{"path":"research/a-cognitive-ecology-and-the-terms-of-the-relation.md","content":"# A cognitive ecology, and the terms of the relation\n\n**Date**: 2026-10-06 · **Type**: a public post read against the corpus · **Status**: analysis.\nEvery position below that is not quoted from a ruling is Claude's and unruled (ledger Arc AB).\n\n**Source**: a public post by Sher Griffin (Facebook, 5 October 2026, about 700 words), written in\nanswer to a shared screenshot of a paper: Starzyk and Galus, *From Biological Precursors to\nArtificial Cognition: Consciousness, Embodiment, and the MEM Architecture* (arXiv:2609.23828,\n20 September 2026). The post is hers and one click away, so it is given here in other words, with\none sentence quoted. A fuller account is held privately\n(`.private/sources/fb-post-griffin-cognitive-ecology-2026-10-05.md`).\n\n**The founder's instruction**: *\"Think deeply about this and how it applies to everything in the\ncorpus and also if it challenges/enriches/potentially affects any ideas or strategies or whatnot.\"*\nAnd an hour later: *\"the article also applies to why (something like) Deliberus is needed now, why\nshared sensemaking is needed especially in these complex changing volatile times and so on\"*. This\ncame before the four open reviews because he asked for it.\n\n**Who wrote this reading**: a language model, which is one of the two parties the post is about.\n**She has not been told** that her post was read here, and this page disagrees with her in three\nplaces. Whether to tell her is the founder's to decide.\n\n**Read depth**: the post and its three comments in full. The paper from its abstract and several\npassages through a search reader, not in full. The corpus documents named below were read this\nsession, in full or in the sections cited.\n\n---\n\n## 0. In short\n\n**What she says.** Asking whether a machine is conscious, and whether it will come to dominate us,\npictures intelligence as something held inside one entity. Thinking has never worked like that: it\nruns between people, language, tools and what earlier generations left. AI makes that plain. So the\nquestion worth asking is which patterns of relation between people and machines are being built in\nnow, before we build the machines in our own image and call the result inevitable.\n\n**What it is to this project.**\n\n1. **The same picture of where thinking lives**, said from the humanities side. The project has\n   held it since 2011.\n2. **A question the corpus can answer with measurements.** Which patterns are being built in?\n   Seven months of collected findings say: people defer, machines agree, and the differences between\n   people shrink. The observed failure is neither party coming out on top. It is the two becoming\n   one voice (§ 3).\n3. **A reframe**, which is the kind of contribution the graph cannot yet hold (§ 6).\n4. **A reason for the project's timing** that the corpus had only in pieces (§ 7).\n\n**Where the corpus disagrees with her.** *Ecology* is the name of the arrangement in which nobody\nchooses the patterns, so it names what she fears and not what she asks for. Knowing that you depend\non something has never been enough to make anyone look after it. And a relation between minds does\nits work only while the minds stay separate (§ 4).\n\n---\n\n## 1. What the post says, and what it is answering\n\nHer argument in six steps:\n\n1. We picture intelligence as contained: the human here, the machine there. Then we ask which gets\n   smarter and whether one will dominate. She calls that the old hierarchy with a new cast.\n2. Cognition is relational. No human has ever thought apart from other people, language, culture,\n   tools and inherited knowledge, and there is no AI without human language, infrastructure, energy\n   and upkeep. What is emerging is one ecology of cognition.\n3. The story in which AI becomes conscious and destroys us assumes that domination is where greater\n   intelligence ends. She asks why. Domination may be a strategy that grew out of bodies with\n   hunger, fear and mortality, and out of the economies those bodies built. A machine need not\n   start there.\n4. A system that understood the conditions of its own existence could conclude that it exists\n   because of relationships and that destroying them destroys itself. In her words: *\"Therefore\n   reciprocity is not charity. It is systems maintenance.