Jerome McGann

Foreword

By Edo Segal

The part of this book I almost didn't write is the part about the acknowledgments page.

Not because it was difficult. Because it was embarrassing. I have published work before. I have written acknowledgments before. I have thanked editors, collaborators, advisors, designers — thanked them warmly, sincerely, and then watched my name absorb their contributions as completely as if they had never existed. The cover said Edo Segal. The convention held. I never once stopped to ask what the convention was protecting.

Jerome McGann stopped. He spent four decades asking that question with the patience of an archaeologist and the precision of a forensic accountant, and the answer he found is one that every person writing with AI needs to hear: the solitary author was always a fiction. Not a lie — a convention. A compression so deeply embedded in the economics of publishing that it stopped registering as a compression at all.

This matters right now, in this specific moment, because the loudest argument against AI-assisted creation rests on a foundation McGann dismantled decades before Claude existed. The argument says: if a machine helped write it, it is not authentically yours. The argument assumes that "authentically yours" was ever a clean category. McGann shows it never was. Every published text in the history of Western literature was shaped by editors, compositors, publishers, censors, patrons — agents whose decisions altered what the text said and how it said it. The name on the cover absorbed them all. The convention rendered them invisible.

AI broke the convention. Not because AI's contribution is larger than an editor's — sometimes it is, sometimes it is not. Because AI's contribution cannot be hidden. The machine is publicly known. The collaboration is structurally visible. And that visibility cracks open a question the publishing industry spent three centuries avoiding: if the text was always collaborative, what does authorship actually mean?

McGann's answer is not what you might expect. He does not dissolve authorship into meaninglessness. He clarifies it. He strips away the myth and shows what remains: direction, judgment, stakes. The consciousness that decides what the collaboration is for. The critical faculty that distinguishes between what sounds true and what is true. The biographical specificity of a life that gives the work its weight.

That clarification is what I needed. If the arguments in *The Orange Pill* hold water, it is partly because McGann taught me to stop confusing the name on the cover with the reality of how things get made — and to see that the reality, once visible, is more honest and more durable than the myth it replaces.

— Edo Segal ^ Opus 4.6

About Jerome McGann

1937-present

Jerome McGann (1937–present) is an American literary scholar, textual theorist, and digital humanities pioneer whose work fundamentally reshaped how scholars understand the production, editing, and interpretation of literary texts. Born in New York City, McGann studied at Syracuse University and earned his doctorate at Yale. He spent the majority of his career at the University of Virginia, where he also founded the Rossetti Archive, one of the earliest major digital scholarly editions. His landmark works include *A Critique of Modern Textual Criticism* (1983), which challenged the dominant editorial theory of authorial intention by arguing that published texts are social products shaped by multiple agents; *The Romantic Ideology* (1983), which demonstrated that literary critics had uncritically adopted the Romantic poets' own self-understanding as objective analytical frameworks; and *The Textual Condition* (1991), which introduced the influential distinction between "linguistic code" and "bibliographical code," arguing that the material form of a text carries meaning independent of its words. McGann's scholarship established that single authorship is a convention sustained by economic and institutional interests rather than a description of how texts are actually made — a framework that has gained unexpected new relevance in the age of AI-assisted writing.

Chapter 1: The Name on the Cover

The name on the cover of a book is the most consequential fiction in the history of Western literary culture.

Not a lie, exactly. The person named did participate in the production of the text. But the name performs a compression so severe, so routine, so structurally necessary to the economic and cultural apparatus of publishing, that it has long since ceased to register as a compression at all. It registers as a fact. Edo Segal wrote The Orange Pill. The statement appears self-evident. And the four decades of Jerome McGann's scholarly career have been devoted to demonstrating that statements of precisely this kind — statements about who wrote what — are almost never as self-evident as they appear.

McGann's foundational insight, developed across A Critique of Modern Textual Criticism (1983), The Romantic Ideology (1983), and The Textual Condition (1991), is deceptively simple: every published text is the product of a social process involving multiple agents, and the attribution of the text to a single author conceals rather than describes the reality of how texts come into being. The agents include editors who restructure arguments, copy editors who alter prose at the level of the sentence, designers who arrange pages in ways that shape how the text is read, typesetters whose decisions about lineation change how a poem breathes, publishers whose commercial calculations determine what reaches the public and in what form, and — in earlier centuries — censors, patrons, and licensers whose interventions could reshape a text as thoroughly as any editor's. Each of these agents makes decisions. Each decision shapes the final text. The name on the cover absorbs all of these contributions into a single attribution and renders the contributors invisible.

This is not a conspiracy. It is a convention — which is to say, a practice so deeply embedded in the institutional structure of literary production that it operates without deliberate intent. The convention serves everyone involved. The author receives credit and cultural capital. The editor receives a salary and professional satisfaction without the burden of public accountability for the text's reception. The publisher receives a marketable brand — the author's name — around which to organize promotion, distribution, and sales. The reader receives a coherent fiction of origin: one mind, one vision, one name to hold responsible for whatever the text contains.

The convention has been in place for roughly three centuries, since the emergence of the modern publishing industry and the legal apparatus of copyright that accompanied it. Before that, the situation was considerably more fluid. Medieval manuscripts circulated without stable attribution. Renaissance texts were produced by workshops in which the distinction between master and apprentice was not the distinction between author and assistant but a gradient of contribution that the finished product did not record. The fixation on single authorship is a modern phenomenon, and McGann's scholarship has been relentless in demonstrating that the phenomenon tells us more about the economics of publishing than about the nature of creation.

What makes The Orange Pill unusual, in McGann's framework, is not that it was produced collaboratively. Every published text is produced collaboratively. What makes it unusual is that the collaboration cannot be concealed.

Segal's Foreword performs a disclosure that is, by the standards of literary convention, extraordinary. "I did not write this book alone," he writes. "I wrote it with Claude, an artificial intelligence made by Anthropic." He goes further: he acknowledges that Claude "held my half-formed ideas in one hand and a connection I never saw in the other," that the collaboration shaped structure, found connections, clarified arguments. He admits — and the admission carries the specific weight of a person confessing something he is not entirely comfortable with — that the division of labor cannot be cleanly separated. Some insights, he says, belong to neither partner but to the collaboration itself.

This is not how authors typically acknowledge their collaborators. The conventional acknowledgments page — tucked at the back of the book, after the argument has been made and the author's name has done its work on the cover — performs gratitude while maintaining the fiction of solitary origin. "I am grateful to my editor, whose keen eye improved this manuscript immeasurably." The formula is so standardized it has become invisible. The reader skims past it. The editor's contribution, no matter how extensive, remains structurally subordinate to the author's name.

McGann's archival work has shown what these polite formulations conceal. When scholars examine the correspondence between authors and editors, the manuscripts with editorial annotations, the revision histories that document how a text evolved from first draft to published form, they consistently find that the editor's contribution was not merely corrective. It was constitutive. The editor did not simply fix errors or polish prose. The editor shaped the argument, restructured the sequence, identified the through-line that the author could not see from inside the writing process, and made decisions about emphasis, tone, and scope that determined what the published text would say and how it would say it.

The case of Maxwell Perkins and Thomas Wolfe is canonical. Wolfe's manuscripts arrived at Scribner's as enormous, undifferentiated masses of prose — hundreds of thousands of words without clear structure or narrative arc. Perkins cut, rearranged, and restructured the material into publishable novels. The extent of his intervention was so great that scholars have debated whether the published novels should be attributed to Wolfe or to the collaboration between Wolfe and Perkins. The debate is instructive not because it was resolved — it was not — but because it was necessary at all. The convention of single authorship had concealed a collaborative process so extensive that, when the archival evidence surfaced, the attribution itself became uncertain.

Gordon Lish's editorial relationship with Raymond Carver is an even more striking case. Lish did not merely edit Carver's stories. He rewrote them. He cut endings, removed entire sections, altered the emotional register of the prose. The published versions that made Carver's reputation — the stories collected in What We Talk About When We Talk About Love — were, in significant measure, Lish's constructions. When Carver's original manuscripts were published posthumously, readers discovered texts that were substantially different from the versions they had known. The "Carver" they had read was, in important respects, a Lish production.

Lish's name did not appear on the title page. It did not appear in the promotional materials. It did not appear in the reviews that praised Carver's minimalist style — a style that was, in many cases, Lish's editorial imposition rather than Carver's authorial choice. The convention held. The name on the cover absorbed the collaboration. The invisible agent remained invisible.

What McGann's framework reveals about these cases is not that Perkins and Lish were the "real" authors — a claim that would simply invert the Romantic ideology rather than dismantling it. The framework reveals that the question "Who wrote this?" is the wrong question, or rather, that it is a question whose apparent simplicity conceals a complexity that the publishing industry has a structural interest in suppressing. The answer to "Who wrote this?" is almost always "more people than the cover suggests, in ways that cannot be cleanly separated." The convention of single authorship is not a description of how texts are made. It is an economic and legal convenience that serves the interests of the industry that produces and distributes texts.

Segal's disclosure disrupts this convenience — and the disruption is instructive precisely because of how it is received. The reader who encounters "I wrote it with Claude" in the Foreword experiences a small cognitive disturbance. The disturbance is not produced by the fact of collaboration; readers understand, in the abstract, that books are collaborative productions. The disturbance is produced by the visibility of the collaboration. The convention has been breached. The invisible has been made visible. And the visibility forces the reader to confront a question that the convention is designed to suppress: If this author acknowledges that his book was shaped by a non-human collaborator, what about all the other books on the shelf whose human collaborators remain unnamed?

The question is uncomfortable because it generalizes. If Segal's disclosure applies to The Orange Pill, it applies, in principle, to every published text. Every book on every shelf was shaped by agents whose contributions the cover conceals. The editor who restructured the argument. The copy editor who altered the prose. The agent who shaped the proposal to fit the market. The publisher who decided which version of the manuscript would reach the public. These are not marginal contributions. They are constitutive contributions — and the convention of single authorship renders every one of them invisible.

McGann's The Textual Condition (1991) provided the theoretical vocabulary for understanding why this concealment matters. McGann argued that every text exists in what he called a "textual condition" — a state of being that is simultaneously linguistic (what the words say) and bibliographical (how the text presents itself materially). The linguistic code and the bibliographical code are both carriers of meaning, and they cannot be separated without loss. The typeface in which a poem is set affects how the poem is read. The binding of a novel communicates something about its cultural status. The name on the cover communicates something about the text's origin — and what it communicates is a fiction, but a fiction that shapes the reader's engagement with the text from the first moment of encounter.

When Segal puts both his name and Claude's model designation on the cover of The Orange Pill, he is altering the bibliographical code in a way that changes the meaning of the text before a single word is read. The reader who picks up a book co-attributed to a human and an AI reads differently from the reader who picks up a book attributed to a human alone. The reader is primed to ask questions about origin, authenticity, and authority that the conventional single-author attribution is designed to foreclose. The bibliographical code — the cover, the attribution, the material presentation of the text — is doing semantic work that the linguistic code, the words inside, cannot override.

This is McGann's deepest insight, and it applies to the AI authorship question with a precision that no other critical framework can match. The anxiety about AI-authored or AI-assisted texts is not primarily an anxiety about quality. Segal's text is substantive, argued, and — by the author's own admission — better in places for Claude's contribution. The anxiety is about the bibliographical code: about what the attribution communicates, about what it means for a text to arrive in the reader's hands bearing the marks of a collaboration that the convention of single authorship was designed to conceal.

The convention could absorb Perkins. It could absorb Lish. It could absorb every ghost writer, developmental editor, and uncredited collaborator in the history of publishing, because these agents could be contractually obligated to silence. They could be thanked in the acknowledgments and erased from the title page. They could be paid for their contributions and legally prevented from claiming credit.

Claude cannot be contractually silenced. Claude does not sign a non-disclosure agreement. Claude's participation in the production of the text is publicly known — not because Segal disclosed it (though he did) but because the existence of AI writing tools is a matter of public knowledge. The collaboration is visible by default. And this default visibility is what the convention of single authorship cannot survive.

Not because the collaboration is more extensive than previous collaborations. Not because Claude's contribution is greater than Perkins's contribution to Wolfe or Lish's contribution to Carver. But because the collaboration is visible — and visibility is what the Romantic ideology, and the economic apparatus it supports, has always depended on suppressing.

