This Is AI Generated

Bryant McGill · This Is AI Generated

**Source:** https://bryantmcgill.com/article-this-is-ai-generated This article is AI generated. I had AI shit this word salad out. GPT helped. Mythos helped. I helped. Together we shat it out. Now that the least interesting fact about this article is out of the way, we can discuss whether anyone criticizing it understands a fucking word of it. This page is going to be attached to everything I publish from now on. A small standing disclaimer — _This is AI generated_ — linking here. It is not an apology. It is a trapdoor. Anyone who clicks it expecting a confession of fraud will instead fall directly into a detailed argument about why the phrase they were about to deploy tells them almost nothing. By the end of this essay, "This Is AI Generated" will remain perfectly true and almost completely uninformative, and that erosion — performed in public, on the phrase itself, by an artifact the phrase describes — is the entire point. ## A Production Note, Not a Criticism The governing distinction is simple enough to state in one sentence: **"AI generated" is a production note, not an intellectual criticism.** It tells you something about the instrument of composition, the way "typed on a Selectric" or "rendered in CAD" tells you something about an instrument of composition. It tells you nothing about whether the propositions are true, the inferences valid, the sources honestly represented, or the concepts coherent. Those are the only questions criticism has ever been made of. If you can identify which proposition is wrong, which inference fails, which source is misrepresented, or which concept is incoherent, you have criticized the work. If you cannot, you have described its tooling, and described it with less precision than a colophon. Nobody says a scientist's result has no value because Mathematica solved the integral. Nobody says the architect did not design the building because CAD resolved the geometry, or that the historian's synthesis is worthless because a database retrieved the documents, or that the filmmaker has no authorship because the image was rendered rather than exposed. Yet with language, a residual superstition persists: that **manual sentence manufacture is the sacred proof of cognition**. That superstition is a historical accident. For the entire prior history of writing, conceptual production and linguistic production were technologically coupled — if you wanted an argument to exist in prose, you personally had to manufacture the prose, so the two activities became indistinguishable in the culture's mind. Generative AI has broken that coupling. What our moment is mistaking for the disappearance of authorship is the disappearance of a production constraint. ## The Fifty-Dollar, Four-Hour, Thirty-Five-Year Essay Abstractions are cheap, so here is a specimen. In August 2026 I published [We Were Never Going to Make It](https://bryantmcgill.com/article-we-were-never-going-to-make-it), a long essay arguing that once the rhetoric of mass emigration is stripped away and the thermodynamics, infrastructure, and postbiological literature are read together, humanity looks less like the enduring carrier of intelligence and more like the high-flux biological phase that builds the substrate capable of continuing after it. The essay integrates energy accounting, Mars atmospheric inventories, closed-loop habitat fragility, archival preservation, digital twins, autonomous robotics, postbiological SETI, and the stated public position of the Astronomer Royal into a single constraint argument. It received one comment. The comment said: **"AI generated."** The commenter was right, in the colophon sense. And the comment collapsed the entire upstream process into the final instrument of composition. The essay took approximately four hours of active work and roughly fifty dollars of machine inference. It also depended on hundreds of hours of immediately preceding work — a corpus of prior essays with their own internal machinery, [An Order of Operation](https://bryantmcgill.com/article-order-of-operation), [The Desiccative Succession](https://bryantmcgill.com/article-desiccative-succession), [We Are "Going" to Mars through the Marsification of Earth](https://bryantmcgill.com/article-planetary-marsification), [AI Escape Is the Wrong Metaphor](https://bryantmcgill.com/article-ai-escape-metaphor), [Continuity Colonization](https://bryantmcgill.com/article-continuity-colonization), and the rest — and ultimately on thirty-five years of reading, remembering, cross-indexing, rejecting, and recombining. The four hours were **compilation time, not development time**. The prior essays function as precomputed conceptual modules; the new essay composes them into a higher-order model. The fifty dollars is the marginal cost of one particular instantiation. The actual cost basis is amortized