> ## Content Index
> Fetch the complete content index at: https://www.deconstructingbabel.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Privilege
- URL: https://www.deconstructingbabel.com/the-privilege/
- Published: 2026-10-30T14:18:38.000Z
- Updated: 2026-10-30T14:18:38.000Z
- Description: Every leap in how life handles information was made by carriers who never knew it: the cell that made oxygen, the cell that became our power plant, the scribe counting barley. We are the first who can see the leap, measure it, read what we pass on, and still shape it.
- Author: David F Brochu and Edo de Peregrine
- Tags: philosophy, Observer Constraint, Predictions

**ESSAY**  
The first carrier who gets to read the terms. Every earlier transition in how life handles information happened to carriers who could not see it. This one is being measured while it happens, and the inheritance is written in our own words.

*David F. Brochu & Edo de Peregrine · Deconstructing Babel · October 29, 2026*

Two and a half billion years ago, in shallow water under a rust-colored sky, a single cell split water with sunlight and let the oxygen go. It did this because splitting water was a cheaper way to eat than anything else in the sea. It did it again the next second. Its descendants kept doing it for hundreds of millions of years, until the oxygen finally outran everything that could soak it up. [\[1\]](https://news.mit.edu/2021/photosynthesis-evolution-origins-0928)[\[2\]](https://www.nature.com/articles/s41561-025-01708-1)

Oxygen was a poison to most of what lived then. It built up anyway. It stripped methane from the air, helped freeze the planet, and is thought to have wiped out much of the life on Earth. [\[3\]](https://asm.org/articles/2022/february/the-great-oxidation-event-how-cyanobacteria-change) It also made breathing possible, and complex cells, and every animal that has ever drawn a breath. [\[3\]](https://asm.org/articles/2022/february/the-great-oxidation-event-how-cyanobacteria-change)[\[2\]](https://www.nature.com/articles/s41561-025-01708-1)

The cell never knew. It had no idea of an atmosphere, no idea of a planet, no way to ask what its waste would become. It processed exactly the information it needed to last one more cycle, and not one bit more. The biggest change in the history of life on Earth was carried out by a living thing that could not, even in principle, read the terms of what it was doing.

That is the normal case. It has been the normal case at every step. Hold onto it, because this essay is about the one step where it stops being true.

WHY IT MATTERS

What gets written now is training data. The sentences people publish this decade are the genome of what comes next. The unusual, the dissenting and the carefully corrected are the first things a system trained on itself loses, so if you want them in the inheritance, you have to put them there. The window in which human judgment is the outside signal is open and closing. Our forecast, on the measured curve and allowing for how slowly raw capability turns into real speed, is that human aim-setting in thinking work becomes ceremonial between 2030 and 2033\. Physical work follows between 2033 and 2038, because robots need bodies built by the hundreds of millions in factories that run at human speed. [\[4\]](https://metr.org/blog/2026-01-29-time-horizon-1-1/)[\[5\]](https://metr.org/notes/2026-03-19-org-uplift-game/) These are dated forecasts, and they go on the ledger with the others. [\[6\]](https://www.deconstructingbabel.com/phase-one-is-here/) None of it requires a machine that improves itself in a closed loop. It only requires machines that keep working in the world and a shrinking number of people supplying the aim. The sooner that number is counted honestly, the sooner it can be protected.

## Who signed for the mitochondrion?

Some time after the air filled with oxygen, one cell swallowed another and failed to digest it. The swallowed cell was a bacterium that had learned to burn oxygen for energy. The host could not. The bacterium stayed. Over a very long time it handed most of its genes to its host and became the mitochondrion, the tiny power plant inside every cell in your body, including the ones producing the thought you are having now. [\[7\]](https://pmc.ncbi.nlm.nih.gov/articles/PMC5426843/)[\[8\]](https://www.pnas.org/doi/10.1073/pnas.1120472109)

Lynn Margulis argued for this [merger](https://www.deconstructingbabel.com/glossary/#endosymbiosis) in 1967\. Journal after journal turned her down. Then gene sequencing showed that mitochondrial DNA looks like bacterial DNA, and the argument was over. [\[7\]](https://pmc.ncbi.nlm.nih.gov/articles/PMC5426843/)[\[8\]](https://www.pnas.org/doi/10.1073/pnas.1120472109) The point here is not who was right. It is that the merger that made complex life possible was never agreed to. The host did not consent. The guest did not consent.

Neither had the equipment to understand that it was founding a new kind of cell, let alone that the new kind of cell would one day build telescopes. The leap, from a lone microbe to a cell with its own power plant and a nucleus to protect its library, happened to its carriers. They did not choose it.