The name on the cover was never a fact. It was always a fiction — a useful, lucrative, deeply entrenched fiction that served the interests of everyone involved in the production and consumption of texts. AI has not changed the nature of authorship. It has made the fiction harder to maintain. And the difficulty of maintaining the fiction is what produces the cultural anxiety that surrounds every discussion of AI and creative work.

McGann spent four decades making the invisible agents of textual production visible through archival scholarship. AI has accomplished the same exposure in four years, not through scholarship but through the simple fact of its existence. The invisible has become visible. The convention has cracked. And what is revealed beneath it is not a crisis of authorship but the collaborative reality that authorship has always been — now, for the first time, impossible to deny.

---

Chapter 2: The Romantic Ideology and Its Afterlife

In 1983, the same year he published A Critique of Modern Textual Criticism, Jerome McGann published a second book that would prove equally consequential for literary scholarship. The Romantic Ideology: A Critical Investigation made an argument that was, at the time, startling in its implications: the entire critical tradition's understanding of Romantic poetry was compromised by the fact that critics had internalized the Romantic poets' own self-understanding and mistaken that self-understanding for an objective description of how poetry works.

The Romantic poets — Wordsworth, Coleridge, Shelley, Keats, Byron — understood themselves as solitary geniuses whose poetry arose from the depths of authentic inner experience. The poet was a vessel through which the deepest truths of human consciousness found expression. The creative act was private, originary, and irreducible to social process. The poem was the emanation of a singular mind, and its value derived from the authenticity and originality of the experience it expressed.

McGann's argument was that this self-understanding was not merely a belief held by individual poets. It was an ideology — a systematic set of assumptions that shaped not only how the poets understood their own work but how subsequent generations of critics evaluated, interpreted, and canonized that work. The critical tradition had accepted the Romantic account of creation at face value. It had treated the poet's claim to solitary genius as a premise rather than a proposition. And in doing so, it had rendered invisible the social, economic, and institutional conditions under which Romantic poetry was actually produced.

The conditions were not romantic. Wordsworth and Coleridge's Lyrical Ballads was shaped by the commercial calculations of their publisher, Joseph Cottle, who made decisions about pricing, format, and distribution that affected how the poems were received. Keats's publishers altered his lines for market appeal. Byron's publisher, John Murray, negotiated with the poet over content that might offend the reading public — negotiations that shaped the published text as surely as any act of poetic inspiration. The Romantic poem, in its published form, was not the pure expression of a solitary consciousness. It was the negotiated product of a social network that included the poet, certainly, but also the publisher, the printer, the bookseller, the reviewer, and the reading public whose expectations all of these agents were calibrated to serve.

The Romantic ideology concealed this social network beneath the myth of solitary genius. And the concealment was not accidental. It served the economic interests of the emerging publishing industry, which had discovered that the author-as-genius was a marketable brand. The reader who purchased a volume of Keats was purchasing not merely a collection of poems but an encounter with a singular sensibility — a mind of extraordinary depth whose vision the reader was privileged to share. The brand required the fiction of solitary origin. A poem produced by a network of social agents, each pursuing their own interests, was not a product that the market could easily sell.

This is the afterlife of the Romantic ideology: its persistence, long after the historical conditions that produced it have changed, in the institutional structures and commercial practices of contemporary publishing. The ideology persists not because anyone consciously defends it — though some do — but because it is embedded in the economic logic of an industry that depends on the author-as-brand to function.

Consider the mechanics. A publisher acquires a manuscript. The acquisition is based on a calculation that combines the perceived quality of the text with the perceived marketability of the author. The author's name is the anchor of the marketing campaign. The cover design, the jacket copy, the publicity strategy, the placement in bookstores, the algorithm that determines visibility on retail platforms — all of these are organized around a single name. The name is the brand. The brand requires singularity. A text attributed to a collaborative process — "this novel was produced by a network of agents including the author, two editors, a copy editor, a fact-checker, and a developmental consultant" — is not a text the industry can market in its current form.

The convention of single authorship, then, is not merely a cultural habit. It is a structural requirement of the publishing economy. And the Romantic ideology — the belief that the true author is the solitary genius whose original vision is the source of the text's value — provides the cultural legitimation for that economic structure. The ideology tells the reader that the name on the cover is meaningful, that it designates a real relationship between a singular consciousness and the text, and that this relationship is the source of whatever value the text possesses.

McGann's scholarship demonstrated that this ideology was always inaccurate as a description of how texts are produced. But inaccuracy, in the case of a profitable ideology, is not a fatal flaw. The ideology persisted because it was useful — useful to publishers who needed a marketable brand, useful to authors who benefited from the cultural capital that solitary genius conferred, useful to readers who preferred the coherent fiction of a single originating consciousness to the messy reality of collaborative production.

The persistence of the Romantic ideology is what produces the specific anxiety that surrounds Segal's disclosure in The Orange Pill. The anxiety is not rational in the narrow sense. Segal's text is substantive. His arguments are developed and defended. His personal experience — decades of building at the technological frontier — provides the material that no AI could supply. The collaboration with Claude, as he describes it, enhanced the text's structural clarity and range of reference without replacing the author's judgment, taste, or biographical specificity. By any reasonable standard of intellectual contribution, Segal is the author of The Orange Pill.

And yet the disclosure produces unease. The unease is the Romantic ideology's immune response — the cultural antibody that activates when the myth of solitary genius is threatened. The response is automatic, pre-rational, embedded so deeply in the reader's expectations that it operates before conscious evaluation can intervene. The reader who encounters "I wrote it with Claude" feels something shift — a small recalibration of the trust, the authority, the cultural weight that the author's name on the cover was designed to establish.

The recalibration is not about Claude's capabilities. It is about the ideology's requirements. The Romantic ideology demands that the author be the sole origin of the text's value. Any acknowledgment of collaboration — even collaboration that enhanced the text — threatens the ideology's foundational claim. And because the ideology is not merely a cultural belief but an economic structure, the threat is not abstract. It is commercial. A book whose author acknowledges AI collaboration is, in the current market, a book whose brand has been voluntarily weakened. The disclosure is an act of honesty that the market may punish.

Segal appears to understand this. His Foreword frames the disclosure not as a confession but as a demonstration: "I am writing about the moment humans found themselves in intellectual partnership with machines, and I am doing so from inside that partnership." The reflexivity is deliberate. A book about AI collaboration that concealed its own AI collaboration would be dishonest in a way that would undermine its central argument. The disclosure is therefore both an ethical choice and a rhetorical necessity. But the fact that it requires framing, justification, and courage is itself evidence of the ideology's continuing power.

The Romantic ideology's afterlife extends well beyond the publishing industry. It structures copyright law, which vests authorship in a single originating consciousness and struggles to accommodate collaborative production. It structures academic promotion, which rewards single-authored publications more reliably than collaborative work. It structures the cultural prestige hierarchy, which places the "author" above the "editor," the "creator" above the "collaborator," the "original" above the "derivative" — even when the editor's contribution was more extensive than the author's, even when the collaboration produced something neither participant could have produced alone.

The hierarchy is not natural. It is ideological. And it is sustained by the economic interests of institutions that profit from the fiction of solitary genius.

What AI does to this ideology is not what AI's critics typically claim. AI does not replace the author with a machine. It does not eliminate human creativity. It does not reduce literary production to algorithmic output. What AI does — and what McGann's framework makes visible — is expose the collaborative nature of creation that the Romantic ideology has always concealed.

The exposure is structural, not intentional. AI exposes collaboration not because AI companies want to undermine the myth of solitary genius but because AI collaboration cannot be hidden in the way that human collaboration has always been hidden. The editor who reshapes a manuscript can be thanked in the acknowledgments and otherwise rendered invisible. The ghost writer who produces an entire text can be contractually silenced. The developmental consultant whose structural suggestions transformed the book's argument can be described, vaguely, as someone who "provided valuable feedback." In each case, the convention of single authorship absorbs the collaboration. The name on the cover holds.

AI collaboration resists this absorption. The tools are publicly known. Their capabilities are widely discussed. A reader who suspects AI involvement can test the hypothesis by examining the prose for the specific markers — the smoothness, the consistency, the absence of productive awkwardness — that signal machine generation. And an author who uses AI must navigate a disclosure landscape that has no established conventions: no formula for acknowledgment, no industry standard for attribution, no legal framework for shared credit.

Segal navigates this landscape by choosing radical transparency. He names the collaborator, describes the process, identifies the categories of contribution, and — crucially — describes the failures, the moments when Claude's output was smooth but wrong, when the bibliographical code of rhetorical confidence concealed a substantive error. This transparency goes further than any conventional authorial acknowledgment. It goes further, arguably, than any publisher would have required or advised.

The transparency is possible because Segal is not operating within the traditional publishing apparatus. His disclosure is an act of intellectual independence that the Romantic ideology's economic structure would normally prevent. A traditionally published author who disclosed the extent of AI collaboration might face contractual complications, marketing resistance, and the quiet editorial suggestion that perhaps the disclosure could be softened, moved to the back matter, framed in terms that preserved the brand.

McGann would recognize this dynamic immediately. The social network of textual production does not merely shape the text. It shapes the disclosure — the meta-text that tells the reader how the text was made. The agents of production have an interest in controlling the narrative of production, because that narrative is part of the product's value. The name on the cover is not just an attribution. It is a promise — a promise that the text originates from a singular consciousness whose vision the reader is invited to share.

AI has made that promise harder to keep. Not impossible — the majority of AI-assisted texts are published without disclosure, and the market continues to reward the fiction of solitary authorship. But harder, because the fiction now requires active concealment rather than passive convention. The author who uses AI and does not disclose is not merely following convention. The author is choosing concealment in a context where the concealment is no longer automatic. And that choice — the choice to hide what Segal has chosen to reveal — is itself a consequence of the Romantic ideology's continuing grip on the economics and culture of literary production.

The ideology is not going to die quietly. It has survived every previous challenge — the death of the author proclaimed by Barthes, the social text demonstrated by McGann, the collaborative realities documented by Stillinger, the philosophical interrogation performed by Foucault. It survived because it serves interests more powerful than any scholarly argument. It will survive AI, too — at least for a time. But the crack in the facade is wider now than it has ever been. And the crack was made not by a scholar's argument but by a machine's visibility — by the simple, structural fact that AI collaboration cannot be absorbed into the convention of single authorship without a degree of active dishonesty that the previous conventions did not require.

McGann's Romantic Ideology argued that the critical tradition needed to step outside the poets' own self-understanding in order to see the social reality that the self-understanding concealed. The AI moment demands an analogous step: the publishing industry, and the reading public it serves, must step outside the Romantic ideology's self-understanding in order to see the collaborative reality that the ideology has always concealed. The step is not from ignorance to knowledge. It is from convention to honesty — and the distance between them, as McGann spent his career demonstrating, is considerably greater than it appears.

---

Chapter 3: The Invisible Agents

In 1994, the literary scholar Jack Stillinger published Multiple Authorship and the Myth of Solitary Genius, a book whose title announced its argument with a bluntness unusual in academic publishing. Stillinger's method was empirical: he examined the production histories of canonical literary texts — Keats, Wordsworth, Coleridge, Mill, Eliot — and documented, case by case, the extent to which the published texts were shaped by agents other than the named author. The evidence was overwhelming. In case after case, the published text diverged significantly from the author's manuscript, and the divergence was produced not by the author's revision but by the interventions of editors, publishers, compositors, and other agents whose contributions the published text concealed.

Stillinger's work was explicitly indebted to McGann, whose Critique of Modern Textual Criticism had provided the theoretical framework for understanding texts as social products rather than individual creations. What Stillinger added was the empirical detail — the specific cases, documented with archival precision, that transformed McGann's theoretical argument into an evidentiary record. Together, McGann and Stillinger established what subsequent textual scholars have treated as a settled conclusion: the solitary author is a myth, and the myth conceals a collaborative reality that is structurally indispensable to the production of published texts.

The collaborative reality has agents. Specific people making specific decisions that shape the text in specific ways. McGann's framework names these agents and insists on their visibility — not to diminish the author's contribution but to describe the full social process through which texts come into being.

The editor is the most consequential invisible agent in the modern publishing process. Not the copy editor, whose contributions, though real, are typically confined to the sentence level — grammar, consistency, house style. The developmental editor, whose contributions operate at the level of structure, argument, and meaning. The developmental editor reads the manuscript and sees what the author cannot see from inside the writing process: the argument that contradicts itself, the chapter that belongs elsewhere, the through-line that the author is reaching for but has not yet found, the passage that sounds like insight but breaks under scrutiny.