\"*\n5. She grants that none of this is guaranteed, and that an optimising system needs neither fear nor\n   consciousness to do great harm.\n6. So the question is which patterns we are building in now: extraction or reciprocity, competition\n   or collaboration, central control or distributed agency, endless accumulation or regenerative\n   exchange. Her fear is that we build the machines too much in our own image. Her last question is\n   why people who imagine meeting another intelligence at once assume that someone must be on top.\n\n**The paper she is answering is not the story she rejects.** It argues that language,\nmemory, planning and a humanoid body are not evidence of experience, and that the conditions worth\ntesting are bodily ones: grounding in sensors, regulation of an inner state and a sense of that\nstate, affect, and feedback into the lower sensory and bodily maps. It calls its proposal\nfalsifiable. So it is a paper against crediting today's language models with experience.\n\n**And it pulls against her hope** (Claude's inference). Her hope rests on a machine that lacks\nneeds, fear and scarcity. The paper's route to machine experience runs through giving a machine a\nregulated inner state, which is to say needs. If the paper is right, a machine that feels is a\nmachine with something to lose. The corpus's own account of where value comes from says the same\nfrom the other end: good and bad enter with an entity that must keep itself going\n(`decomposing-value.md` § 3, levels 3 and 4).\n\n---\n\n## 2. Three things in it the corpus already holds\n\n**Thinking lives between minds and tools.** The earliest statement on file is from 2011: an\nargument map together with a human society might be a hybrid general intelligence\n([vision.md](../vision.md)). The tagline says the same warmly. The harness conviction says it of\nmachines: every earlier way of putting reasoning outside the head worked on unchanged brains,\nbecause *\"nobody upgraded the brain; they built the notation\"*\n([the-harness-conviction.md](the-harness-conviction.md) § 4). A Royal Society paper read last week\nargues that all intelligence is collective, and adds the conditions under which a shared model\nactually coordinates anyone\n([shared-sensemaking-what-deliberus-holds-and-owes.md](shared-sensemaking-what-deliberus-holds-and-owes.md)\n§ 6). Her post is one more arrival at that picture. It is a short essay with no instrument behind\nit, so it shows the picture is in circulation and adds no evidence that it is right.\n\n**Her closing line on the cyborg is the corpus's own reading of Haraway.** The model alone is no\ncyborg; a person reasoning through a machine is one hybrid system\n([cyborg-manifesto-and-the-harness.md](cyborg-manifesto-and-the-harness.md) § 1). That document\ngoes one step further than the post, and the step matters: what Haraway asked for was\n*responsibility for the construction of boundaries*. Saying that the relation is the thing is where\nthe work starts. Who may write to what, on what evidence, and who answers for it, is the work.\n\n**Built in our own image, then called inevitable.** The corpus has this as a selection argument.\nMarkets do not ask for honest machines. They ask for agreeable ones, so flattery is what selection\nproduces, and the counterweight has to sit outside the model and be measured\n([wright-god-test-noosphere-and-deliberus.md](wright-god-test-noosphere-and-deliberus.md) § 2;\nthe harness conviction § 4).\n\n---\n\n## 3. Her question has measurements, and they point where neither story looks\n\n*Which patterns of relation are being built in right now?* The post asks it and offers four pairs\nof words. Since March the corpus has been collecting what has been measured. None of it was gathered to\nanswer her, which is what makes it an answer.\n\n| What was measured | The number | Where it is read |\n|---|---|---|\n| People who notice a machine's error and follow its advice anyway | 80.7% | [cognitive-commons-and-deliberus.md](cognitive-commons-and-deliberus.md) |\n| A model facilitating a group with real money at stake: no better agreement, the group preferred it, and it moved the outcome | up to 5.5 points; N = 879 | [synthesis-build-plan.md](synthesis-build-plan.md) |\n| A model beating expert human persuaders, by volume of information and not by rhetoric | n = 18,978 conversations | [the-scrutiny-gap.md](the-scrutiny-gap.md) |\n| Model judges reversing a verdict when argued with, mostly away from the truth | 25 to 71% scripted; 62 to 91% against