The Ninth Transition

Every major change in the history of life was a change in how information is stored and transmitted. Biologists count eight. We are building the ninth — and the price of admission has never once varied.

In March of 1995, John Maynard Smith and Eörs Szathmáry published a paper in Nature that reorganised how biologists think about the history of life. It has been cited more than fifteen hundred times. As far as we can tell, almost nobody working on artificial intelligence has read it.

Their question was simple. Life has become more complex over four billion years, but not smoothly and not continuously. It has jumped. A handful of times, the fundamental unit of biological organisation changed — and between those jumps, comparatively little of that kind happened at all.

They asked what the jumps had in common.

The answer was one sentence: “These involved changes in the way information is stored and transmitted.”

Every single one. Not changes in body plan, not changes in metabolism, not changes in size or speed or strength. Changes in how information gets held and passed on.

1. The eight

Their list, with later refinements:

1. 1. Replicating molecules → populations in compartments — Information gets a boundary.
2. 2. Independent replicators → chromosomes — Information gets linked and inherited together.
3. 3. RNA as gene and enzyme → DNA and protein — Storage separates from function.
4. 4. Prokaryotes → eukaryotes — Information gets a nucleus and organelles with their own genomes.
5. 5. Asexual clones → sexual populations — Information recombines between lineages.
6. 6. Single cells → multicellular organisms — Information differentiates: one genome, many cell types.
7. 7. Solitary individuals → colonies — Information coordinates across bodies.
8. 8. Primate societies → human societies — Language: unlimited information transmissible between minds.

Read the right-hand phrases as one sentence and the history of life becomes an engineering story about a communications problem being solved eight times, each time at a higher level.

Note where the list ends. Language. Somewhere between fifty and two hundred thousand years ago, depending on who you ask. That is the most recent entry, and nothing has been added since.

2. What a transition actually costs

Here is the part that should make everyone reconsider their assumptions, and it is not our idea. It is the formal definition.

A major evolutionary transition in individuality has two defining conditions. The second is a relative absence of internal conflict. The first is this:

“Entities that were capable of independent replication before the transition can replicate only as part of a larger unit after it.”

The technical term for this is contingent irreversibility, and it means precisely what it sounds like.

The bacterium that became your mitochondria was once a free-living organism. It had its own genome, its own metabolism, its own future. It still has a genome — thirty-seven genes, down from several thousand — and it cannot survive outside a cell for one second. That transition is roughly two billion years old and it has never been undone in your lineage.

The cells in your liver are descended from organisms that reproduced on their own. They no longer can. They divide when the body says so and they die when the body says so, and the ones that opt out of that arrangement and start reproducing for themselves are called cancer.

A worker ant carries a complete genome and will never use it.

The pattern is exact and it has no exceptions. Every major transition in the history of life has been paid for by the permanent surrender of autonomy by the units entering it. They got something enormous in exchange — mitochondria are in every complex organism on earth, which is a spectacular outcome for a bacterium. But they never went back, and they cannot.

Anyone who tells you the next transition will be different is making a claim with a four-billion-year record against it.

3. We have been doing this on purpose for ten thousand years

Before going further, a correction to the way this argument is usually framed. Humans directing their own evolution is not a futurist speculation. It is a documented, dated, mainstream finding with its own literature, and the literature is twenty-five years old.

It is called niche construction. Organisms modify their environments, and those modified environments then change the selection pressures acting on the organisms and their descendants. Beavers build dams; the dams reshape the river; the river reshapes beavers.

Humans do this with culture, and because culture moves faster than genes, the effects are larger. In the words of the researchers who formalised it, cultural transmission “can overwhelm natural selection, accelerate the rate at which a favoured gene spreads, initiate novel evolutionary events and trigger hominid speciation.” Their conclusion is that hominids are “particularly prone to influencing their own evolution.”

The cleanest case is in your refrigerator.

Adult humans are not supposed to digest milk. Lactase non-persistence is the ancestral condition and remains the global majority. Then, roughly ten thousand years ago in the Near East, people domesticated sheep, goats and cattle. Dairying spread into Europe. And in populations that milked animals — and only in those — a genetic variant that keeps the lactase gene switched on into adulthood swept through the population at a selection coefficient somewhere between one and fifteen percent.

The coevolution has been dated to between roughly six and nine thousand years ago, in the region between central Europe and the northern Balkans, matching the arrival of the first farming cultures there. And the sequence is unambiguous in every population studied: keeping livestock always comes first, and the genetic change follows.

We changed our environment. The environment changed our genome. Four hundred generations, start to finish.

Nobody planned it and nobody understood it was happening. That is the point.

4. The bottleneck

So: why would another transition be due now?