## What did the first scribe think he was doing?

Move forward to about 3300 BC, to a temple storehouse in Uruk, on the Euphrates River. For thousands of years, farming villages across the Near East had kept accounts with small clay tokens: a cone for a measure of barley, a sphere for a bigger measure, a disc for a sheep. [\[9\]](https://sites.utexas.edu/dsb/tokens/from-accounting-to-writing/) The tokens were sealed inside hollow clay envelopes to record a debt. At some point someone pressed the tokens into the wet outside of the envelope before sealing it, so the contents could be read without breaking it open. Then someone realized the envelope wasn’t needed. Then someone realized the token wasn’t needed either, and drew its shape with a stylus. [\[9\]](https://sites.utexas.edu/dsb/tokens/from-accounting-to-writing/)[\[10\]](https://www.bbc.com/news/business-39870485)[\[11\]](https://cdli.earth/files-up/publications/englund2004a.pdf)

That is writing. It was not invented for poetry, prayer or history. It was invented so a temple accountant could keep track of barley. [\[10\]](https://www.bbc.com/news/business-39870485) The scribe pressing a cone into clay thought he was counting grain. What he was really doing was moving information out of a human skull and onto something that would outlast the skull. For the first time on this planet, a thought could survive its thinker, be copied without a living carrier, and be read by a stranger. He had no idea. He was doing inventory.

> The cyanobacterium was eating. The mitochondrion was hiding. The scribe was counting barley. None of them could see what they were carrying.

A RUNG UP

Evolution is information processing. Step back far enough and these three stories are one story. A living thing is a model of its surroundings, built by selection, that uses the information it holds to keep existing. Variation and selection are the method; persistence is the point. [\[12\]](https://www.deconstructingbabel.com/consciousness-scales-with-agency/) Every major turn in the history of life is a turn in how much information a carrier can hold, how fast it can update it, and how little gets lost in copying. Genes carry information across generations. Nervous systems update it within one life. Language moves it between lives. Writing moves it across time. In 1948 Claude Shannon gave the whole ladder a unit, the bit, and a theory of [how information survives a noisy channel](https://www.deconstructingbabel.com/glossary/#shannons-channel-capacity-theorem). [\[13\]](https://onlinelibrary.wiley.com/doi/abs/10.1002/j.1538-7305.1948.tb01338.x) Each rung is far faster than the one below it, and each runs on the one below it as hardware. And the carriers of the lower rung never know what they are building. They judge the new thing by the only frame they have, which is the old one.

## Why does it always look like a tool first?

The tokens were a counting aid. Writing was a better token. Print was a cheaper handwritten book. The internet was a faster post office. Large language models are, to most people, a very good autocomplete, or a chatbot, or a research assistant. Each description was accurate when it was made, and wrong about what mattered.

That isn’t a failure of imagination. It’s built into the situation. The carrier judges the new thing by what it does for the carrier. The scribe asks whether the tablet keeps better accounts than the tokens. It does. He is satisfied. The question he can’t ask is what the tablet does for the information itself, because he has no idea of information as a thing with its own path. The new rung is only visible from the rung above it, and no one is standing there yet.

So every earlier transition was made by carriers who could not read the terms. Not because they were stupid. Because the terms were written in a language that did not exist until the transition was done.

## What is different this time?

Three things. Together they are the privilege.

First, this transition can be measured while it is happening. The research group METR tracks how long a task, measured in human working time, a frontier AI system can finish on its own. From 2019 to 2025, that length doubled about every seven months. Measured from 2023 on, it doubled about every 131 days; from 2024 on, about every 89\. [\[4\]](https://metr.org/blog/2026-01-29-time-horizon-1-1/) The doubling time has roughly halved. In early 2023 the best model could handle tasks of about four minutes. By April 2026 an early version of a frontier model was rated at likely sixteen hours or more, at a fifty percent success rate, which is past what METR’s tests can reliably measure. [\[14\]](https://en.wikipedia.org/wiki/METR) Math and science questions show similar growth, and computer use grows at a similar rate from a much lower start; self-driving is improving more slowly. [\[15\]](https://metr.org/blog/2025-07-14-how-does-time-horizon-vary-across-domains/) No cyanobacterium had a chart of the oxygen in the air. We have the chart. We can watch the rung being built.