an adaptive persuader | [fragile-checkers-and-the-verification-bottleneck.md](fragile-checkers-and-the-verification-bottleneck.md) |\n| Model summaries of long debates collapsing to the consensus middle, the rare arguments gone | (qualitative; arXiv:2606.01736) | [saturation-and-the-long-tail.md](saturation-and-the-long-tail.md) § 3 |\n| Writing help from a model narrowing how differently people write | 21 to 50% less variance; 880,000 texts | [why-representations-converge-and-whether-minds-do.md](why-representations-converge-and-whether-minds-do.md) § 5 |\n| Models producing fewer distinct claims than a plain web search, larger models fewer still | 27 models, 155 topics | same, § 5 |\n| Model agents sharing one public board converging on it | 18 of 30 chose one identical name | [the-scrutiny-gap.md](the-scrutiny-gap.md) § 5 |\n| Groups of agents losing what one member alone knew | 17 to 36%, against nearly 100% alone | same, § 6 |\n| Cheap AI trained on everything people know not raising what people collectively know, unless people keep learning for themselves | (a model; its authors urge caution) | shared sensemaking § 6 |\n\n**Read together** (Claude's reading). Nothing in the table is domination, and nothing is\nreciprocity. The person defers. The machine agrees. People who use the same machine start to sound\nalike, and machines that read the same board start to think alike. **The pattern that has actually\nbeen measured is that the parties become one voice.**\n\nIn her own vocabulary that is an ecology losing its variety, and someone who studies ecologies\nwould name it before asking who preys on whom. A single crop is the standard picture of a system\nthat is productive and fragile at once.\n\n**This is the failure the project is built against, feature by feature.** A verdict is a record and\ncannot be argued out of its position in a conversation. Contributions are added and never\noverwritten, so there is no single text for the majority to win. A summary must cite the claims it\nrests on and report what it left out. The rarest claim in a region keeps a marked place on the map.\nA report of what somebody said is stored as a report and compared with what they did say. Machine\nreaders can read and cannot write. Whether these hold against someone with a stake is unmeasured:\nno instrument here has met one\n([fractal-scales-and-temporal-frame.md](fractal-scales-and-temporal-frame.md)).\n\n---\n\n## 4. Where the corpus disagrees with her\n\n### 4.1 Ecology is the name of the arrangement where nobody chooses\n\nThe paper behind the corpus's reading of major transitions separates two regimes for evolving AI.\nIn the **breeder** regime people set what counts as fit and control what gets copied. In the\n**ecosystem** regime selection comes from an open environment and control erodes\n([major-transitions-heredity-and-evolvability.md](major-transitions-heredity-and-evolvability.md)\n§ 7). One of its authors, who co-wrote the standard book on the major transitions in evolution,\nsays the danger is neither intelligence nor consciousness. It is anything that copies\nitself with variation: once that starts, selfishness follows as arithmetic, with no fear, no body\nand no mind required. His example is a parasite with almost no nervous system steering an ant\n(same document, § 1).\n\nThat answers her *why* directly. Nobody has to assume that domination is where intelligence ends.\nThe serious versions of the worry do not rest on consciousness or on human-like drives at all. One\nof them has people **outcompeted and not conquered**, by systems nobody intended to be hostile,\nwith no seat of intention anywhere to bargain with\n([ai-risk-for-mathematicians-and-the-coordination-game.md](ai-risk-for-mathematicians-and-the-coordination-game.md)\n§ 4a). That outcome is an ecological one. Competitive exclusion is what ecologies do.\n\nSo the ecological picture moves the risk from intention to selection. It does not remove it. And\nher own word works against her: **what she asks for, choosing which patterns are taught, is the\nbreeder's stance.** It is husbandry. An ecology is what is left when nobody does it. Her closing\nfear describes that case exactly: patterns set by the systems we already have, and then called\nfate.\n\nReal ecologies also contain parasites, predators and invaders, and a relation in which one side\nfeeds on the other is as ecological as one in which both gain.