Schultz et al. 2023; Copley 2025; McCarthy et al. 2022]. Newer evidence from chromosome organisation pushed one way in 2023; a 2025 statistical critique pushed it back. The animals’ founding lineage is, in real terms, recoverable in outline only. The detail (what its cells actually looked like, what its first body actually was) is gone.

You may notice that the unknown list is not a list of things the field is failing to study. Each item is a question the field has framed precisely, precisely enough to know that the question is open, and to know roughly what kind of evidence would close it (or has already established that it will not close). That precision is itself a kind of progress. The right counterpart to we know the polyphyly thesis is true is we know what we don’t know, and we know why we don’t know it.

The documentary makes this point in the strongest language it uses anywhere in the script: a film that hides the second list from you is not telling the truth. The explainer puts the same point more quietly. The two lists sit side by side on the page. They are roughly the same length. They are roughly equally well-defined. That is what intellectual honesty about a young, rapidly moving field looks like.

→ Continue with Part 4 — A recent gain, and a closing image.

What this part draws on:

Part 4 — A recent gain, and a closing image

The last part might have left you wondering whether the unknown list is only growing, whether the field’s honest report is that it just keeps finding new things it can’t answer. It isn’t. There are open questions that have closed in the last few years. There are open questions that have substantially narrowed. The unknown list is not a one-way ratchet either.

To make that concrete, one area of the picture has sharpened considerably since the early 2020s. It is the right area to close on, because it is where the documentary closes too: the late Ediacaran, the last sixty million years before the Cambrian, and a small fossil from a cliff face in central England.

A short period of geological time, recently re-read

The Ediacaran is the geological period running from roughly six hundred and thirty-five million years ago to roughly five hundred and thirty-eight million years ago [Narbonne 2005]. It sits between the last great Cryogenian glaciations on one side, and the start of the Cambrian on the other. For most of the time it had a name, it was treated as a mystery. The macrofossils it contains (soft-bodied impressions in sandstone, found in places like the Charnwood Forest in England and Mistaken Point in Newfoundland) looked alien. They were assumed, for a long time, to be a separate experiment: multicellular organisms unrelated to anything alive today, an alternative lineage that did not survive.

That reading has steadily come apart. Recent careful work on the morphology and development of the major Ediacaran fossils has placed most of them where the molecular-clock estimates have always said the early animals should sit: on the animal stem, before the major modern animal phyla had split off from each other. The frond-like forms called Charnia and its relatives, which look at first glance like nothing alive today, have been resolved as early relatives of animals like the cnidarians and their kin [Dunn et al. 2021]. Cloudina, an early Ediacaran organism that built a tube-like skeleton and had what looks, in three-dimensional preservation, like a through-gut, has been resolved as an early member of the lineage that contains animals with a two-sided body plan, including, eventually, us [Schiffbauer et al. 2020]. Recalibrated molecular clocks place the deep divergence of the major animal lineages in the early Ediacaran, not after it [Carlisle et al. 2024]. The clean five-hundred-and-forty-million-year line that the popular “Cambrian explosion” story used to draw between no animals before and all the animals after is dissolving in the literature. The animals were already there. The Cambrian is where they got dramatic.

Coupled, not caused

There is a connected reframing about the chemistry of the ocean those animals lived in. The simple story (oxygen rose in the sea, then animals could evolve) has had the causal direction tightened up. The new picture is reciprocal. The Ediacaran biota was, in part, what was doing the oxygenating.

The mechanism, when you spell it out, is more concrete than it sounds. The Mistaken Point fossils, in Newfoundland, preserve a community of sessile frondose organisms living on the seafloor, anchored to the substrate, sticking up into the water column, filtering food out of the water that passed over them. When researchers built computational models of the water flow around reconstructed fronds, modelling the seafloor community as you might model the airflow around the trees of a forest, they found that the bodies of the fronds themselves stirred the water column locally. They generated mixing. They moved oxygenated surface water down into the layer just above the sediment, where the slower-moving microbial communities lived [Gutarra et al. 2024]. (Voice rule 4, for the bookkeeping: when we say the fronds “stirred” or “ventilated” their habitat, we don’t mean they actively did anything. They sat there. Their shapes did the mixing, the way trees in a forest passively change the wind. The metaphor is doing the work; the underlying mechanism is physics, not behaviour.)

The reframed reading: oxygen and complex animal life on the Ediacaran seafloor were rising together, in a feedback loop in which the rising animals were partly doing the work [Lenton & Daines 2018]. Animal multicellularity probably already existed well before this; molecular-clock work places the deep root of the animal lineage at roughly seven to eight hundred million years ago, in an ocean with much less dissolved oxygen than the modern one [Sperling et al. 2018]. What the Ediacaran does is show us the moment that lineage became visible. One of the things that made it visible was that its bodies were now large enough, and densely enough packed, to start changing the chemistry of the water above them.

Animals existed long before they became spectacular. They appear in the rock record at the moment they became spectacular enough to leave one.

The closing image

Among the Ediacaran fossils, one in particular has become the chapter’s closing image. It is small, about twenty centimetres tall, and was found in 2022 in the Charnwood Forest in central England, preserved in fine detail under a layer of volcanic ash that buried the seafloor it lived on, freezing it in place. Its name is Auroralumina attenboroughii.

What it looks like, from the reconstruction: a rigid, polyhedral organic skeleton, branching into two side-by-side polyps at the top, each polyp ringed with what look like densely packed simple tentacles. It lived on the seafloor about five hundred and sixty million years ago [Dunn et al. 2022]. It was a small filter-feeder, almost certainly capturing particles from the water moving past it with those tentacles.