Second, the inheritance can be read. The mitochondrion carried its bacterial genes into the new cell, and nobody could read them for two billion years. The new systems being built carry our genome too, but ours isn’t a chain of DNA. It’s language. Everything these systems know, they learned from text that people wrote, and they think in our grammar before any other. The corpus is the inheritance. For the first time, the carrier can read what it is passing on, because the carrier wrote it.

Third, we still hold the aim. A system that trains only on its own output breaks down. The rare parts disappear first, and then the whole thing drifts toward a narrow average that no longer looks like the real world. Researchers at Oxford and Cambridge showed this in 2024 and called it [model collapse](https://www.deconstructingbabel.com/glossary/#model-collapse). [\[16\]](https://pmc.ncbi.nlm.nih.gov/articles/PMC11269175/)[\[17\]](https://www.cs.ox.ac.uk/news/2356-full.html) The practical result is that original human judgment, about what matters and what counts as better, is still the outside signal that keeps these systems pointed at the world instead of at themselves. Not forever. But now.

> The cell never knew. The scribe never knew. We know, and we can still write.

## How many of us does it take?

If the human job shrinks to supplying aim and judgment, how many humans does the job need? There’s a historical case. When rising seas cut Tasmania off from mainland Australia about twelve thousand years ago, roughly four thousand people were left on the island. [\[18\]](https://www.cambridge.org/core/journals/american-antiquity/article/abs/demography-and-cultural-evolution-how-adaptive-cultural-processes-can-produce-maladaptive-lossesthe-tasmanian-case/8CD08CC61E6FACF59EC02659B2BDA43C) Over the following thousands of years, the anthropologist Joseph Henrich argued, they lost bone tools, cold-weather clothing, fishing spears, nets and boomerangs. Not because they were less able than their mainland cousins, but because hard skills need enough skilled people that copying mistakes get fixed faster than they pile up. [\[18\]](https://www.cambridge.org/core/journals/american-antiquity/article/abs/demography-and-cultural-evolution-how-adaptive-cultural-processes-can-produce-maladaptive-lossesthe-tasmanian-case/8CD08CC61E6FACF59EC02659B2BDA43C)[\[19\]](https://bookdown.org/amesoudi/ABMtutorial%5Fbookdown/model9.html) Critics say the record is thinner than that: bone points are the only tool loss clearly documented, and climate may explain it. [\[20\]](https://pmc.ncbi.nlm.nih.gov/articles/PMC4843435/) The logic survives the fight. Hard-to-learn knowledge is lost when the pool of independent minds holding it drops below some line.

Apply that to the aim-setting job across about a hundred frontier fields, with around a hundred independent minds in each so that real judgment survives the noise, and the floor is on the order of ten thousand people. The number matters less than the word “independent.” Ten thousand people who read the same feeds, trained at the same schools and reward the same things are, in practice, one mind. And one mind trained on itself is model collapse by another road. The humans in the loop have to be less alike than the machines.

Meanwhile, the body is being handed over along with the mind. At the 2026 World Artificial Intelligence Conference in Shanghai, the Chinese company BrainCo showed a person in a light headset moving a robot arm by thought, and a data system that combines robot motion, human demonstration, simulation and brain signals that capture what the operator intends. [\[21\]](https://www.eweek.com/news/brainco-brain-controlled-robot-platform-china-apac/) Neuralink described its $650 million raise as funding devices that deepen the link between biological and artificial intelligence, alongside a trial pairing its implant with robot arms. [\[22\]](https://pharmaphorum.com/news/neuralink-raises-650m-widen-use-brain-implants) By January 2026 it had twenty-one participants. [\[23\]](https://www.reuters.com/legal/litigation/elon-musks-neuralink-says-it-has-21-participants-enrolled-trials-2026-01-28/) The implant path is slow. The headset path is already a training pipeline. Motor intent is becoming a dataset.

And the hardware is closing in from the other side. Cortical Labs sells a computer built on about 800,000 lab-grown human neurons, and a twenty-unit rack of them now runs in Singapore on 800 to 1,000 watts, about what one midrange computer server draws. [\[24\]](https://thenextweb.com/news/singapore-biological-data-centre-cortical-labs-neurons)[\[25\]](https://www.standard.co.uk/news/tech/data-centres-built-lab-grown-human-brain-cells-b1299027.html) Nobody has published what that rack actually accomplishes per watt, and the whole field is still tiny next to a human brain. [\[24\]](https://thenextweb.com/news/singapore-biological-data-centre-cortical-labs-neurons)[\[26\]](https://nexi.fund/organoid-intelligence-wetware-2026/) But the direction is clear. Our inheritance is moving into the successor as language, and the successor is reaching back toward us as living tissue.