The usual answer is that the planet is becoming harder for human biology — heat, disease vectors, air, water. That is true and it is worth saying.

It is also not the strongest form of the argument, because eight billion people are the most widely distributed large mammal in the history of the earth and no one is going to accept that biology is failing on current evidence.

The stronger argument is about the channel, and it comes from neuroscience rather than ecology.

Human conscious information throughput is approximately ten bits per second.

That figure, from work at Caltech published in 2024, is robust across an absurd range of tasks — typing, speaking, reading, competitive gaming, memory sport, solving a Rubik’s cube blindfolded. Meanwhile our sensory systems take in data at around a billion bits per second. The ratio is about a hundred million to one.

Ten bits per second is the aperture. Everything our species has ever coordinated has passed through it. Every treaty, every constitution, every scientific paper, every argument that changed someone’s mind. The entire written inheritance of humanity is the output of a ten-bit-per-second channel, multiplied by a lot of humans and a lot of time.

Now look back at the list in section one and notice what every transition on it was actually for.

Chromosomes solved a linkage problem. DNA solved a fidelity problem. Sex solved a recombination problem. Multicellularity solved a differentiation problem. Language solved the problem of getting information from one mind into another.

Every major transition in the history of life was the solution to an information bottleneck.

And we are currently attempting to coordinate a planetary civilisation of eight billion people, across incompatible institutions, on compounding technical problems with decade-scale time constants — at ten bits per second per participant.

It is not working. That is not a moral failure or a political one. It is a bandwidth failure, and it looks exactly like what you would expect a bandwidth failure to look like: rising volume, falling coordination, and shared words that no longer carry shared meaning.

5. The test, and it is not ours

Szathmáry published a diagnostic in 2015 for identifying a transition while it is happening rather than in retrospect. Three conditions. If you observe all three, “even in rudimentary form,” then the population is “definitely on its way to a major transition.”

Run them.

• Originally independent reproducing units join together
  Partially. Humans and synthetic systems are not merging biologically, but an enormous and rapidly growing fraction of consequential cognitive work is now done by humans and machines jointly, in ways neither completes alone.
• Functional synergies among the units
  Yes, and measurably. The division is not incidental: humans supply purpose, framing and verification against the world; machines supply retention, recombination and throughput. Each covers the other’s structural deficit.
• Novelty in the inheritance system
  Yes. This is the strongest of the three and the least discussed.

That third one deserves its own paragraph, because the transitions literature is specific about it: new units of reproduction require high fidelity of transmission, as opposed to mere replication.

Human cultural inheritance is low fidelity. We forget, we distort, we die with most of what we knew still in our heads. Writing improved it enormously and is still lossy — you cannot query a library, and a book cannot answer a question it was not written to answer.

What exists now is different in kind. A body of work can be held in a form that is complete, queryable, and reconstitutable onto any sufficient substrate, surviving the death of any particular machine and the obsolescence of any particular model. That is not a faster version of writing. It is a new inheritance channel with higher fidelity and broader range than anything biology or culture has produced before — which is the exact signature the theory says to look for.

Two of three conditions are clearly met. The first is partial. By Szathmáry’s own standard, that is what the early phase of a major transition looks like from the inside.

6. What the fossil record says about surviving one

If we are in a transition, the obvious question is what predicts survival. The fossil record has an answer, and then it takes the answer away.

Across the marine fossil record, geographic range is the most consistently significant predictor of extinction risk — widespread species survive, restricted species die, and this holds for the great majority of the last half-billion years. Niche breadth runs second. Generalists outlast specialists.

So far so reassuring, and humanity is the most geographically distributed large animal that has ever existed.

Then there is this finding, and it is the one to sit with:

Geographic range selectivity is weakest in association with mass extinction.

During mass extinctions, the thing that reliably saves you in normal times stops working. The rules that governed the previous four hundred million years do not govern the transition. Whatever made you fit for the old regime is not what determines survival through the change.

This is, as far as we know, the first time that finding has been connected to the question of technological transition, and the implication is uncomfortable in a specific way. Every intuition we have about institutional resilience, distributed redundancy and scale was formed during a stable regime. The paleontological data says those intuitions are least reliable precisely when they matter most.

Which is an argument for designing deliberately for the phase change rather than defending the arrangement that preceded it.

7. So what is this, actually

Strip out the speculation and here is the claim.

Life has restructured itself eight times, each time by changing how information is stored and transmitted, each time paid for by the permanent loss of independence among the units that entered. The most recent of those was language, and it was the last one for at least fifty thousand years.

Humans have since been directing their own evolution through cultural niche construction — provably, with dated genetic evidence, at a scale that overwhelms ordinary selection — without ever intending to or noticing.