What earned it its place at the close of the chapter is what it is: the oldest known crown-group cnidarian. Cnidarian is the lineage that contains the modern jellyfish, the sea anemones, the corals, the small freshwater Hydra. Crown-group and stem are paleontology’s pair of terms for sorting out where in a lineage a fossil sits. The crown of a group is the part of the family tree that contains the most recent common ancestor of every living member and all of that ancestor’s descendants: modern jellyfish, modern corals, modern anemones, and every fossil that branches off after the crown ancestor lived. The stem is the long line leading up to that crown ancestor, the older fossils that are clearly on the way to becoming cnidarians but had branched off before the modern cnidarian groups split from one another. Auroralumina sits in the crown, not on the stem. So Auroralumina is the earliest clear member of a still-living animal phylum, anywhere in the rock record. The earliest moment we can look at a fossil and say: this organism is recognisably an early member of a group whose descendants are still in the ocean today.

It is a remarkable thing. It is also worth being precise about what it is not. Auroralumina is not the first multicellularity. The cyanobacterial filaments we met in Chapter 1 had been multicellular for two billion years before Auroralumina lived. It is not the first animal; the molecular-clock estimates place the deep root of the animal lineage hundreds of millions of years earlier, and other Ediacaran fossils probably do represent earlier members of the animal stem. It is not even the first cnidarian, in any sense that matters; the lineage that contains Auroralumina must have a stem reaching further back than its earliest known fossil. What Auroralumina is is the first moment we can see the trick clearly, in a lineage that is still alive. The first time the lineage that contains modern jellyfish becomes visible to a fossil hunter. The first time, in the rock, that we can point and say this is what it looked like.

Between the first time it happened and the first time we can see it, the chapter closes on the second.

→ Continue with Part 5 — We will probably never know what was.

What this part draws on:

Part 5 — We will probably never know what was

One last time, slowly, the distinction.

The Ediacaran shows us the trick (animal-grade multicellularity, in a lineage whose descendants are still in the ocean today) at the moment it became visible. Auroralumina is what visibility looks like. A small, sessile, two-polyp animal pinned under volcanic ash on a Leicestershire seafloor five hundred and sixty million years ago, preserved in detail clear enough that we can recognise it as an early jellyfish relative. The first time the rock record gives us, unambiguously, a member of a lineage that is still alive.

That is the first time we can see it.

It is not the first time it happened.

By the time Auroralumina lived, the animal stem lineage (the lineage that connects every living animal back to a common ancestor) had probably been making and unmaking some form of multicellularity for two hundred million years already. The molecular-clock work places that deep root at roughly seven to eight hundred million years ago, with credible intervals wide enough to drive a bus through [dos Reis et al. 2015]. Somewhere in that long pre-Ediacaran interval, the lineage crossed the threshold from facultative to obligate multicellularity, the door that Chapter 4 spent a chapter on, the one we still cannot identify the closing mechanism for. Somewhere in there, the first cells on the animal lineage that were truly committed to being part of a body actually did it.

What that organism looked like, what its cells were doing, what selection pressure pushed it across the line: all of that is information the rock record will not give us. The fossil window opens in the late Ediacaran. What came before is a long stretch of small, soft-bodied, low-oxygen-ocean ancestry that left almost no trace. The molecular-clock estimates can tell us, with wide intervals, roughly when. The comparative genomics can tell us, with imperfect resolution, roughly what regulatory parts were available. The closest living single-celled cousins, the choanoflagellates, can tell us, indirectly, what some of those parts still do in cells that never made the transition.

None of that will reach back through the fossil silence to the lineage’s founding cells.

This is true beyond the animal case. It is true everywhere on the bush. The cyanobacterial filaments we met in Chapter 1 are at least two billion years old; the molecular machinery that builds heterocysts is older. We will never know which cyanobacterium first divided without separating. The fungal lineages whose mushroom-forming arms have made the complex-multicellular transition eight or eleven times: we will never see the cells that first committed in any of those eight or eleven cases. Volvox’s germ-soma split is two hundred million years old (recent, by these standards) and we cannot see its founder either. The lineages that reverted left, by definition, the least of all. The dimmest lights on the bush, in Chapter 5’s image, are the ones whose path is most thoroughly erased.

Most of what happened on the bush, then, is not visible to us now. The forty-five well-attested origins and the projected hundred are a count of the cases that are visible: lineages that survived, that left descendants, that became preservable, that someone has gotten around to sampling. The ones we missed are gone in a way that is not just not yet found, but mostly not findable in principle. Geology and biology have erased their own work over billions of years.

The closing chapter of the documentary holds a single image of Auroralumina on the screen for what feels like a long time, in silence, and then ends with one sentence. The explainer is not built for silence. The sentence is the same one. It is the corpus’s epistemic stance compressed into seven words. It does not promise certainty in the future. It does not soften. It does not pretend the field is converging on an answer it cannot, even in principle, find.

We will probably never know what was.

This explainer is dedicated, as the documentary is, to two contemporary researchers whose work has carried more of the project’s weight than any others: Łukasz Lamża, whose 2023 paper made the polyphyly thesis quantitative (the source of the count that escalates from sixteen to forty-five to a hundred, depending on the question), and the Ratcliff lab at Georgia Tech, who built MuLTEE and gave the field its clearest demonstration so far of a ratchet operating in real time [Lamża 2023; Bozdag et al. 2021].

What this part draws on:

End of Chapter 6 (draft state).

This chapter closes the explainer. It is shorter than the others by design (roughly six and a half thousand words across five parts, against the ten thousand or so of Chapter 1) and the voice in it is slower. There is no Chapter 7. The reader who has read all six chapters now has the same picture of the polyphyletic origins of multicellularity that the underlying corpus and the documentary script defend, in a register that assumed no prior biology and built every term in plain English. The closing image is Auroralumina. The closing line is locked. There is nothing after it.