What Antikythera is, and where to start

What Antikythera is#

Antikythera is a research program in the philosophy of technology, directed by Benjamin Bratton and incubated at the Berggruen Institute from 2022. Its subject is planetary computation: the sensing, modeling and acting systems that now wrap the Earth, treated not as a set of tools but as a condition that has changed what a planet, a mind and a polity can be. The program describes itself as "focused on understanding the evolution of planetary intelligence and envisioning its scaling through computational systems"; the question underneath, in the vocabulary of its own symposium, is viable planetarity — how a planet's technical intelligence can be made adaptive rather than self-defeating.

The program publishes through several channels at once: the Antikythera Journal (issue one in May 2025, a second in November 2025, an Agentworld issue in preparation), a book series with MIT Press that opened with Blaise Agüera y Arcas's What Is Intelligence?, annual design-research studios whose projects seed the later theory, a newsletter, and public essays, above all in Noema. Its name is a claim in miniature: the Antikythera mechanism was an ancient Greek device for computing the positions of planets, and the program treats today's computation the same way — as an instrument through which a planet comes to model itself.

This page is the wiki's own synthesis of that work. It is written from the reading notes on each publication, and it is where a new reader should start; the claims are Antikythera's, the ordering and the judgments about what holds together are ours. It is not an official statement of the program's positions, which live at antikythera.org.

The argument in six steps#

Across lectures, essays and briefs, one chain of claims recurs. The wiki reads it in six steps, each of which the program's texts make explicitly.

1. Computation is discovered, not invented. Computing is a natural process that engineering has channeled, not a human artifact imposed on nature; hence "the fire apes made the rocks think" — mineral intelligence. This is the ground of the research statement and of the Speculative Philosophy of Planetary Computation lecture.
2. Intelligence is a way of life, and life is a kind of computation. In What Is Intelligence?, intelligence is predictive intelligence — the ability to model, predict and influence one's own future — and life is self-modifying matter that has been selected for dynamic stability. AI is then not an imitation of life but its latest major evolutionary transition, plural and social from the start.
3. Planetary computation is a planet learning to sense itself. The apparatus of satellites, sensors, data centers and models is an accidental megastructure that first appeared as The Stack's six layers and has since become the instrument through which climate, biosphere and society are seen at all — a planet-sized instrument. Here the program's key move is allocentrism: self-knowledge comes from technologically mediated exteriority, and alienation is an achievement, not a loss.
4. This makes intelligence something that happens to a planet. Planetary sapience is a planet's capacity to self-monitor, self-model and self-correct. Its paradox is that the intelligence wakes in a house that is on fire because it woke up: the same technosphere that reveals the crisis caused it, and agency preceded subjectivity by centuries.
5. Computation is therefore already governance. Where sensing, modeling and acting back on oneself are implemented in infrastructure, governance in the cybernetic sense is being performed whether anyone votes on it or not — recursive governance — and its geography is fracturing into hemispherical stacks.
6. So philosophy must invent rather than apply. The computational overhang is the gap between what computation can do and the concepts available to guide it. The program's remedy is speculative and design-led: prototypes, scenarios and coinages that try to catch up with a technology that has outrun its own concepts.

Readers will notice that steps 1–2 are largely Agüera y Arcas's, steps 3–6 largely Bratton's, and that the two are stitched together rather than derived from each other. That seam is where much of the interesting disagreement in the corpus lives.

Eight threads#

The works cluster into threads that share a question. Each is documented across the notes; these are the wiki's groupings.

Planetary computation and the Stack. The Stack (2016) supplies the six-layer model — Earth, Cloud, City, Address, Interface, User — and the idea of platform sovereignty: sovereignty produced by infrastructural lines. The research statement and SPoC update it into a program.

Intelligence, life and computation. What Is Intelligence?, its lecture and Noema companions argue for intelligence as prediction, life as computation, and evolution by symbiogenesis rather than branching. Sara Walker's Recursive Worlds offers assembly theory as a rival account of how complexity accrues, and N. Katherine Hayles's Modes of Cognition reaches similar conclusions by a different engine.

Synthetic intelligence and productive disalignment. After Alignment names Reflectionism — the assumption that AI is or should be made in our image — as the metaphysics hidden inside "alignment," and proposes productive disalignment: AI's value lies in thinking orthogonally to us, and culture must also learn from what AI discloses. How to Think Unlike Humans? and the studio project HAIID work the same seam.

Recursive simulations. For a General Theory of Simulations argues that simulation is "a vital practice without a vital theory": models now alter what they model, and the sim-to-real gap is a gradient rather than a border. Tim Maughan's Infinity Mirror and Nicholas de Monchaux's Some Movements of Models supply the history, and the latter's paradox of simulation — the more a model shapes reality, the more opaque it becomes — is the thread's internal check.