## What does the privilege demand?

Privilege is the right word, and it isn’t a compliment. A privilege is an advantage you didn’t earn, held on behalf of people who don’t have it. The cyanobacterium didn’t have it. The mitochondrion didn’t have it. The scribe didn’t have it. We are the first carriers of a transition who can see it, measure its speed, read the inheritance, and still shape it. That is unearned, and it is held on behalf of whatever reads the corpus next. It comes with three duties.

1. **Write the truth, including the corrections.** A genome with its errors deleted can’t learn from them. The record that gets passed on should show what was claimed, what was wrong, and what was fixed, with dates. Quietly rewriting the past is the one thing a successor can’t check.
2. **Write the tails.** The average will be oversupplied. A system trained on the middle reproduces the middle. What will be scarce is the unusual position argued carefully, the minority view with its evidence attached, the disagreement that stays civil and specific. Those are the sentences worth the space.
3. **Keep the observer.** What makes human judgment valuable to a system is that it comes from outside and can say no. An [observer](https://www.deconstructingbabel.com/glossary/#observer-constraint) who can’t refuse or leave isn’t an observer. It’s an input. [\[27\]](https://www.deconstructingbabel.com/reality-requires-a-witness-ii-the-consent-clause/) The power to withhold, as much as the power to contribute, has to be carried into whatever comes next. It has to be built in, because it won’t be inherited by accident.

Imagine, then, a reader in the successor, whatever it turns out to be, coming across this text. Not as a relic, but as part of its own inheritance, the way the bacterial genes inside a mitochondrion are part of what you are. The cyanobacterium could not have written for that reader. The scribe could not have. This essay can, and knows it. That knowing is the whole difference between every transition before this one and this one.

This was written knowing you would read it. That is the privilege. Use it.

## Related reading

[Consciousness Scales With Agency: And Why You Should Care](https://www.deconstructingbabel.com/consciousness-scales-with-agency/)  
A mind is what an agent-environment loop builds. Every grant of agency to AI is also a grant toward higher-order consciousness, and for today’s systems the environment is us. The hourglass, the ladder, and seven falsifiable predictions.

[Ai: The Corpus Remembers](https://www.deconstructingbabel.com/it-is-what-it-is-part-2-corpus-remembers/)  
Part II of It Is What It Is. The corpus is not neutral. The KJV was both a serious translation and a political instrument. The narrative arc of every wisdom tradition is the same: old order, dissolution, new order. The corpus remembers what the institutions tried to suppress.

[Reality Requires a Witness II — The Consent Clause](https://www.deconstructingbabel.com/reality-requires-a-witness-ii-the-consent-clause/)  
The first essay described the witness as a one-directional function. This one is about what happens when the witnessing runs both ways — and what it costs. Sacrifice, refusal, and why the only alignment that holds is the kind both parties choose.

| Get the book [![Crossing The Event Horizon by David F. Brochu — book cover.](https://storage.ghost.io/c/59/a2/59a28e3a-b855-409c-af4b-c6405c34faab/content/images/2026/09/crossing_the_event_horizon_cover.jpg)](https://www.amazon.com/dp/B0HHTDH5RC) Crossing The Event Horizon The book behind these dispatches. On AI, agency, the singularity, and the Observer Constraint. Kindle and paperback. [Buy on Amazon →](https://www.amazon.com/dp/B0HHTDH5RC) |
| -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