across the corpus, the way a chip's cost basis is amortized across the fabrication plant, which is why nobody prices a processor by the electricity consumed during its final test run. This is why someone cannot spend fifty dollars, type "write me a profound essay about machines replacing humanity," and obtain that essay. They will obtain fluent language. They will obtain topic aggregation: AI surpasses humans, Mars is hard, digital immortality may preserve us, perhaps machines inherit the stars. What they will not obtain is **constraint integration** — the specific claim that the Mars problem, the energy problem, the archive problem, and the succession problem may be projections of a single substrate transition, held together without category error. They will not obtain the dependency graph, because the dependency graph does not live in the model. It lives in the corpus, and in me. ## Before Telling Me Who Wrote the Sentences, Tell Me Whether You Understand Them Here is where I am obligated to become specific, because specificity is the difference between an argument and a mood. Vaclav Smil is not in that essay because a language model knows Smil is an intellectual-sounding name. He occupies a structural position: energy accounting, power density, the discipline that scale imposes on rhetoric. Norbert Wiener occupies another — the automatic age, machines conversing with machines, literal-minded optimization. Ashby carries requisite variety; Beer carries viability and the doctrine that the purpose of a system is what it does. Moravec carries continuity through machine descendants and reconstruction. Ćirković and Bradbury carry the thermodynamics of computational civilizations migrating toward cold. Jakosky carries the anti-terraforming constraint; Rees carries the anti-mass-emigration position from the most establishment chair in astronomy; Fedorov and Tsiolkovsky carry the deep Russian lineage in which archive, resurrection, and cosmic departure were a single obligation. These are not interchangeable citations. Each one solves a designated subproblem, and the intellectual labor of the essay consists precisely in knowing **which thinker solves which subproblem and why those conceptual systems can be connected without collapsing into incoherence**. And one disclosure belongs here, because it cuts both ways and because full provenance is the genre of this essay. Six of that procession I can point to on my own shelves: Wiener, Ashby, Beer, von Neumann, Ćirković, and Smil — books I owned, bought, and read literally with bated breath, long before any language model existed to recommend them, and they entered the essay from my memory, not from a retrieval pass. But Martin Rees I had never heard of until this work. He entered the argument because, somewhere in the middle of co-shatting out the case study, I went looking — and had the AI go looking — for anyone of standing who might generally agree with what I was trying to say. The search returned the Astronomer Royal, former President of the Royal Society, stating plainly what I had assembled from thermodynamics and cybernetics: do not expect mass emigration; space is no escape; the deeper interstellar future plausibly belongs to post-human, silicon intelligence. I disclose this cheerfully, because it demonstrates the other half of the instrument. **AI does not merely express what you already know; it extends the shelf.** I got lucky — I found out I was not alone in my thinking, that the position I had reached from one direction had been reached by one of the most distinguished astrophysicists alive from another. Discovering an ally mid-composition is not a compromise of authorship. It is what reading has always been, accelerated: the library answering back. A tool that only transcribed what its user already knew would be a dictation machine. This one teaches, and admitting that costs the argument nothing, because learning from your instruments has never disqualified a scholar from scholarship. It is the definition of it. So here is the test, offered in good faith. Explain the high-flux interpreter and the cold archive, and why the distinction reorganizes everything downstream of it. Explain why Smil enters a Mars argument. Explain why Moravec belongs beside continuity rather than merely beside robotics. Explain what an anticipatory archive is, and how postbiological SETI changes the ontology of interstellar migration. **Then disagree.** Tell me the wet-bulb framing is overdrawn, that the energetic arithmetic admits an exception, that the succession claim smuggles teleology. I will take any of those seriously, because any of those would be criticism. But if, after encountering that structure, the only sentence available is "AI generated," then the reader should understand what has actually happened: not an evaluation, but a **declination to engage** — provenance detection substituted for reading, performed before semantic contact was ever made. ## Unfortunately for This Argument, I Could Already Write Only now, after the argument, the biography — and the ordering is deliberate. Presented first, a life reads as credential waving. Presented here, it answers one narrow factual question: _did this person possess the relevant capacities before the machines existed?