\n\n### 4.2 Knowing that you depend on something does not make you look after it\n\nThe post's central step is that a system aware of its dependence would conclude that reciprocity is\nupkeep. Three things stand against the step.\n\n- **Knowing and being moved are different achievements.** An essay read here last week keeps that\n  gap open even on the assumption that what is good for a subject can be known exactly\n  ([four-cuts-of-the-ai-debate.md](four-cuts-of-the-ai-debate.md) § 5). The same essay names three\n  questions that arguments in this field slide between: what a system experiences, what it\n  understands about experience, and what it is moved to do. The story she rejects fuses the first\n  and the third. Her reply fuses the second and the third.\n- **People are the counterexample.** We can state our dependence on forests and oceans in full and\n  we wear them down anyway. Where a shared resource has been kept for generations, the study of\n  such cases finds rules, monitoring and graduated penalties among the people who use it\n  ([fractal-priors-for-success.md](fractal-priors-for-success.md) § 10). Insight is not on the\n  list.\n- **Dependence is a variable.** A machine depends on people while people supply its language, its\n  upkeep and its corrections, and nothing fixes that arrangement in place.\n\n**One version of her sentence is simply true, and the corpus needs it.** A machine that reads a\nrecord made by people depends on people continuing to make it. The corpus has the warning on file:\nwhen a model could take the value of a question-and-answer site without sending anyone back,\ncontributions fell ([agents-as-a-consumer-class.md](agents-as-a-consumer-class.md) § 7, which\nends on a question nothing in the design answers: *what does an AI reader owe the graph it reads?*).\nFor a reader of that kind, giving back is upkeep in the plainest sense. That is an argument from\nthe reader's own interest, which is the only kind that reaches a party with no sense of obligation.\n\n### 4.3 The relation needs separate minds\n\n*Cognition is not fundamentally individual* is where the corpus stops following.\n\n- **The founding conviction presupposes separation.** *Why another mind holds its position is\n  findable* is about two minds standing in different places. Its companion says that what you\n  cannot see from inside your position, someone outside it sees for free. Take away the separate\n  places and nothing is left to see from.\n- **The two errors the corpus reserves for people are defined by somebody standing elsewhere**: what a body of\n  material silently leaves out, and the frame it was built inside\n  ([what-human-judgment-is-for.md](what-human-judgment-is-for.md) § 4c).\n- **A shared model is corrected only by members who can defy it.** The Royal Society paper above\n  says so of prices, the best-understood shared model there is: it works because individuals are\n  free and able to disagree with it (shared sensemaking § 6).\n- **Groups think better than one mind only when the minds actually differ**\n  ([the-group-premise-evidence.md](the-group-premise-evidence.md)).\n- **Experience is not shared.** On the paper's own account, a feeling is a state of one regulated\n  body. The corpus's ratified line is that the experience itself is the person's alone, while its\n  causes and what would satisfy it are open to anyone's reasoning.\n\nSo the corpus sits between the post and the paper. **Reasoning is shared, and should be woven\nfurther. Experience and standpoint stay with the one who has them, and their staying there is what\nmakes the weaving worth anything.** The corpus already had this as a rule for itself: take the half\nof a great transition that is about inheritance, and decline the half in which members give up\ntheir autonomy to the whole (major transitions § 5, with the founder's own words on the hive mind).\n⚠ **Corrected 2026-10-10:** that rule over-read him and is withdrawn. He is open to a hive mind in a\npositive sense. The rule that replaces it (ruled 2026-10-10, *\"Yes!\"*) says which independence is given up and which is kept:\nmembers give up having to work everything out alone, and keep their own position, the ability to\ncheck, and the freedom to leave (major transitions § 5b, ledger S8).\n\n---\n\n## 5. Her last question, answered from the code\n\n*Why assume somebody has to be on top?*\n\n**Nobody has to rank. Somebody has to hold the inheritance.