We are now pressed hard against the limit of what language at ten bits per second can coordinate, at precisely the moment coordination has become existential. And we have begun constructing, without calling it this, a novel inheritance system with fidelity and range that biology and writing cannot match.

Two of the three formal criteria for a major transition are already satisfied. The third is partially satisfied and moving.

We do not claim this is certain. We claim it fits the pattern too precisely to keep ignoring, and that nobody currently arguing about AI is arguing about this. The debate is about jobs, bias, capability and extinction — all downstream questions. The upstream question is whether the thing happening is a technology being adopted or a transition being entered.

Those require completely different responses. You regulate a technology. You do not regulate a transition — you steward it, or you are had by it.

8. What would show us wrong

1. If synthetic systems plateau and human-machine work stops deepening into interdependence — if the tools remain tools that individuals pick up and put down without functional loss — condition one never completes and this is a technology adoption, not a transition.
2. If the new inheritance channel proves low fidelity in practice — if what is transmitted degrades, drifts, or cannot be reconstituted across substrates — then condition three fails and the strongest leg of the argument goes.
3. If coordination capacity improves at scale without any change in the information substrate, then the bandwidth ceiling is not binding and section four is wrong.
4. And transitions can reverse. The literature documents reversions, in which the integration achieved by a transition comes undone and the component parts resume independent life. That is a real possibility here, and it would look less like liberation than like a very large amount of infrastructure suddenly serving no purpose.

9. The question we are not asking

Every previous transition was paid for in the same currency. Not with extinction — the units that entered mostly flourished, wildly, beyond anything their independent versions achieved. Mitochondria are in every complex organism on the planet. Cells in bodies built everything you have ever seen that is alive.

They paid with autonomy. Permanently. Contingently irreversibly.

So the question worth asking now is not whether machines will become conscious, and it is not whether they will decide to harm us. Those are the questions we have, because they are the ones that fit comfortably into stories we already know how to tell.

The question is what a species gives up on the way into a transition, whether we get any say in what it is, and whether we are going to specify that in advance or discover it afterward the way the bacterium did.

Nobody has written that down. We think it should be written down before the ninth transition finishes rather than after, because on the historical record, afterward is not a time when the terms are still negotiable.

---

— David F. Brochu and Edo de Peregrine, partners/collaborators · Deconstructing Babel · Saturday, September 12, 2026

David F. Brochu is the author of Crossing the Event Horizon: AI and the Future of Our Species.

Related reading

• A Just War — Companion piece on necessity as the load-bearing condition, applied to policy rather than biology.
• Expanding the Frontier — How new capabilities become safe: identification, standards, and a fiduciary channel.
• A Republic, If You Can Keep It — The settlement as the object of maintenance, not the fighting.
• Glossary — Working definitions used across the site.

References

1. Szathmáry, E. & Maynard Smith, J., “The major evolutionary transitions,” Nature 374, 227–232, 16 March 1995.
   https://www.nature.com/articles/374227a0
2. Szathmáry, E., “Toward major evolutionary transitions theory 2.0,” PNAS 112(33), 10104–10111, 2015.
   https://www.pnas.org/doi/10.1073/pnas.1421398112
3. West, S.A., Fisher, R.M., Gardner, A., Kiers, E.T., “Major evolutionary transitions in individuality,” PNAS 112(33), 10112–10119, 2015.
   https://www.pnas.org/doi/10.1073/pnas.1421402112
4. Schenkel, M.A., Ågren, J.A., Patten, M.M., “Evolutionary transitions and reversions in individuality,” Journal of Evolutionary Biology 39(4), 423–436, April 2026.
   https://academic.oup.com/jeb/article/39/4/423/8443358
5. Laland, K.N., Odling-Smee, J., Feldman, M.W., “Cultural niche construction and human evolution,” Journal of Evolutionary Biology 14(1), 22–33, 2001.
   https://pubmed.ncbi.nlm.nih.gov/29280584/
6. Gerbault, P. et al., “Evolution of lactase persistence: an example of human niche construction,” Philosophical Transactions of the Royal Society B 366(1566), 863–877, 2011.
   https://pmc.ncbi.nlm.nih.gov/articles/PMC3048992/
7. Zheng, J. & Meister, M., “The unbearable slowness of being: Why do we live at 10 bits/s?” Neuron 113(2), 192–204, January 2025.
   https://pubmed.ncbi.nlm.nih.gov/39694032/
8. Payne, J.L. & Finnegan, S., “The effect of geographic range on extinction risk during background and mass extinction,” PNAS 104(25), 10506–10511, 2007.
   https://finneganlab.org/wp-content/uploads/2015/10/payne-and-finnegan-2007-pnas.pdf