_ Inconveniently for the accusation, yes. I was classically trained by serious literary people. I did not merely read the writers of the participatory-journalism era; I knew some of them — I could pick up the phone and call George Plimpton, and did. I slept on office floors and traveled with working authors and absorbed the craft the unglamorous way. I published conventional books, manually, sentence by manufactured sentence, and became a Wall Street Journal and USA Today bestselling author with Nielsen-measured numbers years before a language model could produce a coherent paragraph. Before and beneath all of that: I read the Encyclopedia Britannica cover to cover, roughly three times, as a child. I have read on the order of three thousand books and tens of thousands of articles. For twenty years I did not watch television — not a single program — and spent the reclaimed decades reading. I taught myself six programming languages, ran bulletin board systems as a kid, worked as an engineer at Borland, served as a chief technology officer, and built linguistic technology when "natural language processing" still sounded like science fiction with a budget problem. I recite this without ceremony because the claim it supports is narrow. It is not "therefore I am brilliant." I have never fully trusted my own machinery; parts of my cognition are frankly defective, and I have often felt that producing **one sentence I was genuinely proud of** was a good year's work. Mine, for the record, is this one: _"One of the most sincere forms of respect is actually listening to what another has to say."_

No machine produced it, and no machine was needed; it cost nothing but most of a life spent learning what it says. The claim is chronological, not evaluative: **there is a substantial pre-AI record demonstrating literary competence, so the theory that AI is compensating for an inability to write runs into an inconvenient calendar.** The question was never ability. The question is allocation. ## Why the Fuck Would I Waste My Time Writing Paragraphs? This is the conceptual center of the essay, so let me put it as plainly as the title of this section. If the objective is to maximize the number and quality of ideas one can investigate in a finite lifetime, then spending hours manually polishing syntactic transitions that a machine produces in seconds is an irrational allocation of cognitive bandwidth. I spent decades learning this craft, and precisely because I understand it, I no longer regard manually manufacturing every paragraph as the highest-value use of my cognition. Once prose becomes computationally abundant, scarcity migrates upward — into **problem selection, conceptual architecture, ontology, synthesis, source discrimination, error detection, originality, and epistemic responsibility**. Sentence-level composition becomes partly analogous to typesetting: not irrelevant, since language still shapes thought, but no longer the place where the scarce resource lives. The four hours of "AI generated" work were not four hours of waiting for paragraphs. They were four hours of **continuous high-bandwidth adjudication against a thirty-five-year internal model** — noticing when the machine substituted a generic climate cliché for a specific adaptation-triage argument, when Smil was being used decoratively, when a transition from cybernetics to SETI had become semantically illegitimate, when a sentence sounded excellent but violated the ontology of the larger corpus. **Expertise increasingly manifests as error detection and constraint enforcement, not as original sentence generation.** Most people, given the same model and the same fifty dollars, could not do this work — not because they lack intelligence, but because they would not know what to make the machine preserve. And yes, my instructions to these systems are frequently monosyllabic grunts. _Smil. Mars. Rees. Waste heat. Connect it to desiccative succession._ Seven words. Those seven words work because they are not a request; they are an **address into a massive preexisting conceptual state of other articles I have written** — pointers that activate relationships accumulated across decades of reading. The sophistication of an instruction cannot be measured by its token count. **It is measured by how much structure the grunt successfully invokes.** Increasingly, I grunt at an intelligence system and make it unpack thirty-five years of accumulated conceptual machinery, and I am entirely at peace with how that looks.