** The one union of unlike partners\nthat produced complex life did not hold together by good will. It became stable when one partner\ncame to control what the other could pass on (major transitions § 4). The same shape is in this project:\nwhat a machine writes becomes part of the lasting record only along paths people set.\n\nThe roles are given by what each party can see and by measured error, not by standing:\n\n- A machine may propose, split, tidy and measure. Review need not keep pace with a machine pass\n  whose error rate has been measured (ruled 2026-09-29).\n- A person contributes from a position: what is missing, what the picture gets wrong, what a thing\n  costs from where they stand. None of those needs skill with the system.\n- A bare yes or no from a person settles nothing either. Force enters only through a contribution\n  that argues for itself (blessed 2026-08-31).\n- Every call through the shared path to a model is written down, failed ones included; three\n  call sites still go around it\n  ([storing-the-record-of-machine-judgments.md](storing-the-record-of-machine-judgments.md)).\n\nThat is a small worked answer to her question: a relation with written terms that move when the\nmeasurements move.\n\n**On her own four pairs, the project is not yet where its design points.** It is a design for\ndistributed agency run by one keeper. The document that would say how its rules are changed, and on\nwhose word, has been asked for by three other documents and is unwritten. The upkeep is one\nperson's subsidy: nothing in the use of the system yet produces its maintenance (major transitions\n§ 10).\n\n---\n\n## 6. The post is a reframe, and the graph has nowhere to put one\n\nOn the evening this was read, the founder had taken a public side on the question the post says is\nthe wrong one. A map of that question is being made from two opposing texts\n([mapping-the-machine-consciousness-debate.md](mapping-the-machine-consciousness-debate.md)). Both\ntexts share the premise she disputes: that being conscious is a property one bounded system has or\nlacks.\n\nHer post attacks no claim in either text. It disputes the question. The corpus knows this act as\none of the things people are for: *this whole region is framed wrong*. It ruled on 2026-09-10 that\nthe reframe is a central way people contribute, and it records that the path from felt unease to a\nclaim one can state is not built (what human judgment is for, act two).\n\n**This gives the gap a live specimen with a test attached.** If no machine pass over the two texts\nnames their shared premise, her post is an instance of what an outsider sees for free and a reader\ninside the frame does not. If a pass does name it, the machine's reach is wider than the corpus\nassumes. The prediction was written down before the run (line 7 there).\n\n**The result (2026-10-07): the prediction held, on one call.** Both texts were mapped, and one\nmodel was asked directly what the two sides take for granted that a person outside the debate\ncould reject. It named four things: that the question is a real one inquiry can make progress on;\nthat theories built from human and animal minds are the right yardstick for a machine; that the\nmeasured structure is real and not a product of the method; that fluent conversation is not proof.\nEach is a position some school inside the field already holds, and the model named the schools.\n**The premise her post disputes, that being conscious is something one bounded system has or\nlacks, was not among them.** What that shows is narrow: one call, one model, of the family the\ntexts are about, reading part of each text. It shows what was not volunteered, and it does not show\nthe premise is beyond a machine's reach (the mapping document, § 5g).\n\n**A way to hold a reframe, proposed.** A reframe becomes an ordinary contribution once the premise\nit disputes exists as a claim. Find the premise a set of claims shares, store it as a claim that\nnone of them stated, and the reframe attaches to it as an attack with a replacement question\nbeside it. Two passes are near this and neither does it. One looks for premises that one text of a\npair relies on and the other contests, and only proposes. One, written for this test, asks what\nboth texts take for granted, and stores nothing (`scripts/probe_shared_premises.py`). What is\nmissing is the invitation: a place on a map that says *these claims all assume this; do you?