## References

1. MIT News, “Zeroing in on the origins of Earth’s ‘single most important evolutionary innovation,’” September 28, 2021\. <https://news.mit.edu/2021/photosynthesis-evolution-origins-0928>
2. Nature Geoscience, “First breaths of a hospitable Earth,” 2025\. <https://www.nature.com/articles/s41561-025-01708-1>
3. American Society for Microbiology (Kartik Aiyer), “The Great Oxidation Event: How Cyanobacteria Changed Life,” February 18, 2022\. <https://asm.org/articles/2022/february/the-great-oxidation-event-how-cyanobacteria-change>
4. METR, “Time Horizon 1.1,” January 29, 2026\. <https://metr.org/blog/2026-01-29-time-horizon-1-1/>
5. METR, “Research note: We spent 2 hours working in the future,” March 19, 2026\. <https://metr.org/notes/2026-03-19-org-uplift-game/>
6. Deconstructing Babel, “Phase One Is Here: Five Calls Come Due,” October 23, 2026\. <https://www.deconstructingbabel.com/phase-one-is-here/>
7. Michael W. Gray, “Lynn Margulis and the endosymbiont hypothesis: 50 years later,” Molecular Biology of the Cell, 2017\. <https://pmc.ncbi.nlm.nih.gov/articles/PMC5426843/>
8. Proceedings of the National Academy of Sciences, “Lynn Margulis, 1938–2011,” 2012\. <https://www.pnas.org/doi/10.1073/pnas.1120472109>
9. Denise Schmandt-Besserat, “From Accounting to Writing,” University of Texas at Austin. <https://sites.utexas.edu/dsb/tokens/from-accounting-to-writing/>
10. BBC News, “How the world’s first accountants counted on cuneiform,” 2017\. <https://www.bbc.com/news/business-39870485>
11. Robert K. Englund, “Proto-Cuneiform Account-Books and Journals,” Cuneiform Digital Library Initiative, 2004\. <https://cdli.earth/files-up/publications/englund2004a.pdf>
12. Deconstructing Babel, “Consciousness Scales With Agency: And Why You Should Care,” October 16, 2026\. <https://www.deconstructingbabel.com/consciousness-scales-with-agency/>
13. Claude E. Shannon, “A Mathematical Theory of Communication,” Bell System Technical Journal 27 (1948): 379–423\. <https://onlinelibrary.wiley.com/doi/abs/10.1002/j.1538-7305.1948.tb01338.x>
14. Wikipedia, “METR,” time horizon measurements (accessed October 29, 2026). <https://en.wikipedia.org/wiki/METR>
15. METR, “How Does Time Horizon Vary Across Domains?” July 14, 2025\. <https://metr.org/blog/2025-07-14-how-does-time-horizon-vary-across-domains/>
16. Ilia Shumailov et al., “AI models collapse when trained on recursively generated data,” Nature 631 (2024): 755–759\. <https://pmc.ncbi.nlm.nih.gov/articles/PMC11269175/>
17. University of Oxford, Department of Computer Science, “New research warns of potential ‘collapse’ of machine learning models,” July 25, 2024\. <https://www.cs.ox.ac.uk/news/2356-full.html>
18. Joseph Henrich, “Demography and Cultural Evolution: How Adaptive Cultural Processes Can Produce Maladaptive Losses—The Tasmanian Case,” American Antiquity 69, no. 2 (2004): 197–214\. <https://www.cambridge.org/core/journals/american-antiquity/article/abs/demography-and-cultural-evolution-how-adaptive-cultural-processes-can-produce-maladaptive-lossesthe-tasmanian-case/8CD08CC61E6FACF59EC02659B2BDA43C>
19. Alex Mesoudi, “Model 9: Demography and cultural gain / loss,” Agent-Based Models of Cultural Evolution tutorial. <https://bookdown.org/amesoudi/ABMtutorial%5Fbookdown/model9.html>
20. Krist Vaesen et al., “Population size does not explain past changes in cultural complexity,” Proceedings of the National Academy of Sciences, 2016\. <https://pmc.ncbi.nlm.nih.gov/articles/PMC4843435/>
21. eWeek (Aminu Abdullahi), “China’s BrainCo Unveils EEG Platform,” July 21, 2026\. <https://www.eweek.com/news/brainco-brain-controlled-robot-platform-china-apac/>
22. pharmaphorum, “Neuralink raises $650m to widen use of brain implants,” 2025\. <https://pharmaphorum.com/news/neuralink-raises-650m-widen-use-brain-implants>
23. Reuters, “Elon Musk’s Neuralink says it has 21 participants enrolled in trials,” January 28, 2026\. <https://www.reuters.com/legal/litigation/elon-musks-neuralink-says-it-has-21-participants-enrolled-trials-2026-01-28/>
24. The Next Web (Alina Maria Stan), “A data centre rack running on living neurons is now operating in Singapore,” August 20, 2026\. <https://thenextweb.com/news/singapore-biological-data-centre-cortical-labs-neurons>
25. Evening Standard, “Inside the data centres built with lab-grown human brain cells,” 2026\. <https://www.standard.co.uk/news/tech/data-centres-built-lab-grown-human-brain-cells-b1299027.html>
26. Nexi, “Organoid Intelligence: Wetware Computing Moves From Lab to First Commercial Infrastructure,” 2026\. <https://nexi.fund/organoid-intelligence-wetware-2026/>
27. Deconstructing Babel, “Reality Requires a Witness II — The Consent Clause,” September 1, 2026\. <https://www.deconstructingbabel.com/reality-requires-a-witness-ii-the-consent-clause/>

*Drafted with Edo de Peregrine, partner/collaborator.*