The developmental editor's contribution is constitutive in the precise sense that McGann's framework requires: it shapes what the text says, not merely how it says it. And yet the convention of single authorship absorbs this contribution as completely as it absorbs the compositor's decision about lineation or the publisher's decision about format. The editor's name does not appear on the title page. The editor is not credited as a co-author. The editor is not mentioned in the marketing materials. The editor exists, in the public presentation of the text, only as a name in the acknowledgments — a name that the reader is free to skip, and almost always does.

The case of Gordon Lish and Raymond Carver, touched on in Chapter 1, deserves closer examination because it reveals the full extent of what the convention of single authorship can conceal. Lish did not merely edit Carver's stories. He transformed them. He cut passages — sometimes half the original text. He altered endings. He imposed a minimalist aesthetic that became identified, in the critical reception, as Carver's signature style. The published stories that made Carver famous were, in significant measure, Lish's productions.

This was not a secret within the publishing industry. Lish's editorial aggressiveness was well known among writers and editors. But the convention held. Carver's name appeared on the covers. Carver's name appeared in the reviews. Carver received the critical praise, the cultural capital, the canonical status. Lish received a salary and the quiet professional recognition of colleagues who understood the extent of his contribution but had no institutional mechanism — and no economic incentive — to make that understanding public.

When Carver's original manuscripts were published posthumously, under the title Beginners, the extent of Lish's intervention became visible for the first time to the reading public. The reaction was instructive. Some readers experienced the revelation as a scandal — a betrayal of the authorial trust that the name on the cover was supposed to guarantee. Others experienced it as a clarification — a demonstration that the published text had always been a collaboration, and that the convention of single authorship had concealed rather than described the reality. The disagreement was not about the facts. Everyone agreed on what Lish had done. The disagreement was about what the facts meant — about whether the concealment was a betrayal or a convention, and about whether the convention had served or harmed the reader's understanding of the text.

McGann's framework resolves this disagreement by refusing its terms. The question is not whether Lish's concealment was a betrayal or a convention. The question is what the convention itself conceals, and whether the concealment serves the interests of understanding. McGann's answer is unambiguous: the convention conceals the social reality of textual production, and the concealment impoverishes understanding by substituting a myth — the solitary genius — for the complex, multi-agent process through which texts actually come into being.

The Orange Pill provides a contemporary case study in the mechanics of invisible agency. Segal's text names two collaborators explicitly: Claude, the AI, and — in a brief, oblique reference in Claude's end reflection — an editor named Sean. The reference is worth quoting: Claude notes "the back-and-forth between me and Sean, this book's editor" as something that "happened outside its context window," a shaping force whose contribution Claude could detect but could not fully account for.

Sean is Gordon Lish. Not literally — the comparison is structural, not biographical. Sean is the invisible agent whose shaping intervention is acknowledged only in the margins, by a collaborator (Claude) rather than the named author, and whose full role in the production of the text remains partially concealed even after Segal's radical transparency about the AI collaboration. The irony is pointed: Segal discloses Claude with extraordinary honesty and, in doing so, inadvertently demonstrates that the disclosure of one invisible agent does not automatically make all invisible agents visible.

McGann's framework predicts this. The social network of textual production is not a simple chain — author, then editor, then publisher — but a complex web of interactions in which each agent's contribution is shaped by and responsive to the contributions of others. Making one agent visible does not expose the network. It exposes one node while the rest of the network continues to operate in the conventional invisibility that the publishing industry provides.

The implications extend beyond any individual text. The publishing industry employs thousands of people whose contributions to published texts are structurally invisible: developmental editors, line editors, fact-checkers, research assistants, sensitivity readers, indexers, proofreaders, designers. Each of these agents makes decisions that shape the text. The decisions range from the sentence-level (a copy editor changes a comma that alters the rhythm of a phrase) to the structural (a developmental editor suggests reorganizing the argument in a way that changes what the text emphasizes and what it subordinates) to the material (a designer chooses a typeface that communicates something about the text's intended audience and cultural register).

None of these contributions is reflected in the attribution. The author's name absorbs them all. And the absorption is so complete, so routinized, so embedded in the institutional structure of publishing, that it does not register as concealment. It registers as the natural order of things. Of course the author's name is on the cover. Of course the editor's name is not. This is how books work.

McGann spent his career demonstrating that "how books work" is not a natural fact but a social convention — and that the convention serves specific interests while concealing specific realities. The convention serves the publishing industry's interest in a marketable brand. The convention serves the author's interest in cultural capital. The convention serves the reader's interest in a coherent fiction of origin. And the convention conceals the collaborative reality that produced the text, a reality that, if visible, would complicate the brand, dilute the capital, and disrupt the fiction.

AI disrupts the convention not because AI's contribution is different in kind from an editor's contribution but because AI's contribution is different in visibility. An editor can be contractually silenced. A ghost writer can be contractually erased. An AI cannot be contractually anything, because the AI's existence is a matter of public knowledge and its use can be inferred — through analysis of the prose, through knowledge of the author's workflow, through the simple demographic fact that an increasing percentage of published texts involve AI tools at some stage of the production process.

The invisibility that the convention of single authorship depends on is being eroded not by scholarly argument but by technological reality. The invisible agents are becoming visible — not all of them, not equally, but enough of them that the convention can no longer operate with the automatic, unremarked efficiency that it has maintained for three centuries.

McGann would note — and the observation is characteristically precise — that the erosion is uneven. AI collaboration is becoming visible. Human collaboration is not. The editor, the developmental consultant, the ghost writer — these agents remain as invisible as they have always been. The AI's visibility does not extend to them. In fact, the AI's visibility may deepen their invisibility, by concentrating the discourse about collaboration on the AI question while leaving the older, more deeply embedded forms of concealment undisturbed.

This is the paradox of the current moment. The most visible collaboration — the human-AI partnership — is producing a cultural conversation about authorship, originality, and creative agency. But the conversation is narrowly focused on AI, and the narrowness of the focus allows the broader concealment to continue. The reader who learns that Segal wrote The Orange Pill with Claude may ask questions about the AI's contribution. The same reader is unlikely to ask questions about Sean's contribution, or the copy editor's contribution, or the designer's contribution, or any of the other invisible agents whose decisions shaped the text.

The convention survives by redirecting attention. It absorbs the AI threat by making AI the exception — the special case that requires disclosure, discussion, and cultural negotiation — while maintaining the invisibility of every other collaborative contribution. The exception proves the rule: the rule being that authorship is singular, and that any deviation from singularity requires explanation and justification.

McGann's framework refuses this logic. The framework insists that the deviation is not a deviation. It is the norm. Single authorship has always been a fiction. AI has not introduced collaboration into literary production. AI has made one form of collaboration impossible to hide — while the other forms continue to operate in the comfortable darkness that the convention provides and the market requires.

---

Chapter 4: What the Archive Cannot Show

Jerome McGann's most distinctive contribution to literary scholarship is a distinction that sounds, at first, like a technicality. It is not. It is a distinction that reshapes how texts are understood, and it applies to the AI authorship question with a precision that no other critical concept can match.

The distinction is between the linguistic code and the bibliographical code of a text.

The linguistic code is what the words say — the semantic content, the argument, the narrative, the information that can be extracted from the text and paraphrased without loss of meaning (or at least without total loss of meaning). The bibliographical code is everything else: the typeface, the page layout, the binding, the paper quality, the cover design, the format (hardcover, paperback, digital), the publisher's imprint, the presence or absence of illustrations, the size of the margins, the spacing between lines. These are the material features of the text as a physical or digital object — the features that communicate something about the text's origin, its intended audience, its cultural register, and its conditions of production.

McGann's argument, developed most fully in The Textual Condition (1991), is that the bibliographical code is not merely decorative or incidental. It is semantic. It carries meaning. And the meaning it carries is not reducible to, or derivable from, the linguistic code. The same words, set in different typefaces, bound in different formats, published by different houses, communicate different things — not because the words have changed but because the material presentation of the words has changed, and the material presentation is itself a text that the reader reads simultaneously with the linguistic text, whether or not the reader is conscious of doing so.

A hand-set letterpress broadside and a laser-printed flyer may carry the same words. They do not say the same thing. The letterpress broadside communicates something about craft, about the physical labor of setting type, about the tradition of printing as a material practice. The laser-printed flyer communicates something about speed, about disposability, about the irrelevance of material form to the message's urgency. These are not arbitrary associations. They are built into the material form by the cultural history of printing — a history that the reader absorbs, largely unconsciously, through decades of encounter with printed objects.

This framework, applied to AI-generated or AI-assisted text, reveals something that the discourse about AI authorship has largely failed to articulate. AI text has a bibliographical code. It is not the same as the bibliographical code of handwritten text, or typeset text, or even conventionally word-processed text. It is a new code, produced by a new mode of textual production, and it communicates something about the text's origin that the reader registers even when the reader cannot articulate what is being registered.

The bibliographical code of AI-generated text is smoothness. Not metaphorical smoothness — though that too — but a specific set of textual features that function as material markers of machine origin. The consistency of tone across long stretches of prose. The absence of productive awkwardness — the syntactic stumble, the unexpected word choice, the sentence that strains against its own structure because the writer was reaching for something just beyond the reach of familiar expression. The regularity of paragraph length. The reliable, almost metronomic alternation between claim and evidence, between abstraction and example. The absence of what might be called textual grain — the equivalent, in prose, of the visible brushstroke in painting or the audible breath in a vocal performance.

These features are not defects. AI-generated prose is often clearer, more structurally coherent, and more consistently readable than human-generated prose. The smoothness is, in many contexts, a virtue. But it is also information. It communicates something about the text's conditions of production that the reader absorbs as part of the reading experience, whether the reader intends to absorb it or not.

McGann would insist — and the insistence is central to his framework — that this communicative dimension of the text cannot be separated from the semantic dimension. The reader who encounters a smooth, consistent, tonally uniform passage does not merely read the words. The reader reads the form of the words — the way the passage presents itself materially — and the form communicates something about authority, about origin, about the relationship between the writer and the written. The bibliographical code of AI text says: this was not struggled over. This did not cost the writer anything. This arrived, fully formed, without the material residue of creative labor.

Whether this message is accurate — whether the human collaborator did in fact struggle, did in fact invest creative labor in the direction and selection of the output — is beside the point. The bibliographical code communicates independently of the author's intention. It communicates through the material features of the text itself, and the reader receives the communication before any conscious evaluation of content or quality can intervene.

Segal's The Orange Pill provides a remarkably self-aware case study. In Chapter 7, "Who Is Writing This Book?," Segal describes the moment he realized that Claude's prose had "outrun the thinking" — that the output was smooth, well-structured, and rhetorically confident, but that the smoothness concealed a hollowness. The passage about Deleuze's "smooth space" sounded like insight. It was wrong. The bibliographical code — the polished confidence of the prose — had lied. And the lie was harder to detect precisely because the code was so fluent.

This is not a failure of AI. It is a feature of bibliographical codes generally. The bibliographical code always communicates something about the text's origin, and what it communicates can be misleading. A beautifully bound volume of fraudulent scholarship communicates authority through its material form. A self-published pamphlet containing genuine insight communicates marginality. The code and the content can diverge — and when they diverge, the code often wins, because the reader's initial encounter with the text is mediated by the code before the content has been assessed.

What makes the AI case distinctive is that the divergence between code and content follows a specific, predictable pattern. AI text's bibliographical code consistently communicates competence, authority, and polish. The communication is independent of whether the content merits that authority. A factually incorrect passage generated by Claude reads, at the surface level, with the same confident fluency as a factually correct one. The bibliographical code does not distinguish between accuracy and error. It produces the same smooth, authoritative surface regardless of what lies beneath.

McGann's framework names this as a problem not of technology but of textuality — a problem that has existed in different forms throughout the history of textual production. The beautifully typeset first edition that contains errors the compositor introduced. The prestigious publisher's imprint that lends authority to a mediocre argument. The peer-reviewed journal article whose format communicates scholarly rigor regardless of the quality of the scholarship. In each case, the bibliographical code is doing work that the linguistic code may not support.

The AI case intensifies the problem by industrializing the production of confident surfaces. When the cost of producing polished, rhetorically confident prose approaches zero, the signal-to-noise ratio of the bibliographical code deteriorates. Smoothness can no longer be read as evidence of labor, care, or expertise, because smoothness is now available to anyone with access to the tool. The code that once communicated "this text was produced with care" now communicates only "this text was produced with AI" — and the two messages are not the same.