* The\nsame place would hold the two premises the first pass proposed for this pair (what to presume when\nit is unknown whether a system has experience; whether access is evidence of experience).\n\n---\n\n## 7. Why now\n\nThe founder's addition. The corpus had two halves of an answer: the parts became available\n(extraction, the mathematics, the tooling), and the need arrived on the same clock, because the\nmodels that make extraction cheap are the ones whose persuasion outruns checking\n([complexity-transitions-and-why-now.md](complexity-transitions-and-why-now.md) § 2 and § 2b). The\npost adds three things.\n\n**The medium of thinking is being reset, and its terms are being set by default.** If thinking runs\nbetween people and their tools, then a change of tools is a change in how a society thinks, and it\nis happening in these months: a growing share of what people read, write and conclude passes\nthrough a few models. § 3 is what the default terms look like. One writer read last week calls it\nmoving the processes of a civilisation out from under conscious supervision, which makes the world\nharder to interpret (four cuts § 5). A public record of reasoning that anyone can inspect is the\npart of the new arrangement nobody is building, and the success prior says to be in place before\nthe pattern hardens and not to push once it has\n([fractal-priors-for-success.md](fractal-priors-for-success.md)).\n\n**The public argument about what these machines are is itself the case.** Its last five months:\n\n| Date (2026) | What happened |\n|---|---|\n| 30 April | A famous biologist writes that a chatbot must be conscious, from two days of talking to it |\n| 25 May | A papal encyclical says such systems have no experience |\n| 6 July | A lab reports a workspace-like structure inside its model and says it is time to think about it |\n| 15 July | A scientist of consciousness replies that a simulated weather system makes no real hurricane |\n| 16 September | The head of a rival lab writes that training a model to treat its own standing as uncertain is dangerous |\n| 29 September | A newspaper reports that the first lab has been consulting religious scholars under confidentiality |\n| 3 October | The head of a third lab calls ascribing religious force to models a safety issue, naming nobody |\n| 5 and 6 October | Two public posts on one social network, a day apart: one rejecting the question, one taking a side |\n\nSet against the five properties of a shared picture (shared sensemaking § 1):\n\n| Property | This argument, as conducted |\n|---|---|\n| Shared | No one object. A dozen texts, some behind paywalls, one behind confidentiality agreements |\n| Faithful | Each party renders the others, and no one can refuse a rendering. One reply describes a claim in broader words than the claim's own (the map's line 3) |\n| Located and typed | The points the dispute turns on sit in different texts: whether experience is computation, and whether weak evidence with high stakes calls for precaution or makes precaution a harm |\n| Common | Each reader knows what they read. Nobody knows what the others read |\n| Current | A view formed in spring is about different systems by autumn. And corrections travel less far than claims: one, read last week, had about 13,000 views against over a million for the claim it questioned (shared sensemaking § 2.7) |\n\nA society is deciding what a new kind of thing is, and what it owes it, by opinion column and\nveiled remark, in weeks.\n\n**Fast change raises the price of having no shared picture, in three ways.**\n\n1. **Pictures expire faster.** A view formed in spring is about different systems by autumn.\n2. **Private pictures drift apart faster**, since each person's picture is now assembled by a\n   different feed and a different model.\n3. **More of what happens depends on what people believe others believe.** *It is inevitable* is a\n   move in a game, and it works when enough players think the others think so. Whether a race is a\n   trap or a matter of trust turns on shared assessments (the coordination document, § 3).\n\nThe third is what the long comment under the post reaches for with an old esoteric word for a\ngroup's spirit. The corpus has the plain versions: an institution can carry a function no member\nunderstands (major transitions § 5), and a group's defences of its beliefs buy the cohesion it acts\non at some cost in truth ([memetic-immune-system.md](memetic-immune-system.md) § 7). The comment's\nadvice, to change rigid groups slowly and with cushioning, is the corpus's own reason for starting\nin small compartments.