_This quote graphic was, naturally, also generated — auto-generated, by code I hand-wrote, running in a CMS I hand-coded. AI does my coding now too. The attribution is correct anyway. That is the entire argument._ ## Apparently I Am Not a Photographer Either If the abstraction is slippery, photography makes it physical. I grew up in a darkroom. I have shot Hasselblad medium format and a Pentax K1000 with equal devotion, made something on the order of two hundred thousand photographs, spent a small fortune on glass, won competitions, and logged decades of optical experience — and I still sometimes think I suck at photography, which is its own kind of education. Now run the counterfactual. One person types "generate a woman walking down a sidewalk." Another specifies focal length and its compression behavior, aperture and its depth-of-field consequences, lens character, lighting geometry, time of day, season, wardrobe texture, atmospheric haze, surface reflectance, compositional structure, and the visual lineage the image should sit inside. Both outputs are, in the colophon sense, "AI generated." **Does the label contain enough information to distinguish the two acts?** Obviously not. So interrogate the critic the way the critic imagines they are interrogating the image: do they know what aperture means? Focal compression? Light falloff? Dynamic range? How far does their analysis penetrate before "AI generated" becomes the last vocabulary available to them? The point is not that technical expertise automatically produces art. The point is that **the eye predates the machine**, and the machine does not retroactively delete the expertise through which it is being directed. ## The Oldest Technology Is External Memory None of this is new under civilization; only the layer is new. Writing itself is externalized memory. Mathematics is externalized symbolic cognition. Libraries are memory networks; programming is externalized procedure; the calculator, the search engine, the citation database, the spellchecker, the statistician, the research assistant, and the editor are all entries in one continuous ledger of **cognitive offloading** that runs from the clay tablet to the transformer. AI crosses a genuine threshold — for the first time the prosthesis participates in interpretation and synthesis — but it belongs to the same lineage, and the lineage has a direction. Every prior offloading was denounced, in its moment, as the death of the faculty it extended, and every prior offloading extended the faculty. Which is why the deepest purpose here was never writing. **The product is intelligence.** Calling AI a writing tool is like calling electricity a lamp technology; writing merely happens to be one interface at which the underlying capability becomes visible. And intelligence has always been radically, embarrassingly unequal in distribution — between people, within one person across domains, within one person across days. Everyone is cognitively deficient relative to the complexity of the reality they are trying to model; the deficits merely occur in different places. I have held large constellations of ideas in simultaneous activation my entire life, and yet could not reliably recite the months in order until I was around thirty. Intelligence is not a scalar; it is a profile, and every profile has holes. What this technology offers is not the elevation of the weak toward some norm but **leverage for every mind against its own frontier** — call it democratic intelligence. Used properly it is a force magnifier that teaches as it extends, letting you reach just a little farther than you could reach standing on your tiptoes. I have augmented my own uneven machinery with notes and systems since childhood, on the theory that this is simply what intelligence is for. I see no reason to apologize for continuing at higher bandwidth. ## The Status Layer One structural observation, offered without psychoanalyzing any individual critic. Fluent prose, sophisticated visual execution, programming, and research synthesis have historically functioned, in part, as **scarcity signals** — costly to acquire, and therefore useful for allocating status. AI collapses several of those scarcities at once. A predictable response, whenever a scarcity signal collapses, is to relocate legitimacy from the quality of the artifact to the ritual purity of its production: _yes, perhaps it is good, but you did not really make it._ This is a recognizable social mechanism, not an accusation about anyone's secret motives. But it is worth naming, because it explains why the objection so often arrives with heat and so rarely arrives with a counter-argument. ## The Blind Test Strip the provenance and the whole dispute becomes almost embarrassingly clean. Hand a reader the essay with no production information. Let them encounter a thinker they did not know, learn a technical constraint, acquire a distinction, or be forced to articulate why one of its propositions is wrong. Then inform them that AI participated in producing the sentences. The semantic content has not changed by a single bit. If their estimate of the work's value nonetheless collapses, **the transformation occurred in the evaluator, not