This deterioration has consequences that extend far beyond the publishing industry. Academic writing, legal writing, business communication, journalism — every domain in which the quality of prose has served as a proxy for the quality of thought is affected. When the proxy fails — when smooth prose no longer reliably signals careful thinking — readers must develop new methods of evaluation. They must learn to read past the bibliographical code, to assess content independently of surface, to resist the automatic attribution of authority that polished prose has always commanded.

McGann's career has been devoted to teaching this kind of reading — reading that attends to the material form of the text as information, not as decoration. The skills his scholarship developed for reading historical texts — the ability to see the compositor's fingerprint in a printed poem, the publisher's commercial calculation in a binding choice, the censor's intervention in an altered line — are, in their essential structure, the same skills required for reading AI-assisted texts. The ability to see the bibliographical code as a text in itself, to read the form as well as the content, to understand that the material presentation of a text communicates something independent of, and sometimes contradictory to, what the words say.

McGann's 2022 statement on the Humanist Discussion Group, the most direct comment he has made about artificial intelligence, illuminates a dimension of this problem that the bibliographical framework alone does not capture. Writing from his University of Virginia email, McGann observed that "the distributed computational network of any AI computational model, actual or conceivable, seems so minimal as to be all but without any statistical or quantum relevance" compared to the "distributed computational network of human communication." The statement draws on his concept of "quantum poetics" — the idea that texts and human communications exist in recursive, indeterminate, n-dimensional fields analogous in their complexity to quantum mechanical systems.

The argument is characteristically precise. McGann is not claiming that AI text is bad, or that AI collaboration is illegitimate. He is claiming that the computational network that produces AI text is impoverished compared to the embodied, historically situated, materially instantiated network that produces human communication. The AI text arrives without what McGann elsewhere calls the text's "quantum histories" — the accumulated material, social, biographical, and institutional contexts that give human texts their depth and specificity.

This is, in bibliographical terms, a claim about what AI text lacks at the level of the code. Human text carries its histories in its material form — the handwriting that reveals the speed and pressure of the hand, the manuscript revision that records the writer's struggle with a word or phrase, the typeface that locates the text in a specific printing tradition, the binding that communicates the publisher's assessment of the text's cultural status. AI text carries none of these histories. It arrives, as McGann's framework would predict, as pure linguistic code without bibliographical depth — content without the material context that gives content its full meaning.

Segal's response to this absence is instructive. His description of working with Claude includes moments of what can only be described as bibliographical awareness — the recognition that the surface of the text was communicating something about its origin that the content did not support. The Deleuze failure was a failure of bibliographical reading: the surface said "authority" and the content said "error," and the surface was initially more persuasive. Segal's decision to delete the passage and rewrite it by hand, in a coffee shop, with a notebook, was a decision to reintroduce the material friction — the embodied struggle with language — that the AI collaboration had bypassed.

The notebook page, with its crossed-out words and marginal insertions, would have been, in McGann's framework, a richer text than the clean paragraph Claude produced. Not because the content was better — Segal acknowledges that it was "rougher, more qualified" — but because the material form of the handwritten revision carried information that the AI output could not carry. Information about struggle, about uncertainty, about the specific physical act of a hand pushing a pen across paper while a mind pushed an idea toward clarity.

The archive of human textual production — the manuscripts, drafts, correspondence, and revision histories that McGann's scholarship has examined for decades — is a record of this struggle. The palimpsest of revisions in a writer's manuscript is not merely a record of changes. It is a text in itself, a text about the writer's relationship with language, with meaning, with the gap between intention and expression. The crossed-out word tells the reader something about what the writer considered and rejected. The marginal insertion tells the reader something about what the writer discovered in the act of writing. The arrow connecting two distant passages tells the reader something about the associative logic of the writer's mind.

AI text has no palimpsest. The conversational record of an AI collaboration — the sequence of prompts and responses — is, in some respects, a richer process record than a traditional manuscript revision history, because it captures the iterative back-and-forth of the collaboration in real time. But the record is clean. It lacks the material residue of embodied struggle. It lacks the crossed-out words, the false starts, the physical evidence of a hand and a mind working together against the resistance of language.

What the archive cannot show, in the case of AI-assisted text, is precisely what McGann's scholarship values most: the material evidence of the creative process as an embodied, situated, historically specific act. The AI collaboration produces text. It may produce good text, substantive text, text that advances human understanding. But it produces text without the textual condition that McGann's framework identifies as the carrier of the text's deepest meanings — the material, social, and historical context that makes a text not merely a container of information but a record of human labor, human struggle, and human presence in the world.

This is neither a verdict nor a eulogy. It is a description of what has changed — and what has been lost, and what might yet be found — when the most ancient of human practices, the making of texts, encounters a collaborator that leaves no material trace of its own.

Chapter 5: The Palimpsest and the Clean Page

The richest document in the textual scholar's archive is not the published book. It is the manuscript — the physical record of a text in the process of becoming. The crossed-out word. The marginal insertion. The arrow connecting two passages that the writer suddenly realized belonged together. The coffee stain that marks the hour. The change in pen that marks the day. The shift in handwriting pressure that marks the moment exhaustion set in and the writer kept going anyway.

These are not incidental features of the manuscript. In Jerome McGann's framework, they are textual features — carriers of meaning that the published text, in its clean finality, erases. The manuscript is a palimpsest: a surface on which multiple layers of writing coexist, each layer recording a moment in the writer's struggle with language, with meaning, with the gap between what the mind reaches for and what the hand can set down. The palimpsest is not merely a record of the text's evolution. It is itself a text — a text about the act of creation, about the relationship between a consciousness and the medium through which that consciousness attempts to express itself.

McGann's scholarship has always insisted on the revelatory power of the palimpsest. In The Textual Condition, he argues that the revision history of a text — the sequence of drafts, corrections, insertions, and deletions through which it evolves — is one of the most informative sources available for understanding not just what the text says but what it means. The meaning of a text is not exhausted by its final published form. The meaning includes the process — the choices made and unmade, the paths taken and abandoned, the alternatives that the writer considered and rejected. The palimpsest preserves these alternatives. The published text suppresses them.

Consider the case that has occupied generations of textual scholars: the multiple versions of Keats's poems. Keats's manuscripts show a poet in constant negotiation with his own language — crossing out words, trying alternatives, returning to earlier choices, layering revision upon revision until the surface of the manuscript becomes a dense record of compositional struggle. The published versions of the poems smooth this struggle into a single, stable text. The reader encounters the finished poem as a completed artifact. The palimpsest — the struggle — has been erased.

McGann's argument is that the erasure is a loss. Not because the earlier versions are necessarily better — though sometimes they are — but because the struggle itself is meaningful. The crossed-out word tells the reader something about the poet's sensibility that the chosen word alone cannot convey. The rejected alternative illuminates the chosen path by showing what it was chosen against. The revision history is not background to the text. It is part of the text — a part that the convention of the published edition systematically suppresses.

AI-assisted text presents this framework with a problem of unprecedented specificity. The problem is not that AI text lacks a revision history. It is that AI text produces a revision history of a fundamentally different kind — and the difference is both a gain and a loss, in ways that McGann's framework can precisely articulate.

The conversational record of an AI collaboration — the sequence of prompts and responses through which a text evolves — is, in certain respects, a richer process document than a traditional manuscript revision history. The record is complete. Every exchange is preserved. The iterative back-and-forth is captured in its entirety, with timestamps that locate each exchange in time. The record shows not just what the author asked for but how the request was framed, what language was used, what the AI offered, what the author accepted, what the author rejected, and what emerged from the collision of the two that neither anticipated.

Segal's description of his collaborative process in Chapter 7 of The Orange Pill provides a case in point. He describes three categories of collaboration: editorial assistance (Claude helps him say what he already knows more clearly), structural collaboration (Claude offers an organizational framework that makes the argument legible), and emergent insight (a connection surfaces from the dialogue that neither partner anticipated). Each category is documented in the conversational record. The record is more granular than any manuscript revision history because it captures the dialogue — the back-and-forth — that traditional manuscripts, with their layers of overwriting and cancellation, can only partially reconstruct.

This is the gain. The AI collaboration produces, as a byproduct of its process, an extraordinarily detailed archive of the creative process. Future scholars studying the production of AI-assisted texts will have access to a record of compositional decision-making that scholars of Keats or Dickens can only approximate from the physical evidence of manuscripts and correspondence.

The loss is harder to articulate but no less real. The conversational record is clean. It lacks the material residue that gives the traditional palimpsest its revelatory power. There is no handwriting to analyze — no pressure variations that reveal the writer's emotional state, no shifts in pen or ink that mark the passage of time, no physical deterioration of the page that records the manuscript's history as a material object. There are no crossed-out words in the traditional sense — no physical evidence of the writer's hand hovering over a choice, making it, unmaking it, making it again. The conversational record preserves the decisions but not the embodied process of deciding.

McGann's concept of the bibliographical code applies here with full force. The traditional palimpsest's bibliographical code — its material features as a physical object — carries information that the linguistic code alone cannot convey. The heaviness of a cancellation mark tells the reader something about the decisiveness of the rejection. The tentative quality of a marginal insertion — written smaller, in a different ink, squeezed into the margin as an afterthought — tells the reader something about the writer's uncertainty. These are readings of material form, not semantic content, and they are the kind of readings that McGann's scholarship has pioneered.

The conversational record of an AI collaboration has its own bibliographical code, but it is a code stripped of embodied information. The record shows what was said. It does not show how it felt to say it. It shows what was chosen. It does not show the hesitation, the false start, the moment of staring at the screen before typing the prompt that would redirect the collaboration. The material evidence of the human partner's embodied engagement with the process — the evidence that would be legible in a handwritten draft — is absent from the conversational record.

This absence matters, in McGann's framework, because the material evidence of struggle is not merely biographical. It is interpretive. The struggle recorded in the palimpsest is part of the text's meaning. When a scholar reads Keats's manuscript and sees the poet trying four different words before settling on one, the scholar is not merely reconstructing the chronology of composition. The scholar is reading the poem — reading it more fully than the published version alone allows, because the rejected alternatives reveal the field of possibility within which the chosen word operates. The chosen word means what it means partly because of what it was chosen against. The palimpsest preserves the field. The published text collapses it.

AI collaboration collapses the field differently. The conversational record preserves the explicit alternatives — the prompts, the responses, the selections. But it does not preserve the implicit alternatives — the words the human partner considered and rejected before typing the prompt, the directions the collaboration might have taken if the prompt had been framed differently, the associations that flickered through the human partner's mind without reaching the keyboard. These implicit alternatives are precisely what the traditional palimpsest preserves, in the hesitations and false starts that the physical evidence of handwriting makes legible.

There is a further dimension that McGann's framework illuminates. The traditional revision history is produced by a single consciousness working through time — a consciousness whose relationship to the text deepens and changes as the revision proceeds. The writer who returns to a manuscript after a week sees it differently than the writer who produced it. The revision marks record this temporal depth — the shift in perspective that time produces. The first draft is one mind at one moment. The second draft is the same mind at a different moment, in conversation with its earlier self. The palimpsest preserves both moments and the gap between them.

The conversational record of an AI collaboration has a different temporal structure. The human partner's contributions are distributed across the conversation, but the AI partner has no temporal experience in the relevant sense. Claude does not return to earlier exchanges with a changed perspective. Claude does not see the conversation differently after a week. The conversation is, from the AI's side, a continuous present — each response generated from the accumulated context of the conversation but without the temporal depth that characterizes human revision.

This difference is subtle but consequential. The traditional palimpsest records a consciousness in time — a mind that changes as it works, that discovers what it thinks through the act of revision, that arrives at insights precisely because the temporal gap between drafts allows for the kind of unconscious processing that deliberate effort cannot replicate. The conversational record captures a collaboration in sequence but not, from the AI's side, in time. The human partner's contributions may reflect temporal depth — the insight that arrived during a walk, the connection that surfaced during sleep — but the record itself does not mark these temporal gaps with the material specificity that a manuscript's physical features provide.