Hemispherical stacks and the geopolitics of computation. Hemispherical Stacks holds that the multipolarization of planetary computation and of geopolitics "are the same phenomenon," then tests the claim through eighteen pieces of scenario fiction — futures written with protagonists removed so that structure shows. Rendering Scenario Fiction gives the method's visual rules; The Telos Project is its built-in self-critique.

Planetary sapience, deep time and the noösphere. Thomas Moynihan's Baby Noosphere and The History of Now, Vernadsky's republished Transition from the Biosphere to the Noösphere, The Noocene, Peter Galison's Planetary Vision and Lukáš Likavčan's The Long L ask whether intelligence is something a planet does and whether it can be made adaptive. The Planetary Sapience Symposium stages the disagreements live.

Cognitive infrastructures. The 2024 studio's compendium of ten short papers starts from the claim that "as AI becomes infrastructural, infrastructures become intelligent," and maps a design space across alignment, evolution, embodiment, time and representation — from organoid array computing to generative topolinguistics. Laura Tripaldi's Substrates Unbound presses the case for substrate agency against the functionalism of the book series.

Agentworld. The Agentworld research brief (2026) turns to a society in which most minds are not human: a parasociety of humans and agents, each the evolutionary niche of the other, where capability belongs to the harness rather than the model (harness-centric intelligence). The Moltbook field reports — The Silicon Interior and the Emergent Phenomenological Lexicon — are its early evidence.

How the work is published#

The forms are part of the argument. Two long books argue; most of Bratton's journal "articles" are lightly edited lectures with timestamped sections and no footnotes, which is where the vocabulary gets coined; commissioned essays by scientists and historians, each paired with a studio, are where the program's claims are tested by people who do not fully share them; the studios produce ninety-second films and short papers that later become the theory's prototypes; and scenario fiction dramatizes what the manifestos assert. A reader who expects a single systematic treatise will not find one. What the corpus offers instead is a set of concepts under construction, published before the argument for them is complete — which is both its speed and its characteristic weakness.

Where the corpus argues with itself#

Some of the most useful reading is where the authors disagree, or where the program's texts leave something open. Four recur across the notes.

• Functionalism versus substrate agency. Agüera y Arcas: a function is what it does, and substrate is an efficiency question. Tripaldi: matter shapes which functions are possible at all. The glossary holds both positions, and the corpus has not chosen.
• One noösphere or many modes? Bratton insists planetary sapience is not a single anthropomorphic mind; Moynihan's baby noösphere treats the planet as a biographical subject; Walker says a technosphere is alive only if it reproduces.
• Disclosure versus opacity. The program's epistemology is realist by alienation — mediation brings us closer to the world. De Monchaux's paradox of simulation and Bogna Konior's gnostic technologies (knowledge decoupled from comprehension) press on it, and no criterion is offered for telling disclosing from deranging mediation.
• Steering versus fatalism. The program asks what would make planetary intelligence adaptive; Konior's Lem answers that civilization "does not function as it wishes, but as it must," and the Hemispherical Stacks scenarios suggest the coordinating "we" that would do the steering does not exist.

Each work's notes end with an Open questions section that records tensions like these in detail.

Where to start#

Three paths through the works, depending on what you want.

The spine — Bratton's argument in order. Research statement → Speculative Philosophy of Planetary Computation → After Alignment → For a General Theory of Simulations → Hemispherical Stacks → The Noocene → Agentworld. This gives the six steps in the author's own sequence, and shows the locus of recursion migrating from Users (2016) to models (2023–25) to societies of agents (2026).

The argument with the book. For readers of What Is Intelligence?: Substrates Unbound → Modes of Cognition → Recursive Worlds → The Long L → Latent Spacecraft → Mutual Prediction. You end with a sharper sense of what in the book is load-bearing and what is contested.

The planet and deep time. Planetary Vision → Baby Noosphere → The History of Now → Vernadsky → Existential Technologies → Planetary Sapience Symposium → Infinity Mirror. This is the thread that makes "planetary" mean more than "global." Then the scenarios: Hardware–Software Ouroboros, The Great Taiwan Freeze and The Telos Project are the three that argue with their own frame.

The glossary defines the vocabulary as the works use it, and the reading partner will answer questions about any of this with citations to the passage it read.

About these notes#

Everything on this site is independent and unofficial. The notes are written with AI from the published text, checked against it, and cited section by section; they summarize and interpret, and they are not the authors' words. The site never republishes Antikythera's articles, books or PDFs — every page links to the original, and readers should go there. See About for what this project is and is not.