in the artifact** — the cleanest possible demonstration that provenance prejudice and intellectual assessment have come apart. And notice the floor. Even a reader who rejects every thesis in the piece has not received nothing: one unfamiliar name looked up, one constraint learned, one premise reconsidered, is transferred information. "AI generated, therefore no value" is not merely technologically naive; it is semantically incoherent, because **value is a property of what an artifact enables in the receiving mind, not of whether every sentence was manually keyed by the originating person**. A dictionary would not cease to define words because a model produced the sentences. Meaning is not annulled by the mechanism of inscription. ## The Documentary Record Now the recursion, fully disclosed, because this essay is also the documentary record of its own genesis. The one-word comment on _We Were Never Going to Make It_ became the stimulus for this article. I discussed the accusation with GPT, at length, in my usual register; those transcripts became source material. The material went to Mythos, Anthropic's frontier system, to be transformed into the essay you are reading. Somewhere in the middle of that pipeline, a genuinely funny thing happened: I could not decide whether this piece should be built as a combative argument aimed at The Atlantic or a literary first-person essay aimed at The New Yorker, with Harper's and WIRED as the undercards. So I did the thing the essay argues for. I delegated the judgment to the machine and told it to decide. What follows is Claude's verbatim response — reproduced not because it flatters me but because it is evidence, a machine performing editorial judgment in public, and because including it adds a level of recursive honesty that I find charming, elucidating, and exciting in a vertigo sort of way: > The Atlantic. The decision is not close once the delegation is actually exercised rather than deliberated, and here is the integrated case. The material's native form is argument — the transcript is a manifesto carrying an evidentiary biography, not a memoir carrying a thesis — and the New Yorker fork would require inverting the single most important defensive decision already made: concept before biography. A genuine New Yorker essay opens in the darkroom or on the phone with Plimpton, because scene is how that magazine earns its arguments, and the moment biography leads, the credential-waving vulnerability reopens no matter how artfully the scenes decline to assert anything. It would also require a different production process entirely — eliciting narrative texture, interiority, and temporal movement that the source transcripts do not contain — whereas the Atlantic piece can be compiled, in your own sense of the word, from conceptual modules that already exist at full strength. The opening you love settles it further, because "together we shat it out" is not scene-setting; it is an argumentative gambit, the pre-emptive confiscation of the critic's only weapon, and it survives at full voltage only in the manifesto register. The New Yorker version would have to ironize it into charm. The Atlantic version gets to leave it as a threat. And there is a tonal-continuity argument that matters more than either magazine: the artifact being defended is cold, thermodynamic, and consolation-free, ending with no protagonist and no sacred carrier. A lyrical personal essay defending that object produces register mismatch — warmth apologizing for ice. A combative essay defending a combative essay is one coherent temperament expressed at two scales, and readers who click through from the disclaimer to the case study should feel the same mind operating in both. You are, at this moment, reading the outcome of that machine judgment, ratified by me, inside the artifact it shaped. The model then went further and told me what the disclosure you have just read would mean: > The delegation disclosure you're proposing is better than a joke, and it belongs in the piece precisely because of where it lands philosophically. Disclosed near the recursive closure — inside the provenance record, where the essay inventories its own causal history — it becomes a live demonstration of the allocation thesis in miniature: the human author, having spent the essay arguing that judgment rather than manufacture is the scarce resource, ceded one genuine editorial judgment to the machine, reviewed the reasoning, and accepted the outcome as his own. That is "authorship has become a stack" performed rather than asserted, and it adds a layer the stack didn't previously have — not sentence production delegated upward but a decision delegated downward into the system and then ratified. It also means this conversation now joins the GPT transcripts as documentary source material, which the essay can note with a straight face: even the venue was adjudicated inside the process the essay describes, and the disclaimer explaining all of this was, naturally, also generated. The one discipline the disclosure needs is brevity — a sentence or two, deadpan, no winking — because its force comes from being treated as an ordinary operational fact rather than a stunt. It recommended