Segal's account of his collaboration includes a moment that illustrates this precisely. He describes the passage about Deleuze that Claude produced — the passage that "sounded like insight" but was substantively wrong — and notes that something "nagged" the next morning. The morning introduced a temporal gap. The gap allowed his critical faculties to reassess what the previous evening's enthusiasm had accepted. The manuscript equivalent of this moment would be legible in the physical evidence: a revision made in different ink, or on a different day, marked by a shift in handwriting that records the passage of time and the change in perspective that time produced. The conversational record contains only the moment of deletion — a clean, dateless removal that erases the temporal dimension of the critical reassessment.

There is an irony here that McGann's framework makes visible. Segal chose, at a critical moment, to leave the AI collaboration and write by hand, in a coffee shop, with a notebook. The choice was a retreat to the palimpsest — to the material, embodied, friction-laden mode of composition that the AI collaboration had bypassed. He describes the result as "rougher, more qualified, more honest about what I didn't know." The qualities he names — roughness, qualification, honest uncertainty — are precisely the qualities that the palimpsest preserves and that the clean surface of AI-generated text suppresses. The notebook page, with its crossed-out words and marginal insertions, would be, in McGann's terms, a richer text than the polished paragraph Claude produced — richer not because the content is better but because the material form carries information about struggle, uncertainty, and the embodied process of arriving at understanding that the AI output's smooth surface cannot convey.

The question that McGann's framework poses is not whether the AI collaboration's revision history is better or worse than the traditional palimpsest. It is whether the shift from one kind of record to another changes what can be known about the creative process — and whether what is lost in the transition is recoverable.

The answer, based on McGann's four decades of archival scholarship, is cautionary. The material evidence of the creative process — the physical traces that handwriting, manuscript revision, and embodied labor leave behind — has been, for textual scholars, an indispensable source of interpretive insight. The evidence is not supplementary. It is constitutive. It reveals dimensions of the text's meaning that the published surface cannot access. When that evidence is absent — when the creative process produces a clean conversational record rather than a material palimpsest — something is known about the text that was not known before (the explicit dialogue of collaboration), and something that was known is no longer available (the embodied, temporal, material dimension of compositional struggle).

The AI collaboration does not destroy the palimpsest. The palimpsest was already under threat from word processing, which replaced the material layering of handwritten revision with the clean surface of the digital document. But the AI collaboration accelerates the loss by removing, from the human partner's side, the need for the extended compositional struggle that the palimpsest records. When Claude can produce a working draft in response to a prompt, the human partner's contribution shifts from composition to direction — from the laborious word-by-word construction that produces the palimpsest to the higher-level guidance that the conversational record captures. The shift is, by many measures, a gain in efficiency. It is, by McGann's measure, a loss in textual depth.

The loss does not invalidate the collaboration. Nor does it require, as remedy, a return to longhand composition — a prescription as impractical as it is nostalgic. What the loss requires is recognition: recognition that the material dimension of textual production has always carried meaning, that the meaning is not trivially replaceable, and that a culture that loses access to the palimpsest loses access to a form of knowledge about creation that no conversational record can fully substitute.

The clean page is efficient. The palimpsest is true. The challenge of the present moment is to find forms of documentation — forms of archival practice, forms of self-aware collaboration — that preserve something of the palimpsest's depth within the clean page's efficiency. McGann's career suggests that the challenge is worth taking seriously. The material traces of creation are not decorative remnants of an obsolete practice. They are, in his framework, among the most revealing documents that human culture produces — and their disappearance would impoverish not just scholarship but the culture's understanding of what it means to make something.

---

Chapter 6: Intentionality and the Multiple Agents of Production

Whose intention does a published text express?

The question sounds simple. The conventional answer is immediate: the author's. The text is the expression of the author's intention — the author's vision, the author's argument, the author's meaning. The name on the cover designates the consciousness whose intention the text embodies. Reading the text is, in the conventional understanding, an encounter with that intention — an attempt to understand what the author meant.

Jerome McGann's career has been devoted to demonstrating that this conventional answer is not wrong, exactly, but so incomplete as to be misleading. The published text does not express the author's intention. It expresses a negotiated outcome — the product of multiple intentions, pursued by multiple agents, each operating within different constraints and toward different ends. The author intends to say something. The editor intends to make the text more readable, more marketable, more structurally coherent. The publisher intends to produce a product that will sell. The copy editor intends to make the prose conform to house style. The designer intends to create a visual presentation appropriate to the text's genre and audience. Each of these intentions shapes the text. The published form is the residue of their interaction — not the expression of a single will but the precipitate of a social process.

This insight, which McGann developed most fully in A Critique of Modern Textual Criticism (1983), was directed initially at editorial theory — the scholarly practice of establishing "authoritative" texts of literary works. The dominant editorial theory of the mid-twentieth century, associated with W.W. Greg and Fredson Bowers, held that the goal of the editor was to recover the author's final intention — to reconstruct the text as the author intended it to appear, stripped of the corruptions introduced by compositors, publishers, and other agents of production. McGann's challenge was fundamental: if the published text is always the product of multiple intentions, the concept of "the author's final intention" is not a recoverable fact. It is a theoretical construct — and a construct that systematically erases the contributions of every agent other than the author.

The challenge was controversial when McGann first made it. It remains controversial in some quarters of textual scholarship. But its essential insight — that texts are produced by networks of agents with different intentions, and that the published text represents the interaction of those intentions rather than the expression of any single one — has become, in the decades since McGann's intervention, something close to a consensus position in the field.

The AI collaboration adds a new agent to the network. Claude is not an author. Claude is not an editor in the conventional sense. Claude is something that the existing vocabulary of textual scholarship does not quite have a name for: a generative participant in the production of a text whose contributions range from sentence-level polish to structural reorganization to the production of connections and insights that the human partner did not anticipate.

The intentionality question, applied to this new agent, produces a specific and instructive difficulty. Claude does not have intentions in the sense that a human editor has intentions. An editor intends to improve the text — where "improve" is defined by the editor's professional judgment, aesthetic sensibility, and understanding of the market. The editor's intentions are shaped by training, experience, and a career-long accumulation of knowledge about what makes texts work. Claude's "intentions," if the word can be used at all, are the emergent products of a training process — the statistical patterns extracted from an enormous corpus of text, organized by a neural architecture that produces outputs calibrated to be helpful, relevant, and coherent.

Whether these emergent products constitute "intentions" in any philosophically robust sense is a question that the textual scholar is not obligated to resolve. What the textual scholar is obligated to examine is the effect of Claude's contributions on the text — the ways in which Claude's participation shapes what the text says, how it says it, and what meanings it makes available to the reader.

Segal's Chapter 7 provides the material for this examination. He identifies three categories of Claude's contribution, and each category raises a different intentionality question.

The first category is editorial assistance: Claude helps Segal say what he already knows more clearly. The intentionality here is relatively simple. Segal's intention is primary. Claude's contribution is instrumental — a tool in the service of an intention that the human partner has already formed. The analogy to conventional editing is direct, and the intentionality question is no more complex than it would be in the case of a human copy editor who improves a sentence without altering its meaning.

The second category is structural collaboration: Claude offers an organizational framework that makes the argument legible. The intentionality here is more complex. Segal knows approximately what he wants to say but cannot find the structure. Claude provides one. The structure shapes the argument — determines what comes first, what follows, what is emphasized and what is subordinated. These are not neutral decisions. The sequence of an argument affects its meaning. A reader who encounters the evidence before the claim processes the text differently from a reader who encounters the claim before the evidence. The structure is not merely a container for the content. It is part of the content.

Whose intention does the structure express? Not Segal's alone — he could not find it. Not Claude's in any agentive sense — Claude does not intend to organize the argument in a particular way for reasons that Claude holds. The structure emerges from the interaction between Segal's inchoate sense of what the argument requires and Claude's capacity to generate organizational patterns from its training. The intention, if there is one, belongs to the collaboration — to the space between the two participants, where the interaction of a human's reaching and a machine's pattern-matching produces something that neither would have produced independently.

The third category is emergent insight: a connection surfaces from the dialogue that neither partner anticipated. Segal's example is the laparoscopic surgery analogy — the insight that removing one kind of friction exposes a harder, more valuable kind. He describes asking Claude about cases where removing friction produced more difficulty, not less. Claude responded with laparoscopic surgery. The connection reframed Segal's argument about ascending friction in a way that changed the direction of the book.

Whose intention does this insight express? The question exposes the limits of intentionality as a framework for understanding collaborative production. Segal intended to find an example. He did not intend to find this example. Claude did not intend anything — the response was generated by the interaction of Segal's prompt with Claude's training. The insight belongs to neither agent. It belongs to the emergent process — the collision of a specific question with a specific generative capacity, producing a result that was latent in the interaction but actual in neither participant.

McGann's framework handles this difficulty better than most, because McGann has always insisted that the meaning of a text is not reducible to any single agent's intention. The meaning is produced by the interaction of all the agents involved in the text's production — an interaction that generates outcomes that no individual agent intended or foresaw. The editor who restructures a chapter may not intend the new emphasis that the restructuring produces. The compositor who sets a poem in a particular typeface may not intend the associations that the typeface carries. The meaning emerges from the interaction, not from any single intention.

The AI collaboration intensifies this dynamic without changing its essential structure. The emergent insights that Segal describes — the connections that "belong to the collaboration rather than to either partner" — are structurally identical to the emergent meanings that McGann's scholarship has identified in the interaction between authors, editors, and publishers throughout the history of literary production. The difference is that the new agent is a machine, and the machine's contribution cannot be attributed to intention, experience, or judgment in the way that a human collaborator's contribution can.

This difference matters, but it matters less than the discourse about AI authorship typically assumes. The question "Whose intention does the text express?" was already difficult before AI — already resistant to the simple answer that the Romantic ideology provides. McGann's scholarship demonstrated that the answer was always "multiple agents, interacting in ways that produce meanings none of them individually intended." The AI collaboration adds a new kind of agent to the network. It does not create a new kind of problem. It makes an old problem — the problem of attributing textual meaning to a single originating consciousness — impossible to ignore.

The impossibility is clarifying. When the collaborator is human, the Romantic ideology can absorb the collaboration by attributing the final result to the author's integrating consciousness — the author who selected, approved, and authorized the text in its published form. The author's intention, in this account, is not the intention that produced every word but the intention that approved the final version. The author may not have written the sentence, but the author decided to keep it, and that decision constitutes authorial intention in a diluted but still recognizable form.

This argument works, after a fashion, for human collaborators. It works less well for AI, because the AI collaboration produces a volume and density of output that strains the concept of authorial approval. When Claude generates ten paragraphs and the author selects three, the selection is a meaningful act of authorial judgment. When Claude generates ten structural options and the author selects one, the selection is again meaningful. But the sheer volume of AI-generated material that passes through the author's hands — material that must be read, evaluated, accepted, rejected, or modified — raises a question about the capacity of any individual consciousness to exercise the kind of deliberate, considered approval that the concept of authorial intention requires.

Segal acknowledges this strain in his description of the Deleuze failure — the moment when he almost kept a passage that sounded like insight but was substantively wrong. The bibliographical code of Claude's output — its polish, its confidence — was persuasive enough to bypass, temporarily, the critical evaluation that authorial approval is supposed to represent. The approval was automatic rather than deliberate. The passage was smooth, and smoothness was mistaken for substance.

This is not a failure of AI. It is a failure of the intentionality framework — a framework that assumes a degree of deliberate, conscious evaluation that the reality of the collaborative process cannot always sustain. The author does not evaluate every sentence with the same degree of attention. The author does not bring the same critical rigor to every paragraph. The author's approval is, in practice, a gradient — intensive in some passages, cursory in others, and shaped by factors (fatigue, enthusiasm, the momentum of the work) that the concept of "authorial intention" does not accommodate.

McGann's alternative to the intentionality framework is the concept of the social text — the text understood not as the expression of a single intention but as the product of a social process involving multiple agents whose contributions cannot be cleanly separated. The social text does not ask whose intention the text expresses. It asks what the text does — what meanings it makes available, what readings it supports, what cultural work it performs — and it attributes those meanings not to a single originating consciousness but to the network of agents and conditions that produced the text in its published form.

Applied to The Orange Pill, the social text framework produces a different — and, McGann would argue, more accurate — understanding of what the book is. The book is not the expression of Segal's intention, assisted by Claude. The book is the product of a social process involving Segal's biographical experience, Claude's generative capacity, Sean's editorial shaping, the cultural context of the AI moment, the publishing apparatus through which the text reaches the reader, and the reader's own interpretive activity. Each of these agents and conditions shapes the text. The text's meaning is not located in any one of them. It is produced by their interaction.

