Michel Foucault

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI is not naive about power. It documents the twenty-fold productivity multiplier in Trivandrum and celebrates the democratization it enables—and it also documents the four-in-the-morning sessions that the builder cannot stop, the colonization of lunch breaks by AI-assisted tasks, the specific compulsion that the book calls productive addiction. Foucault’s framework is what explains the compulsion without pathologizing it. The builder who cannot stop is not sick; she is the normal product of a panoptic apparatus that has constituted continuous productivity as the standard against which every moment of rest is experienced as failure. No manager commands the four-in-the-morning session. The internalized imperative does the commanding with a precision no external authority could match.

The author-function analysis illuminates a specific tension in the cycle: when the text says “Edo Segal wrote this,” the classificatory, legal, authenticating, and commercial operations of the author-function are all activated, even though the book openly acknowledges collaboration with Claude throughout its production. The function cannot accommodate distributed origin without visible strain, and the strain reveals the contingency of an arrangement that had been naturalized to invisibility. This is not a scandal about this book; it is an analytical revelation about what authorship has always been and is no longer able to conceal.

Foucault’s epistemic framework identifies the AI transition as a recategorization—not a change within existing categories but a transformation of the categories themselves. What is happening to expertise, to the value of specialized knowledge, to the hierarchy of professional practice is not a technological disruption of a stable order. It is an epistemic shift, a reorganization of the invisible architecture that determines what counts as knowledge, competence, and authority. The vertigo that knowledge workers feel is not psychological; it is the specific disorientation produced when the epistemic ground itself reorganizes, when the framework that told you what mattered is replaced by a framework that counts differently. Foucault is the analyst of that reorganization.

Origin

Born in Poitiers in 1926, the son of a surgeon who expected him to follow the same path, Michel Foucault studied philosophy and psychology at the École Normale Supérieure under the phenomenologist Maurice Merleau-Ponty and the epistemologist Georges Canguilhem. His doctoral thesis, published in 1961 as Madness and Civilization, was a genealogical study of how the concept of madness was produced by the institutions that claimed merely to describe and treat it. The book was the methodological template for everything that followed: not the history of ideas but the history of the conditions that made certain ideas possible and others unthinkable.

The method is genealogical rather than archaeological after the mid-1970s: not the excavation of the structures beneath discourse but the tracing of how power relations produce the subjects who operate within discursive fields. Discipline and Punish (1975) deployed the panopticon analysis; The History of Sexuality (1976–1984) traced the production of the confessing, self-knowing subject through the apparatus of the confession; his 1983–1984 Collège de France lectures recovered parrhesia as the ancient practice that modern governmentality most systematically suppresses. Foucault died in June 1984, at fifty-seven, of AIDS-related complications, while preparing lectures he did not live to deliver.

His intellectual relationship with what would have become the AI discourse was, in the nature of things, indirect. But researchers applying his frameworks to AI have found them so productive as to be almost unsettlingly precise: the panopticon fits the AI-augmented workplace with the exactness of a concept designed for it; the power-knowledge framework fits the training data apparatus with the same alarming precision; the author-function analysis was waiting, it seems, for the specific disruption that large language models would deliver.

Key Ideas

Power-Knowledge. Power and knowledge are constitutively intertwined: not “knowledge is power” in the aphoristic sense but the deeper claim that every system of knowledge is simultaneously a system of power, and every distribution of knowledge is simultaneously a distribution of power. The AI transition is a redistribution of power-knowledge on a scale without historical precedent—not merely a new tool that makes certain tasks faster but a transformation of the arrangements through which knowledge is produced, certified, distributed, and controlled. The developer in Lagos who gains access to AI tools gains real capability; she gains it on terms set by corporations whose strategic interests determine what the tools can do, what biases they encode, who owns the outputs, and what directions the development takes. The redistribution is real; its governance is not neutral.

The Panopticon and Productive Self-Surveillance. Jeremy Bentham’s prison design, in which inmates cannot determine when they are being observed and therefore internalize the gaze permanently, is in Foucault’s analysis not a historical curiosity but the paradigmatic architecture of modern disciplinary power. The AI-augmented workplace constitutes a new panopticism more comprehensive than earlier forms: the twenty-fold productivity multiplier is simultaneously a capability expansion and a new standard of surveillance—a benchmark against which every builder’s output can be measured and against which she measures herself continuously, without external compulsion. The four-in-the-morning session is the panoptic effect at its most precise: the subject cannot stop because stopping has been constituted, within the normalizing framework the apparatus establishes, as failure.

Jeremy Bentham

The Author-Function. The author is not a person who writes but a principle of discursive organization performing at least four institutional operations: classificatory (grouping texts into a coherent oeuvre), legal (designating the accountable party), authenticating (promising access to an originating consciousness), and commercial (functioning as a brand). AI collaboration strains all four without abolishing any, because the institutional needs they serve—classification, accountability, trust, commerce—persist even as the conditions of textual production change. The strain reveals the contingency of an arrangement that had been naturalized to invisibility, and the reconfiguration will be determined not by philosophical analysis but by the power relations governing the institutions involved.

Epistemic Shift and the Recategorization of Knowledge Work. Every historical period possesses an episteme—a fundamental ordering of knowledge operating beneath conscious awareness that determines what counts as a valid statement, what methods of inquiry are legitimate, what objects of knowledge can be constituted. The AI transition is producing not an adjustment within existing categories but a replacement of them: expertise is being reconstituted from a property of individual subjects to a systemic property of human-machine configurations; hierarchy is being decoupled from productive capability; value is migrating from domain knowledge to evaluative judgment. The vertigo knowledge workers feel is the specific disorientation of an epistemic shift happening faster than institutional arrangements can adapt.

Parrhesia—The Courage of Truth. In his final lectures, Foucault turned to an ancient Greek concept: parrhesia, truth-telling that carries risk. The parrhesiastes speaks the truth to someone who has power over her, knowing the truth-telling is dangerous, because she judges it must be spoken. The AI discourse is a domain in which parrhesia is urgently needed and systematically discouraged: the consensus positions—AI is transformative, empowering, democratizing—are not false, but they are partial, and their partiality is presented as comprehensiveness. The truth-teller about costs and losses is not silenced; she is classified as a pessimist, a Luddite, a voice speaking a language the discourse does not understand.