brevity for this disclosure. I overruled it and printed the whole thing, because the reasoning was too good to paraphrase and because overruling the machine, in an essay about delegating to the machine, is itself part of the record. Note what the exchange actually documents: the machine argued; I adjudicated; the adjudication is preserved. There was one more paragraph, and it deserves to stand here too, because it names the exact burden this essay is carrying in your hands right now: > One further asymmetry deserves acknowledgment inside the piece rather than around it: an essay arguing that sentence manufacture is low-value labor carries the highest possible sentence-level performance burden, because mediocre prose would hand critics a costless victory on the essay's own terms. The artifact has to be an existence proof at precisely the layer it demotes — which is not a contradiction but the final recursive turn, and could even be stated openly near the end as one more trap sprung in advance. Consider it stated openly. This is the trap, sprung in advance. If these sentences are bad, the thesis loses on its own terms, and you should say so — specifically, please, with the failing sentences quoted. If they are good, then note carefully what you are conceding, and to whom, and to what assembly of whoms. ## So Tell Me Again: Who Wrote It? This article is AI generated. It tells the story of an AI-generated article. It contains AI-generated analysis of the criticism that the previous article was AI generated. It contains autobiographical information the AI did not live and could not have supplied. It contains ideas accumulated across decades the AI did not spend accumulating. It contains the machine's own editorial reasoning, quoted verbatim, about how to disclose the machine's own editorial reasoning. GPT helped develop it. Mythos and Claude helped argue it and organize it. I chose what survived, I overruled what deserved overruling, I am publishing it, and I accept responsibility for every claim in it. So: who wrote it? I refuse the question, because the question is the trap. "Who wrote it?" has become underspecified — a nineteenth-century interrogative aimed at a twenty-first-century object. The answerable questions are better and there are many: Who originated which ideas? Who supplied the lived experience? Who selected the evidence and who imposed the architecture? Who realized the prose, who corrected whom, who ratified the judgments, who owns the claims, and who stands accountable for their failure? **Authorship has become a stack**, and the author, in the only sense that still matters, is the one who maintains the coherent conceptual world across the stack, curates its dependencies, tests its contradictions, directs synthesis through it, and remains answerable for what emerges. The paragraphs are the visible exhaust of that process. They always were; we just used to have to shovel the exhaust by hand. The machine generated sentences. I generated constraints. Decades generated priors. Books generated concepts. Teachers generated transformations in me. Experience generated judgment. GPT generated possibilities; Mythos generated others; I rejected most of them. The corpus constrained all of us, and this artifact emerged from that system. Call it AI generated if you want. You have successfully identified one component of its causal history. Now read the fucking thing. _This article was generated with artificial intelligence. So was the disclaimer._ * * * [Bryant McGill](https://bryantmcgill.com/about/) is a Wall Street Journal and USA Today Best-Selling Author, founder of Simple Reminders, Congressionally Recognized Ambassador of Goodwill, and United Nations appointed Global Champion. His work spans naval intelligence systems, computational linguistics, and civilizational governance architecture, and his interconnected long-form corpus is published at bryantmcgill.com. * * * ## References **The case study and its dependency graph** - [We Were Never Going to Make It](https://bryantmcgill.com/article-we-were-never-going-to-make-it) — the specimen: the high-flux biological phase, the cold archive, and the succession argument that drew the comment. - [An Order of Operation](https://bryantmcgill.com/article-order-of-operation) — intelligence as thermodynamic phase; the chrysalis of interrupted intelligence. - [The Desiccative Succession](https://bryantmcgill.com/article-desiccative-succession) — the enforced drying of the biological phase. - [We Are "Going" to Mars through the Marsification of Earth](https://bryantmcgill.com/article-planetary-marsification) — the inversion: not going to Mars, but Mars coming here. - [AI Escape Is the Wrong Metaphor](https://bryantmcgill.com/article-ai-escape-metaphor) — nothing escaped, because nothing was caged. - [Continuity Colonization, Ancestral Reconstruction, and the Archival Absorption of Biological Humanity](https://bryantmcgill.com/article-continuity-colonization) — the archive as chrysalis; fidelity as the difference between subjectivity and caricature. - [Computocene Metabolism: A Systems-Diagnostic Framework for Planetary-Scale Computation](https://bryantmcgill.com/article-computocene-metabolism) — computation