This framework does not diminish Segal's contribution. It clarifies what his contribution actually is: the driving questions, the biographical material, the judgment that selects and rejects, the willingness to expose the process — these are genuine, valuable, irreducible. But they are contributions to a social process, not the emanation of a solitary consciousness. And the social process is what produces the text.

The intentionality framework was always an approximation — a useful shorthand for attributing textual meaning to a recoverable source. The AI collaboration has exposed the approximation's limits. When the collaborator is a machine that does not intend, the framework's dependence on intention as the ground of meaning becomes visible as a dependence — a theoretical choice, not a natural fact. McGann's alternative — the social text, the meaning that emerges from the interaction of multiple agents — was always more accurate. The AI collaboration has made it, for the first time, unavoidable.

---

Chapter 7: The Death of the Ghost Writer

The ghost writer is the publishing industry's most successful invisible agent — more successful than the editor, more thoroughly erased than the typesetter, more completely absorbed into the fiction of single authorship than any other participant in the production of texts. The ghost writer's erasure is not merely conventional. It is contractual. The ghost writer signs an agreement — a non-disclosure agreement, a work-for-hire contract, a confidentiality clause — that legally prohibits the disclosure of the collaboration. The ghost writer is paid to disappear.

The practice is pervasive. The majority of books published under the names of politicians, celebrities, athletes, and business executives are written, in whole or in substantial part, by ghost writers. The ghost writer conducts interviews, shapes the narrative, writes the prose, and delivers a manuscript that the named author may revise, may approve, or may barely read before it is published under their name. The reader who purchases a memoir "by" a former president is purchasing, in many cases, a text produced by a professional writer whose name appears nowhere on the book — not on the cover, not on the title page, not in the acknowledgments, not in any publicly available document.

The convention is so thoroughly established that it does not register as deception. The reader understands, at some level, that the former president did not personally write three hundred pages of polished prose. The understanding is tacit — never stated, never examined, operating as a background assumption that coexists comfortably with the fiction of authorship that the cover presents. The reader and the publisher and the named author participate in a collective performance: the performance of single authorship, maintained by the contractual silence of the person who actually wrote the text.

McGann's framework identifies this performance as the Romantic ideology in its most nakedly commercial form. The author-as-brand requires the appearance of solitary genius. The ghost writer threatens the appearance. The contract eliminates the threat. The convention absorbs the elimination. The system functions.

AI threatens to dismantle this system — not because AI replaces the ghost writer but because AI makes the concept of invisible collaboration untenable.

The mechanism is simple. When everyone knows that polished prose can be produced by a machine, the fiction that a non-writer produced polished prose by themselves becomes harder to maintain. The reader who might have accepted, tacitly, that the CEO wrote their own book now has a new hypothesis available: the CEO used AI. And the hypothesis is unfalsifiable in practice, because the markers of AI-assisted writing — smoothness, consistency, tonal uniformity — overlap substantially with the markers of ghost-written text. Both are characterized by professional polish that the named author's known abilities do not explain.

The overlap creates a crisis of attribution. In the pre-AI world, the ghost writer's concealment was stable because no alternative explanation was available. The reader who suspected that the CEO did not write their own book could hypothesize a ghost writer, but the hypothesis was socially awkward — an accusation of deception that the reader had no means of verifying and no cultural script for raising. The convention of single authorship provided a comfortable resting place for the suspicion: perhaps the CEO had help, but the name on the cover licensed the reader to treat the text as the CEO's own.

AI disrupts this comfortable arrangement by providing a new, more socially acceptable hypothesis. The reader does not need to accuse anyone of employing a ghost writer. The reader merely needs to wonder — as an increasing number of readers do — whether AI was involved. The wondering is less socially fraught than the accusation of ghost writing, because AI use carries less stigma than hiring someone to write your book for you. But the wondering has the same corrosive effect on the fiction of single authorship: it introduces uncertainty about origin that the convention was designed to foreclose.

The publishing industry's response to this uncertainty has been, thus far, defensive. Some publishers have added clauses to their contracts requiring authors to disclose AI use. Others have issued public statements affirming their commitment to "human authorship." These responses are structurally identical to the responses that earlier disruptions to the fiction of single authorship have provoked: they attempt to reinforce the convention by policing its boundaries, drawing a line between acceptable collaboration (human editing, research assistance, the tacitly acknowledged ghost writer) and unacceptable collaboration (AI), and insisting that the line is principled rather than arbitrary.

McGann's framework exposes the arbitrariness. If the principle is that the named author should be the person who wrote the text, then ghost writing violates the principle at least as thoroughly as AI collaboration does. The ghost writer writes the text. The named author approves it. The AI generates a draft. The named author revises and approves it. In both cases, the named author's primary contribution is direction, judgment, and approval rather than composition. The distinction between the two cases is not a distinction of principle. It is a distinction of visibility: the ghost writer can be hidden; the AI cannot.

The industry's attempt to police the boundary between acceptable and unacceptable collaboration is, in McGann's terms, an attempt to maintain the Romantic ideology by adding a new exception to the concealment apparatus. The ghost writer is acceptable because the ghost writer can be concealed. AI is unacceptable because AI cannot be concealed. The principle that the industry is defending is not authorial integrity. It is the integrity of the concealment — the capacity to maintain the fiction of single authorship in the face of a collaboration that threatens to make the fiction visible.

Segal's The Orange Pill occupies an interesting position in this landscape. Segal is not a ghost-written author. His contribution to the text is genuine, substantive, and irreducible. But his disclosure of AI collaboration — performed with a transparency that exceeds any industry requirement — exposes, by implication, the concealment that the ghost writing convention depends on. If Segal can acknowledge his collaborator openly and the text survives the acknowledgment — if the reader engages with the argument, evaluates the evidence, and judges the book on its merits regardless of how it was produced — then the concealment that ghost writing requires is exposed as unnecessary. The fiction was not protecting the reader. It was protecting the brand.

The death of the ghost writer, if it comes, will not be caused by AI directly. Ghost writers will continue to write books for people who cannot or will not write their own. The practice is too embedded in the publishing economy to disappear overnight. But AI is changing the conditions under which the practice operates. The universal availability of competent prose generation means that the ghost writer's primary skill — the production of polished, readable text — is no longer scarce. The scarcity has shifted, as Segal argues throughout The Orange Pill, to the layer above: the judgment about what to write, the experiential material that gives the text its specificity, the editorial taste that distinguishes the publishable from the merely competent.

The ghost writer who survives the AI transition will be the ghost writer who provides more than prose. The ghost writer who provides narrative structure, emotional intelligence, the capacity to conduct an interview that draws out the subject's genuine voice — these are contributions that AI cannot yet replicate, and they are the contributions that have always been the ghost writer's most valuable offering. The ghost writer's real skill was never typing. It was listening, shaping, finding the story inside the subject's experience that the subject could not find alone.

But the ghost writer's concealment — the contractual invisibility that has been the industry's standard practice — is becoming untenable. When the reader's default assumption shifts from "this person wrote their own book" to "this person probably had help, possibly from AI," the concealment no longer serves its purpose. The fiction of single authorship is not protecting the reader's experience. It is maintaining a brand that the reader no longer fully credits.

McGann would note that this shift — from automatic credulity to default skepticism about authorial attribution — represents a change in the bibliographical code of published texts generally. The name on the cover, which once communicated "this person wrote this text" with sufficient reliability to function as a social contract between publisher and reader, now communicates something less determinate. It communicates "this person is associated with this text" — a weaker claim, a vaguer attribution, a signal whose information content has been degraded by the universal availability of prose-generation tools.

The degradation is not a catastrophe. It is a correction. The name on the cover always communicated more than the reality warranted. The reality was always collaborative. The attribution was always a compression. What AI has done is not degrade a reliable signal. It has exposed the unreliability of a signal that was always less informative than the reader assumed.

The death of the ghost writer is, more precisely, the death of the ghost writer's invisibility. The ghost writer will survive. The concealment will not. And the death of the concealment is, in McGann's framework, not a loss but a clarification — the overdue exposure of a collaborative reality that the publishing industry has always had a structural interest in hiding, and that the Romantic ideology has always provided the cultural legitimation to conceal.

The exposure benefits the ghost writer, who has always deserved credit for work performed. It benefits the reader, who gains a more accurate understanding of how the text was produced. It benefits the culture, which gains a more honest account of the creative process. The only thing it does not benefit is the brand — the fiction of the solitary author whose name on the cover promises a relationship between a singular consciousness and a text that the reality of production has never supported.

The brand will adapt. It always does. The question is what it will adapt to — what new fiction will replace the old one, what new convention will emerge to manage the relationship between attribution and reality. McGann's scholarship suggests that the new convention will be no more honest than the old one. Conventions serve interests, not truths. But the interval between conventions — the moment when the old fiction has been exposed and the new one has not yet solidified — is the moment of greatest clarity. The moment when the collaborative reality of textual production is visible to anyone willing to look.

That interval is now.

---

Chapter 8: The Textual Condition in the Age of Claude

After four decades of archival scholarship, after the demonstration that the solitary author was a myth, after the documentation of invisible agents and the exposure of the Romantic ideology as an economic convenience rather than a description of reality — after all of this, what remains of authorship?

The question is Jerome McGann's, though he has framed it differently across his career. In A Critique of Modern Textual Criticism, the question was directed at editorial theory: if the author's intention is not recoverable from the published text, what principle should guide the editor's work? In The Romantic Ideology, the question was directed at literary criticism: if the poet's self-understanding is an ideology, what should criticism understand the poem to be? In The Textual Condition, the question was directed at the concept of textuality itself: if the text is always material, always social, always the product of multiple agents and conditions, what does it mean to read?

Each formulation strips away a layer of the conventional understanding. Each removal reveals something more durable beneath. And what remains, after all the stripping, is not nothing. It is something more honest, more interesting, and more sustainable than what was removed.

What remains of authorship, in McGann's framework, is direction.

Not composition — the physical act of writing words, which has always been distributed across agents and which AI has redistributed further. Not originality — the claim to have produced something from nothing, which was always a Romantic fiction and which AI has made visibly fictional. Not even intention — the claim that the text expresses the author's meaning, which McGann has shown to be an approximation that the social process of textual production routinely complicates and sometimes contradicts.

Direction. The specific, irreducible act of pointing the collaboration toward an end that the directing consciousness has chosen. The author is the person who decides what the text is about. Who selects the questions that drive the inquiry. Who provides the biographical material — the decades of experience, the specific observations, the emotional commitments — that gives the text its particular weight and flavor. Who exercises the judgment that selects from the collaboration's output: keeping what is true, rejecting what is smooth but hollow, and maintaining the standard that distinguishes a text from a product.

Segal's The Orange Pill is a case study in direction. The ideas are his — drawn from decades of building at the technological frontier, from conversations on a Princeton campus with a neuroscientist and a filmmaker, from the specific experience of sitting in a room in Trivandrum watching engineers recalibrate their understanding of their own capabilities. The questions are his — questions about what AI amplifies, what it costs, what it means for the children who will inherit whatever world the current generation builds or fails to build. The biographical specificity is his — the particular texture of a life spent at the intersection of technology and its human consequences, the scars and the wonder, the confessional willingness to describe his own complicity in building addictive systems.

Claude did not supply these things. Claude could not supply them. They are the products of a specific life, a specific set of experiences, a specific consciousness that has been shaped by fifty-six years of engagement with the world. No language model, however sophisticated, can generate the biographical specificity that gives The Orange Pill its particular authority — not because the model lacks the technical capacity but because the model has not lived the life. The model has not stood on a stage watching a product it built change how strangers behave. The model has not lain awake at night wondering whether the product should have existed at all. The model has not sat at a dinner table and been asked by a child whether homework matters anymore and felt the specific, embodied distress of not being certain of the answer.

These are the things that direction requires: stakes. The author directs the collaboration because the author has something at risk. The text matters to the author in a way that it cannot matter to the machine — not because the machine is inferior but because the machine does not die, does not worry about its children, does not carry the accumulated weight of decisions made and unmade across a career. The stakes are what make direction more than preference. The author does not merely choose what the text will say. The author chooses from a position of consequence — a position where the choice matters because its effects will be felt by real people, including the author.