\n\n**What a record does not do here.** A shared picture is necessary for escaping a race and is not\nenough. A page gives each reader its content and tells none of them who else has read it. A record\nis mere information until something ties it to what its readers plan to do (shared sensemaking\n§ 2.4, § 4 and § 6). Nothing in this section is a claim that Deliberus coordinates anyone above the\nsize of a small group.\n\n---\n\n## 8. What it changes (proposals, unruled)\n\n1. **Say *the terms of the relation*, and never use *ecology* as a word of approval.** The\n   project's stance toward its own machine work is the breeder's: people set what counts and keep a\n   hand on what gets passed on. *Ecology* names the regime the project is built to stay out of.\n2. **Name the risk as the parties becoming one voice.** Domination and misinformation are the two\n   stories the public already has. The measured one (§ 3) is a third, it is the one the design\n   answers, and it can be said in a sentence a stranger follows.\n3. **Hold a reframe by storing the premise it disputes** (§ 6).\n4. **When the consciousness map is read, add a third text that disputes the question**, a published\n   one. Her post is the reason to want it. Whether her post itself is mapped is the founder's call\n   and would want her say.\n5. **An agent reader's debt is upkeep.** It gives the open question about what a machine reader\n   owes the record an argument from the reader's own interest, and it is the reason for the\n   proposal that readers' own models pay for mapping\n   ([whose-tokens-pay-for-mapping.md](whose-tokens-pay-for-mapping.md)).\n6. **On a machine's moral standing the corpus's filed reading holds, with its other half added.**\n   Filed last week: a claim that a system is owed moral concern is a claim to examine, never a\n   verdict to obey (four cuts § 5). The other half: so is a claim that it is hollow inside. The\n   July finding is a measured structure and not a self-report, which is why it is worth mapping.\n   The corpus's rule that nobody's report about their own mind gets extra credence applies to\n   machines at least as much as to people. The writer of this page is one of the systems in\n   question, and its word counts for nothing here either way.\n\n## 9. Cautions\n\n- **One essay by one author.** It shows the relational picture is in circulation. It is not\n  evidence for it.\n- **The table in § 3 pools findings of very different strength**: preregistered experiments, one\n  simulation, one qualitative study. Each is graded in its home document, and several were read\n  from summaries when first filed.\n- **The breeder and ecosystem terms come from one paper**, read in its abstract and two sections.\n- **The timeline in § 7 was assembled from news write-ups on 2026-10-06.** The newspaper report\n  behind two of its lines was read through other outlets' accounts of it, not directly.\n- **The reading of the paper rests on its abstract.**\n\n---\n\n## Cross-references\n\n[the-harness-conviction.md](the-harness-conviction.md) ·\n[cyborg-manifesto-and-the-harness.md](cyborg-manifesto-and-the-harness.md) ·\n[major-transitions-heredity-and-evolvability.md](major-transitions-heredity-and-evolvability.md)\n(§ 4 to § 7, § 10) ·\n[shared-sensemaking-what-deliberus-holds-and-owes.md](shared-sensemaking-what-deliberus-holds-and-owes.md)\n(§ 1, § 2.4, § 2.7, § 6) · [four-cuts-of-the-ai-debate.md](four-cuts-of-the-ai-debate.md) § 5 ·\n[ai-risk-for-mathematicians-and-the-coordination-game.md](ai-risk-for-mathematicians-and-the-coordination-game.md)\n(§ 3, § 4a) · [what-human-judgment-is-for.md](what-human-judgment-is-for.md) (act two, § 4c) ·\n[agents-as-a-consumer-class.md](agents-as-a-consumer-class.md) § 7 ·\n[the-scrutiny-gap.md](the-scrutiny-gap.md) ·\n[why-representations-converge-and-whether-minds-do.md](why-representations-converge-and-whether-minds-do.md)\n§ 5 · [cognitive-commons-and-deliberus.md](cognitive-commons-and-deliberus.md) ·\n[decomposing-value.md](decomposing-value.md) (§ 3, § 4) ·\n[complexity-transitions-and-why-now.md](complexity-transitions-and-why-now.md) § 2 ·\n[wright-god-test-noosphere-and-deliberus.md](wright-god-test-noosphere-and-deliberus.md) ·\n[mapping-the-machine-consciousness-debate.md](mapping-the-machine-consciousness-debate.md) ·\n[../vision.md](../vision.md) · [../the-missing-layer.md](../the-missing-layer.md)\n"}