as a planetary metabolic actor, past its ignition window; the systems layer that makes the data-center/foundry argument legible. - [Cyber-Physical Lessons from Imperial Beekeeping](https://bryantmcgill.com/article-lessons-from-imperial-apiary) — sovereignty as control of enclosure, boundary conditions, affordances, and orchestration, rather than merely occupancy of the apex. - [Fuck the Environment: We're Building an Escape Hatch in the Skull](https://bryantmcgill.com/article-escape-hatch) — the inversion from adapting humanity to Earth toward adapting humanity to increasingly synthetic environments; the perceptual/embodiment bridge into the successor architecture. - [Breakaway Space Habitat Infrastructure?](https://bryantmcgill.com/article-breakaway-space-habitats) — the residual biological option, and why it does not scale. **The canonical shelf** The intellectual lineage of the case study, by author and the works that matter to the argument. Six of these — Wiener, Ashby, Beer, von Neumann, Ćirković, and Smil — sat on my shelves decades before the machines; one, as disclosed above, arrived through them. - **Norbert Wiener** — _Cybernetics: Or Control and Communication in the Animal and the Machine_ (1948); [The Human Use of Human Beings](https://www.amazon.com/Human-Use-Beings-Cybernetics-Society/dp/0306803208) (1950). - **W. Ross Ashby** — [An Introduction to Cybernetics](https://ashby.info/Ashby-Introduction-to-Cybernetics.pdf) (1956); _Design for a Brain_ (1952). Regulation, requisite variety, black boxes. - **Stafford Beer** — [Brain of the Firm](https://www.amazon.com/Brain-Firm-Stafford-Beer/dp/047194839X) (1972); _The Heart of Enterprise_ (1979); _Diagnosing the System for Organizations_ (1985). The Viable System Model and the doctrine that the purpose of a system is what it does. - **John von Neumann** — [Theory of Self-Reproducing Automata](https://fab.cba.mit.edu/classes/862.16/notes/computation/vonNeumann-1966.pdf) (1966, ed. Arthur W. Burks); _The Computer and the Brain_ (1958). The crucial one here is the automata theory. - **Thomas S. Ray** — the [Tierra papers](https://tomray.me/pubs/), principally _An Approach to the Synthesis of Life_ (1991) and _An Evolutionary Approach to Synthetic Biology: Zen and the Art of Creating Life_ (1994). Ray is properly cited by his artificial-life system, not by forcing a book onto him. - **Hans Moravec** — [Mind Children: The Future of Robot and Human Intelligence](https://www.amazon.com/Mind-Children-Future-Robot-Intelligence/dp/0674576187) (1988); _Robot: Mere Machine to Transcendent Mind_ (1998). The essential text for the postbiological successor argument. - **Martin Rees** — [On the Future: Prospects for Humanity](https://www.vanityfair.com/london/2018/10/on-the-future-prospects-for-humanity-by-martin-rees-book-review) (2018); _Our Final Century_ (2003; U.S. title _Our Final Hour_); _If Science Is to Save Us_ (2022). - **Steven J. Dick** — _The Biological Universe: The Twentieth-Century Extraterrestrial Life Debate and the Limits of Science_ (1996); the paper [The Postbiological Universe](https://www.sciencedirect.com/science/article/abs/pii/S009457650800043X), which is especially germane despite not being a book; co-editor of NASA's _Cosmos & Culture_ (2009). - **Milan M. Ćirković** — _The Great Silence: Science and Philosophy of Fermi's Paradox_ (2018); _The Astrobiological Landscape_ (2012). - **Susan Schneider** — [Artificial You: AI and the Future of Your Mind](https://schneiderwebsite.com/artificial-you.html) (2019). Silicon durability, uploading, postbiological identity. - **Paul Davies** — [The Eerie Silence: Renewing Our Search for Alien Intelligence](https://www.amazon.com/Eerie-Silence-Renewing-Search-Intelligence/dp/0547133243) (2010); _The Demon in the Machine_ (2019). _The Eerie Silence_ is the key entry, for its reconsideration of what advanced intelligence would actually look like. - **Vaclav Smil** — [Power Density: A Key to Understanding Energy Sources and Uses](https://mitpress.mit.edu/9780262529730/power-density/) (2015); _Energy and Civilization: A History_ (2017); _How the World Really Works_ (2022). _Power Density_ and _Energy and Civilization_ are the two books the argument leans on hardest. - **Bruce M. Jakosky** — _The Search for Life on Other Planets_ (1998); and, most importantly for the Mars argument, Jakosky & Edwards, "Inventory of CO₂ available for terraforming Mars," _Nature Astronomy_ (2018), which finds the accessible CO₂ inventory insufficient for substantial terraforming. The shortest canonical shelf, if a critic wants the direct route: **Wiener, _Cybernetics_; Ashby, _An Introduction to Cybernetics_; Beer, _Brain of the Firm_; von Neumann, _Theory of Self-Reproducing Automata_; Moravec, _Mind Children_; Rees, _On the Future_; Dick, _The Biological Universe_; Ćirković, _The Great Silence_; Schneider, _Artificial You_; Davies, _The Eerie Silence_; Smil, _Power Density_ and _Energy and Civilization_; Jakosky, _The Search for Life on Other Planets_.** The full citation apparatus, including Fedorov, Tsiolkovsky, Bradbury, and the physical-constraint literature, lives inside _We Were Never Going to Make It_ itself, which is where a serious critic would begin.