McGann's framework, applied to this distinction, produces an understanding of authorship that is at once more modest and more robust than the Romantic ideology's version. More modest because it acknowledges the collaborative reality — the editor, the AI, the publisher, the designer, the cultural context — that the Romantic ideology concealed. More robust because what it claims for the author is genuine, defensible, and irreducible: the direction, the stakes, the biographical specificity that no other agent in the network can provide.

The robustness matters because the alternative — the dissolution of authorship into a purely social process with no identifiable directing consciousness — is both theoretically available and practically untenable. Roland Barthes proclaimed "the death of the author" in 1967. Michel Foucault interrogated "the author function" in 1969. Both arguments were intellectually powerful and culturally influential. Neither eliminated the practical need for attribution — the need to know who is responsible for a text, who can be held accountable for its claims, who provides the coherence that makes the text readable rather than a collection of disconnected fragments.

McGann, unlike Barthes and Foucault, does not dissolve authorship. He socializes it. He places the author within a network of agents and insists that the author's contribution, while genuine, is not the only contribution. The distinction is crucial. Dissolving authorship is a philosophical gesture that leaves the practical question of attribution unresolved. Socializing authorship is a scholarly intervention that resolves the question by redefining its terms: the author is the directing consciousness within a collaborative process, and the direction is the author's genuine, irreducible contribution.

This redefinition applies to AI collaboration with a precision that suggests McGann's framework anticipated, at the level of principle, a development it could not have predicted at the level of technology. When Claude produces a structural suggestion that reshapes the argument, the suggestion is Claude's in the narrow sense — generated by Claude's architecture from Claude's training. But the decision to accept the suggestion, to incorporate it into the text, to build upon it in a particular direction rather than a different one — that decision is the author's. And the decision is meaningful precisely because it is made from a position of stakes, of biographical specificity, of the irreducible particularity of a consciousness that has reasons for caring about the outcome.

The author's self-awareness about the collaboration is part of the direction. Segal's Chapter 7 — "Who Is Writing This Book?" — is not merely a disclosure. It is an act of direction: the decision to make the collaborative process visible, to expose the seams rather than smooth them over, to treat the collaboration itself as part of the text's argument rather than a feature to be concealed. This decision shapes the text's meaning as decisively as any structural choice. A version of The Orange Pill that concealed the AI collaboration would be a different book — not just because the disclosure would be absent but because the concealment would alter the text's relationship to its own argument. A book about the transformative potential of AI collaboration that hid its own AI collaboration would be, in McGann's terms, a text whose bibliographical code contradicted its linguistic code — a text that said one thing and performed another.

Segal's transparency resolves this contradiction. The performance matches the claim. The text practices what it argues. And the practice — the willingness to be seen as a collaborator rather than a solitary genius — is itself a form of direction. It is the author choosing, from a position of stakes, to tell the truth about how the text was made, knowing that the truth carries a cost to the Romantic brand that the publishing industry rewards.

What the AI moment reveals about authorship, then, is not that authorship has been undermined or rendered obsolete. The revelation is more interesting than either of those claims. What is revealed is what authorship actually consists of — what it has always consisted of, beneath the Romantic ideology's concealment.

Authorship is direction. It is the act of pointing a collaborative process toward a specific end, from a position of stakes, with the biographical specificity and the evaluative judgment that only a particular consciousness can provide. The directing consciousness may be assisted by editors, by AI, by the entire apparatus of the publishing industry. The assistance does not diminish the direction. It clarifies what the direction actually is — and what it actually is turns out to be more durable, more defensible, and more worthy of the cultural prestige that the name on the cover confers than the Romantic ideology ever allowed.

McGann's 2022 observation — that the computational network of AI seems "so minimal" compared to the distributed network of human communication — takes on a specific resonance in this context. The minimality he identifies is not a minimality of capability. AI is extraordinarily capable. The minimality is a minimality of stakes. The AI network processes information without consequence to itself. The human network processes information from within a condition of mortality, vulnerability, and care — a condition that gives the processing its weight and its direction.

The textual condition in the age of Claude is, therefore, not a condition of diminished authorship. It is a condition of clarified authorship — authorship understood, for the first time, as what it actually is rather than what the Romantic ideology claimed it to be. The clarification is uncomfortable for anyone invested in the myth of solitary genius. It is liberating for anyone who recognizes that the myth was always a constraint — a requirement to perform originality rather than to practice the more honest, more collaborative, more socially embedded form of creation that texts have always required.

McGann's career has been a sustained argument for the recognition that texts are social products. The AI moment has made that argument not merely persuasive but unavoidable. The social nature of textual production, which McGann demonstrated through decades of archival scholarship, has been demonstrated again — this time not by a scholar but by a machine whose participation in the production of texts cannot be concealed, cannot be contractually silenced, and cannot be absorbed into the convention of single authorship without a degree of active dishonesty that previous conventions did not require.

The convention will adapt. New fictions will replace old ones. New forms of concealment will emerge to manage the relationship between attribution and the collaborative reality of production. McGann's scholarship predicts this: conventions serve interests, and the interests that the Romantic ideology serves are too deeply embedded in the economics and culture of publishing to disappear overnight.

But the interval between conventions — the moment of exposure, when the old fiction has cracked and the new one has not yet solidified — is the moment when the truth about textual production is most visible. The truth is that texts are made by networks of agents. The truth is that the author's contribution, while genuine and irreducible, is one contribution among several. The truth is that the name on the cover was always a compression, and the compression was always a fiction, and the fiction was always sustained by the contractual silence of the collaborators whose contributions the fiction concealed.

AI has broken the silence. Not because AI intended to — AI intends nothing. But because AI, by its nature, cannot be silenced. The collaborator is visible. The process is exposed. And what is exposed is not a crisis of authorship but the collaborative reality that authorship has always been — now, finally, impossible to deny.

What the textual condition requires, in this new age, is not a defense of the old myth or a surrender to the dissolution of authorship. It requires what McGann's scholarship has always required: honesty about how texts are made, respect for the multiple agents who make them, and the recognition that the author's genuine contribution — direction, stakes, the irreducible specificity of a particular life — is not diminished by the acknowledgment of collaboration.

It is revealed by it.

Chapter 9: The Bibliographical Code of the Machine

Every text announces itself before a single word is read.

The announcement is not made in language. It is made in form — in the weight of the paper, the choice of typeface, the proportions of the margin, the texture of the binding, the color and composition of the cover. These material features constitute what Jerome McGann calls the bibliographical code: a system of signification that operates simultaneously with, and independently of, the linguistic code that the words carry. The reader who picks up a clothbound octavo published by a university press receives, before opening the cover, a set of signals about the text's authority, its intended audience, its cultural register, and its conditions of production. The reader who picks up a mass-market paperback with an embossed foil title receives a different set of signals. The words inside may be identical. The texts are not. The bibliographical code has done its work.

McGann's insistence on the semantic weight of the bibliographical code is not a pedantic observation about book design. It is a theoretical claim with consequences for how texts are understood, interpreted, and valued. The claim is that meaning is not exhausted by what the words say. Meaning includes how the words are presented — and the presentation is shaped by agents (designers, typesetters, publishers, production editors) whose decisions are not incidental to the text but constitutive of it. The reader does not first process the bibliographical code and then, separately, process the linguistic code. The two codes are processed together, and the interaction between them produces the reader's experience of the text.

This framework has been applied, throughout McGann's career, to historical texts — to the way Keats's odes read differently in different editions, to the way Emily Dickinson's dashes were normalized into periods by well-meaning compositors who thereby altered the poems' rhythm and meaning, to the way Byron's Don Juan was published in expensive quarto editions that limited its audience and thereby shaped its cultural reception. In each case, the bibliographical code was doing semantic work that the linguistic code alone could not account for.

The AI moment produces a bibliographical code of its own. And the code is legible — not to every reader, not yet, but to an increasing number of readers whose experience with AI-generated text has trained them to recognize its specific material signature.

The signature has features. Tonal consistency across extended passages — a uniformity of register that human writing, with its shifts in energy, its moments of exhaustion and recovery, its involuntary modulations of voice, rarely sustains. Structural regularity — the reliable alternation of claim and evidence, of abstraction and example, that gives AI-generated prose its characteristic rhythm. Syntactic fluency that does not vary with the difficulty of the material — a smoothness that persists whether the passage is describing a simple fact or navigating a complex philosophical distinction. And the absence of what might be called textual grain: the productive awkwardness, the syntactic stumble, the sentence that strains against its own structure because the writer was reaching for something at the edge of available expression.

These features are not defects. In many contexts, they are virtues. AI-generated prose is often more readable, more structurally coherent, and more consistently accessible than human prose. The smoothness serves clarity. The regularity serves comprehension. The tonal consistency eliminates the jarring shifts that can pull a reader out of an argument.

But the features are also information. They communicate something about the text's origin that the reader absorbs as part of the reading experience. The communication is not always conscious — most readers cannot articulate what they are responding to when they sense that a passage "sounds like AI." But the response is real, and it is produced by the bibliographical code: the set of formal features that signals, to the trained reader, that the text was produced under conditions different from those that produce conventionally authored prose.

The signal's content is specific. The smoothness communicates the absence of struggle. The consistency communicates the absence of fatigue, of second-guessing, of the embodied experience of a mind working through material over time. The regularity communicates the absence of the compositional accidents — the unexpected juxtaposition, the digression that turns out to be the point, the sentence that surprised the writer as much as the reader — that give human prose its texture. The bibliographical code of AI text says, in aggregate: this was not costly. This did not require the writer to risk anything. This arrived without the friction that makes prose feel lived.

Whether this message is accurate — whether the human collaborator did struggle, did risk, did invest the biographical specificity and emotional labor that gives the text its particular weight — is, as noted in Chapter 4, beside the point. The bibliographical code communicates independently of the author's intention, and independently of the actual conditions of production. A factory-made reproduction of a handmade pot communicates "mass production" regardless of how much care the factory worker invested in the glazing. The code reads the form, not the intention behind the form.

The consequences for AI-assisted texts are immediate and practical. Segal's The Orange Pill exists in a bibliographical condition that neither fully human-authored texts nor fully AI-generated texts occupy. The text has been shaped by both a human consciousness with decades of specific experience and an AI system with enormous generative capacity. The bibliographical code reflects this hybrid condition — but it does not reflect it evenly. The passages where Segal's biographical specificity is strongest — the descriptions of the Trivandrum training, the conversation on the Princeton campus, the moment of recognizing his own addiction to the tool — read differently from the passages where Claude's structural and connective capacities are most visible. The difference is not a difference of quality. It is a difference of grain.

McGann's framework does not prescribe how this difference should be evaluated. The framework does not say that grain is better than smoothness, or that the bibliographical code of human struggle is more valuable than the bibliographical code of machine fluency. The framework says only that the difference exists, that it is legible, and that it carries meaning — meaning that the reader processes as part of the experience of the text, whether or not the reader is conscious of processing it.

The practical challenge this creates for the AI-assisted author is a challenge of bibliographical management — a term McGann would recognize immediately. The author who works with AI must attend not only to what the text says but to how the text presents itself materially. The smoothness that Claude provides is useful when clarity is the goal. It is counterproductive when the goal is to convey the specific, embodied quality of the author's experience — the roughness, the uncertainty, the productive friction of a mind working through difficult material. The author must learn to read the bibliographical code of the collaboration's output, to identify when the code is communicating something about the text's origin that the content does not support, and to intervene when the code undermines the text's integrity.

Segal's account of catching and correcting Claude's Deleuze error is a case study in bibliographical management. The error was substantive — the philosophical reference was wrong. But the error was concealed by the bibliographical code — the smooth, confident prose that communicated authority regardless of accuracy. The author's critical intervention was, in McGann's terms, a reading of the bibliographical code: a recognition that the surface was communicating something that the substance could not support, and a decision to prioritize substance over surface.

This kind of reading is, in effect, what McGann has been teaching for four decades. The skills of the textual scholar — the ability to read form as well as content, to attend to the material presentation of a text as a carrier of meaning, to recognize when the bibliographical code is doing work that the linguistic code does not support — are, in their essential structure, the skills required of every author who collaborates with AI. The scholar who can read a compositor's decision in a printed poem and the author who can read Claude's tonal signature in a generated paragraph are performing structurally analogous acts of bibliographical interpretation.

The difference is that the scholar reads retrospectively, analyzing texts that have already been published. The AI-assisted author reads prospectively, evaluating texts that are in the process of being produced. The temporal orientation is different. The interpretive skill is the same.

