When Edo Segal writes in [YOU] on AI that he cannot always tell where his thinking ends and Claude’s processing begins, he is not confessing a personal confusion—he is describing, in phenomenological detail, exactly what Hayles’s framework predicted decades before the technology arrived. The liberal humanist subject is defined as the origin of its own thoughts, the locus from which ideas emerge and to which they can be attributed. The moment a builder discovers that a crucial connection arose from the interaction between her question and the machine’s associative processing—that neither partner could claim it as wholly their own—the foundational assumption of that framework has been empirically disrupted. Not as a philosophical argument. As a Tuesday afternoon.
Hayles provides the theoretical architecture that the cycle reaches for throughout. Her concept of the cognitive assemblage—a collectivity in which cognition flows between biological neural networks and silicon-based processing in patterns that neither component could sustain alone—names what the cycle describes experientially. The engineer in Trivandrum who built a complete frontend feature in two days, with domain knowledge that was entirely hers and implementation capability that was entirely Claude’s, was operating as exactly the kind of distributed cognitive system Hayles describes. Agency was not diminished. It was redistributed. She retained the vision, the judgment, the care. Claude provided the execution. The feature was a property of the assemblage.
Her insistence on what she calls material literacy—the ability to read the substrate through its outputs, to hear the characteristic patterns and tendencies of the system that produced a text rather than treating the text as substrate-independent—is the practical discipline the cycle names when it warns against prose that sounds better than it thinks. The large language model’s characteristic smoothness is not a stylistic choice but a material property: the output of an optimization process that rewards coherence without requiring understanding. The builder who cannot read the substrate will accept polished interpolation as insight—a failure of material literacy with consequences that compound.
Hayles stands in the cycle’s gallery as the scholar who supplies the vocabulary that experience keeps outrunning. Where Judea Pearl provides the logical instrument that measures where the machines stand on his ladder of causation, and Norbert Wiener provides the cybernetic frame that locates intelligence in feedback loops rather than individual components, Hayles provides the phenomenological and philosophical account of what it actually means for a human subject to be embedded in a cognitive system that extends beyond its skin. The posthuman she describes is not the robot of science fiction. It is the builder at the keyboard, discovering that the edges of the self are not where she assumed.
Hayles trained in chemistry before turning to English literature, and the double formation marks everything she has written: a scientist’s attention to material processes combined with a literary theorist’s attention to how meanings are made and unmade by the systems that carry them. Her early work on chaos theory and narrative in Chaos Bound (1990) established her method: trace the movement of scientific concepts into literary and cultural forms, and trace the reciprocal shaping of scientific ideas by cultural assumptions. The movement was never one-directional. Information theory had shaped how the 1950s thought about the self; literary forms had shaped how information theory imagined its own scope.
The decisive encounter came from following the concept of information from Shannon’s 1948 paper through the Macy Conferences on cybernetics—the extraordinary series of interdisciplinary meetings, held between 1946 and 1953, at which Norbert Wiener, John von Neumann, Warren McCulloch, and others built the conceptual foundations of what would become computer science, cognitive science, and artificial intelligence. What Hayles found at those conferences was the moment of original sin: the moment when information was defined as pattern independent of substrate, when the engineer’s legitimate abstraction was elevated into a metaphysical claim about the nature of mind. If information was substrate-independent, then the mind could be uploaded, the brain was a computer, and the body was an accident rather than a constitutive feature of intelligence. How We Became Posthuman traced how that mistake propagated through the culture and what it costs.
Her subsequent books extended the project into new territory: how digital media change what text can do and mean (Writing Machines), how cognitive processes extend into technical systems that operate below the threshold of consciousness (Unthought), and how biological, ecological, and artificial cognizers are bound into the same planetary cognitive assemblage (Bacteria to AI). The trajectory is consistent: a sustained insistence that mind is always material, always embodied, always distributed—and that the AI moment makes this insistence not philosophical luxury but practical necessity.
The disembodiment thesis and its costs. Shannon’s mathematical definition of information as pattern independent of substrate was a legitimate engineering simplification: the telephone system does not need to understand conversations to transmit them faithfully. Hayles’s foundational argument is that this engineering insight was generalized, incorrectly and consequentially, into a theory of mind. If information is substrate-independent, the mind can be uploaded, the brain is optional hardware, and the body is irrelevant to cognition. Every one of these extrapolations is wrong, in Hayles’s analysis, for the same reason: information is never actually disembodied. It is always instantiated in a material medium whose specific properties shape what can be said, received, and understood. The embodied cognition her framework defends is not sentimentalism about biology; it is a correction of a foundational error.
The flickering signifier. Print fixes the signifier: the words on the page persist, unchanged, across readings. The digital text does not persist—it is continuously re-generated from underlying data structures, rendered by software, displayed by hardware, subject to alteration at every level of the stack without leaving a trace. The signifier flickers. The AI-generated text is the limiting case: it flickers into existence, token by token, through stochastic selection from a probability distribution. The same prompt submitted to the same model a moment later produces a different text. This radical contingency is a medium-specific property with consequences for how such texts should be read and evaluated—consequences the disembodiment thesis systematically conceals by treating digital text as equivalent to print.
The cognitive assemblage. Drawing on Edwin Hutchins’s research on distributed cognition and Andy Clark’s extended mind thesis, Hayles developed the concept of the cognitive assemblage—a collectivity in which cognition is distributed across human minds, technical systems, and biological processes. The assemblage does not respect individual boundaries: its outputs are properties of the system rather than of any single component. The human working with Claude is not a person using a tool; she is a node in a cognitive assemblage whose other nodes include the model’s parameters, the training corpus, and the optimization function that shaped both. This reframing changes the question from “What does the machine do?” to “What does the system do?”—and the system includes the human.
Material literacy. Because the AI text is produced by a substrate with specific material properties—a training process optimized for plausibility, a generation procedure that cannot distinguish between what it has encountered and what it is interpolating toward—reading it well requires what Hayles calls material literacy: the ability to hear the substrate in the output, to detect the characteristic smoothness of interpolation, to feel the pull toward the statistical center. The builder who reads with material literacy evaluates AI output against an independent standard maintained by embodied judgment. The builder who lacks it accepts polished surface as substantive depth, and the amplifier carries the resulting noise as faithfully as it carries signal.
The posthuman condition. Hayles insists on the distinction between her posthumanism and the transhumanism that seeks to preserve the liberal humanist subject through technological enhancement—same autonomous individual, but faster and smarter. Her posthumanism reconceives the human as always already embedded in systems that exceeded its comprehension. The boundaries of the self were never where the humanist tradition assumed; they were always permeable, always distributed across tools, relationships, and environments. AI makes this permeability visible rather than creating it. The posthuman she describes is not a future to be feared or anticipated—it is a present to be understood.