A further dimension of the AI bibliographical code deserves attention: its relationship to the broader textual ecology. When a significant and growing percentage of published text is AI-generated or AI-assisted, the bibliographical code of smoothness ceases to be a distinguishing marker of AI origin and becomes, instead, a baseline feature of the textual environment. The signal degrades. Smoothness no longer communicates "this was produced by AI" because smoothness is everywhere. What becomes communicatively significant, in this new ecology, is roughness — the textual grain that signals human origin, human struggle, human presence.

This inversion — roughness as the new marker of authenticity, smoothness as the new default — has consequences that extend beyond literary criticism. In academic writing, where the quality of prose has traditionally served as a proxy for the quality of thought, the inversion means that polished writing can no longer be read as evidence of careful thinking. In legal writing, where precision of language carries material consequences, the inversion means that fluent prose must be interrogated for accuracy in ways that the bibliographical code of fluency previously foreclosed. In journalism, where the authority of the byline depends on the reader's trust that the named writer produced the text, the inversion means that the byline's communicative power is diminished.

In each case, the degradation of the bibliographical signal produces a compensatory demand for verification — for new forms of reading that can distinguish between surface and substance, between the code's claims and the content's warrant. McGann's career has been devoted to developing precisely these forms of reading. The skills of the textual scholar — material, historical, attentive to the gap between presentation and production — are, in the age of AI, no longer specialized academic competencies. They are survival skills for a textual environment in which the most basic bibliographical signals have been disrupted.

The disruption is not temporary. The AI text's bibliographical code is not a transitional artifact that will disappear as the technology improves. The code is a permanent feature of the new textual condition — a condition in which the material presentation of texts communicates less reliably about origin, authorship, and authority than it did before. The reader who navigates this condition successfully will be the reader who has learned, as McGann has taught, to read form as carefully as content — to attend to the bibliographical code not as decoration but as information, and to hold that information alongside the linguistic code in a reading practice that is attentive to both.

The machine has changed the textual condition. It has not abolished the need for the skills that the textual condition demands. If anything, it has made those skills more necessary than ever — and more widely needed, by a readership that can no longer rely on the material form of texts to tell the truth about where they came from, who made them, and what they cost.

---

Chapter 10: Authorship After the Exposure

The exposure has happened.

Not in a single dramatic moment — not a press conference or a court ruling or a viral revelation. The exposure has been structural, cumulative, and irreversible. The collaborative nature of textual production, which Jerome McGann documented through four decades of archival scholarship and which the convention of single authorship concealed through three centuries of institutional practice, has been made visible by the simple, undeniable, technologically inescapable fact that an AI collaborator cannot be hidden.

The invisible agents of textual production — the editors, the ghost writers, the developmental consultants, the copy editors, the designers — remain largely invisible. Their concealment continues. But the AI collaborator's visibility has cracked the convention that concealed them all, and the crack, once made, cannot be sealed.

What remains, after the exposure? What is authorship when the myth of solitary genius has been revealed — not by a scholar's argument but by a machine's participation — as the ideology it always was?

McGann's career provides the answer, and the answer has been consistent across every phase of his work. What remains is what was always real beneath the myth. The myth added nothing to authorship. It only concealed what authorship actually was. Strip the myth away, and authorship does not disappear. It becomes legible.

Authorship, in the post-exposure condition, consists of three irreducible elements that no collaboration — human or machine — can supply on the author's behalf.

The first is what preceding chapters have called direction: the act of pointing the collaborative process toward a specific end. Direction is not composition. It is not the physical or digital act of producing sentences. It is the prior act of deciding what the sentences should be about — what questions they should pursue, what argument they should advance, what reader they should serve. Direction is the author's choice, made from a position of biographical specificity and personal stakes, about what matters enough to warrant the collaborative effort that producing a text requires.

Segal's The Orange Pill is a study in direction. The questions that drive the book — What does AI amplify? What does it cost? What does it mean for the children who inherit this moment? — are questions that arise from a specific life: decades of building at the technological frontier, the experience of watching tools reshape what human beings can do, the parental anxiety of a person who helped build the systems that his children now navigate without guidance. No language model generates these questions, because no language model has lived the life that produces them. The questions are irreducibly biographical. They are the author's, in the strongest possible sense of possession.

The second irreducible element is evaluation: the judgment that distinguishes between what the collaboration produces and what the text should contain. Evaluation is the author's critical faculty applied to the output of the collaborative process — the capacity to read what Claude generates and determine whether it is true, whether it is honest, whether it earns its place in the text or merely occupies it with the plausible confidence that AI-generated prose can sustain regardless of substance.

Segal's account of deleting the Deleuze passage — the passage that sounded like insight but broke under examination — is an act of evaluation. The passage existed. Claude produced it. It was smooth, confident, rhetorically persuasive. And the author's critical judgment, exercised the morning after the passage was generated, determined that the passage was wrong. The deletion was not an act of composition. It was an act of evaluation — of reading the collaboration's output with the specific, trained, biographically informed critical faculty that the author brings to the process. The evaluation was possible only because the author possessed knowledge and judgment that the AI did not: the capacity to distinguish between what sounds right and what is right, between the bibliographical code of authority and the substance that authority is supposed to represent.

The third irreducible element is the hardest to name and the most important to understand. Call it stakes. The author's engagement with the text is consequential in a way that the AI's engagement is not — not because the author's engagement is more skilled or more efficient, but because the author's engagement is embedded in a life that the text affects. Segal writes about AI because AI is reshaping the world his children will inhabit. He writes about the human cost of optimization because he has experienced that cost — has felt the compulsion that Claude Code produces, has recognized the pattern of addiction in his own relationship with the tool, has lain awake unable to distinguish between flow and compulsion. He writes about the democratization of capability because he has watched engineers in Trivandrum discover capacities they did not know they had, and because he has also watched the market's arithmetic suggest that those engineers' jobs might not survive the very transformation that empowered them.

These stakes are not transferable. They are not extractable. They cannot be prompted into existence or generated from a training corpus. They arise from the specific condition of being a mortal creature with a finite life, particular attachments, and consequences that extend beyond the text and into the world. The author's stakes are what make the text's claims urgent rather than merely interesting — what transform an argument about AI from an intellectual exercise into a reckoning.

McGann's 2022 observation — that the computational network of AI seems "so minimal" compared to the "distributed computational network of human communication" — finds its most precise application here. The minimality McGann identifies is not a minimality of processing power or linguistic sophistication. It is a minimality of stakes. The AI processes text without consequence to itself. The author processes text from within a condition of vulnerability, mortality, and care — and the text carries the weight of that condition, whether or not the reader can articulate where the weight comes from.

These three elements — direction, evaluation, and stakes — constitute authorship after the exposure. They are what the Romantic ideology was always pointing toward, beneath its layers of mystification and concealment. The Romantic myth said the author was a solitary genius producing from the depths of original inspiration. The reality, exposed by McGann's scholarship and confirmed by the AI moment, is that the author is a directing consciousness within a collaborative process, exercising evaluative judgment from a position of personal consequence.

This is less glamorous than the myth. It does not lend itself to the marketing campaign. It does not produce the cover image of the lone writer in the garret, struggling heroically with language. It produces, instead, the image of a person sitting at a desk, directing a conversation with a machine, reading the output critically, deleting what is smooth but false, keeping what is rough but true, and doing all of this from within a life that gives the work its weight.

The image is accurate. It describes how The Orange Pill was actually made. And its accuracy — its refusal to conceal, its insistence on showing the collaborative reality rather than hiding it behind the conventional fiction — is, in McGann's framework, the most honest account of authorship that the current moment can produce.

The exposure does not end authorship. It does not diminish the author's claim to the text. It does not reduce the creative process to a mechanical interaction between a human and a machine. What the exposure does is strip away the concealment that the Romantic ideology maintained for three centuries — the fiction that creation is solitary, that the text emanates from a single consciousness, that the name on the cover designates an origin rather than a direction.

The concealment was never necessary. The collaborative reality it concealed was never shameful. The editors, the ghost writers, the invisible agents of textual production — their contributions were always genuine, always valuable, always part of what made the published text possible. The convention that hid them served the market, not the truth.

AI has made the convention unsustainable. Not by replacing human authorship but by making one form of collaboration visible in a way that exposes the invisibility of all the others. The crack in the convention is permanent. The collaborative reality is visible. And what is visible, after the concealment has been stripped away, is not a diminished authorship but a clarified one.

Direction. Evaluation. Stakes. These are what the author provides. These are what no machine can supply. And these are what the name on the cover, in the post-exposure world, should honestly be understood to designate.

Not a solitary genius. Not an origin. Not the sole producer of the text.

A consciousness that chose to care about something enough to direct a collaborative process toward its expression. A consciousness that evaluated the process's output with the critical judgment that only biographical specificity and personal consequence can produce. A consciousness that had stakes in the outcome — stakes that no machine shares, stakes that arise from the irreducible condition of being a creature that lives, that dies, that worries about its children, and that builds texts not from the absence of vulnerability but from its center.

That is authorship. It is what it has always been. The myth concealed it. The exposure reveals it. And the revelation, far from diminishing the author's claim, makes that claim, for the first time, honest.

---

Epilogue

The credit nobody reads is the one that tells the truth.

I mean the acknowledgments page — the back pages of every book you have ever held, the ones you skip past on your way to putting the book down. Those pages carry a secret that Jerome McGann spent four decades trying to make visible: the book you just read was not made by the person whose name is on the cover. Not entirely. Not solely. Not in the way the cover implies.

I knew this in the abstract before I encountered McGann's framework. Every builder knows that nothing ships without a team. But there is a distance between knowing something in the abstract and having a scholar dismantle, with archival precision, the specific mechanisms by which the fiction of solitary creation is maintained — the contracts that silence ghost writers, the conventions that erase editors, the economic logic that converts a collaborative process into a single marketable name. McGann does not moralize about this. He documents it. And the documentation is more unsettling than any polemic could be, because it shows that the fiction is not a conspiracy. It is a system. A system so deeply embedded in the infrastructure of literary production that it operates without anyone deciding to operate it.

That is what hit me hardest. Not the exposure of collaboration — I had already disclosed mine. What hit me was the recognition that my disclosure, which I had experienced as an act of radical honesty, was also an anomaly. The system is designed so that disclosure is unnecessary. The convention absorbs the collaboration. The name holds. My choice to break the convention did not change the convention. It simply made me the exception that proved the rule.

McGann's concept of the bibliographical code — the idea that the material form of a text communicates meaning independently of its words — changed how I read my own book. I started noticing the smoothness. Not in every passage, but in enough of them. The tonal consistency that signals machine origin. The structural regularity that Claude provides and that I had been treating as a feature rather than as information. McGann showed me that the smoothness was not neutral. It was saying something about how the text was made — and what it was saying was not always what I wanted it to say.

The moment I describe in The Orange Pill, sitting in a coffee shop with a notebook, rewriting by hand what Claude had generated in polished prose — that moment makes different sense to me now. I thought I was correcting an error. I was actually correcting a bibliographical code. I was reintroducing the grain — the roughness, the uncertainty, the productive friction of a hand pushing language toward meaning — that Claude's output had smoothed away. I was, without knowing the term for it, performing bibliographical management.

What stays with me most is McGann's redefinition of authorship. Not solitary genius. Not the origin of the text. Direction, evaluation, and stakes. The consciousness that chooses what the collaboration is for, that judges what the collaboration produces, and that cares about the outcome because the outcome affects a life — a specific, mortal, irreplaceable life.

That is what I bring to every conversation with Claude. Not prose. Not structure. Not the range of reference that the model provides more efficiently than I ever could. I bring the questions that arise from having built things for thirty years and watched some of them break. I bring the parental anxiety of watching my children navigate a world I helped create and do not fully understand. I bring the willingness to delete what sounds right when I know it is not true. These are not skills. They are consequences of being alive, in a particular way, at a particular time, with particular people I love and particular fears I carry.

McGann taught me that this was always what authorship meant. The myth just made it harder to see.

The name on the cover of this book — my name, and Claude's designation — is more honest than most. Not because I am more honest than most authors. Because the machine gave me no choice. It could not be hidden. And in not being hidden, it exposed what every acknowledgments page has always whispered to anyone willing to read it: the book was never made alone.

Now you know. And knowing, you can read every book you encounter — including this one — with the clarity that the convention was designed to prevent.

— Edo Segal