---
title: "Infinity Mirror"
description: "Philip Maughan argues that simulation is a fundamental component of planetary intelligence, and makes the case historically."
type: "reading-notes"
authors:
  - "Philip Maughan"
published: "2025-05-10"
original_url: "https://infinitymirror.antikythera.org/"
doi: "10.1162/ANTI.5CZU"
canonical_url: "https://antikythera.wiki/work/journal/infinitymirror"
md_url: "https://antikythera.wiki/md/work/journal/infinitymirror"
last_updated: "2026-08-23"
site: "Antikythera Wiki"
---

# Infinity Mirror

> Independent, unofficial reading notes from Antikythera Wiki (https://antikythera.wiki). Written with AI from the published text, with citations; not by the work's author and not affiliated with Antikythera (https://antikythera.org). Read the original at the link below.

- **Authors:** Philip Maughan
- **Published:** 2025-05-10
- **Kind:** Journal · Journal article
- **DOI:** https://doi.org/10.1162/ANTI.5CZU
- **Original:** https://infinitymirror.antikythera.org/
- **Length:** 16k words
- **This page:** https://antikythera.wiki/work/journal/infinitymirror
- **Notes generated:** 2026-08-23

## Summary

[Philip Maughan](https://antikythera.wiki/people/philip-maughan) argues that simulation is a fundamental component of planetary intelligence, and makes the case historically. The essay follows simulation technology from the Prussian war game Kriegsspiel through Pentagon wargaming, the video-game-driven rise of the GPU, military whole-earth training environments, and the Earth digital twins of the 2020s, and sets this history against a theory of cognition in which brains themselves are prediction engines built from spatial models. The framing device is retrospective: the essay is written as if from a future in which an orbital supercomputer, the Infinity Mirror, has been built to run the simulations on which Earth's self-regulation depends, and its final section tips into scenario fiction to describe that machine's construction. The historical claim is that tools built for killing, playing, and selling graphics cards became the substrate of the models that let a civilization foresee and manage its own planetary effects, and that this convergence was not planned by anyone.

## The argument

Maughan opens by addressing what he calls the simulation paradox. Western culture carries a suspicion of doubles and representations, from Narcissus to Dostoevsky to Plato's cave and its modern upgrades in Descartes and Chalmers's simulation hypothesis. Against this he sets a working definition: a simulation is the execution of a model over time, a constructive tool for situations where existing knowledge breaks down. Following Margaret Morrison, he denies that laboratory experiment deserves epistemic priority over idealized or even impossible mathematical models. The paradox is that the same machines and algorithms that produce cosmological models also produce closed, brittle worlds (conspiracies, regimes, nostalgias) that resist updating, and that simulations have become sovereign actors in governance.

The second step supplies the theoretical stakes. Drawing on Vernadsky's noosphere and on Adam Frank, David Grinspoon, and Sara Walker's definition of planetary intelligence as collective knowledge integrated into the function of coupled planetary systems, Maughan claims that such intelligence is impossible without simulation. Frank and colleagues' planetary classes culminate in a speculative Class V planet whose technosphere has entered a stable, intentional relationship with the biosphere; the transition from an immature to a mature technosphere is one of building feedback loops, as the Archean biosphere did in becoming the Proterozoic one. Maughan then brings in predictive coding and Jeff Hawkins's cortical-column theory to argue that individual brains already work this way, running thousands of spatial reference frames and settling on perceptions by voting. Planning, at personal and planetary scale alike, is the attempt to minimize prediction error, and the technosphere's proliferation of spatialized computer models is the same process at a larger scale.

The long middle of the essay is the material history. Military training produced simulation as an institution (Kriegsspiel, Lanchester's laws, the Pentagon's 1954 wargaming facility). The US Army's collaboration with Epic Games on America's Army tied military and entertainment development together. Consumer demand for better graphics produced the GPU, whose parallel architecture turned out to be the hardware AI research needed. Bohemia Interactive's Virtual Battle Space series scaled the synthetic battlefield to the whole planet. The same engines and chips then migrated into digital twins, from industrial assembly lines to Synth 3D, kilometer-scale climate models, NVIDIA's Earth-2, and the EU's Destination Earth. Maughan reads these as the conditions Frank and colleagues set out for planetary intelligence, arriving in software.

The final section generalizes: hardware and simulation coevolve, each demanding and enabling the other, and Moore's Law is an aspiration sustained by investment rather than a natural law. The essay then crosses into fiction. A "simulation point omega" arrives in 2034 when the representational complexity of the models exceeds the data available to ground them; energy demand for computation grows six orders of magnitude; and the response is to move computation off-planet into a solar-powered orbital fleet, the Infinity Mirror, leaving the biosphere to recover beneath its technological shell. The machine's galactic models eventually predict technosignatures around distant stars, which is counted as contact of a kind.

## Section by section

### 1 The Simulation Paradox

The opening subsection, on the fear of the double, assembles the literary and philosophical evidence for a persistent distrust of likenesses, and offers a counterexample in the Tang dynasty tale of the divided daughter, whose two selves reunite gratefully. Maughan treats the simulation hypothesis as the latest form of this paranoia and cites Nietzsche's dismissal of a hidden substrate as "the worst proved supposition in the world" alongside Frank Wilczek's question of what a simulated world would be simulated in.

The next subsection defines terms. A model is an informative representation selected for salient features according to the maker's purpose; a simulation is a model, or several overlapping models, run over time. One builds a model and runs a simulation. Simulations can be perturbed, idealized, or made stochastic, and Maughan places them alongside in vitro and other experiment types as equivalent rather than inferior, noting fields (cosmology, planet formation) where simulation is the only method available. Morrison's argument that impossible mathematics, such as infinite systems in statistical mechanics, yields real discoveries supports the point. The flight simulator, more complex than the aircraft it trains for, is the everyday illustration.

The third subsection sorts models into concrete (the Army Corps of Engineers' 1957 hydraulic model of San Francisco Bay, which showed that John Reber's scheme to dam it would flood and waste resources), mathematical (Lotka–Volterra predator–prey dynamics), and computational (Schelling's segregation grid). The fourth introduces the paradox proper. Many simulations seek ground truth and many divert from it; cultural models such as regimes, conspiracies, and nostalgias are simulations too brittle to update. The question Maughan leaves open is how parallel technologies sharing the same computers and algorithmic ancestry end up with opposite relationships to truth, at a moment when simulations have become sovereign actors in factories, cities, and geotechnology.

### 2 Planetary Intelligence

Maughan states the project's aim as chronicling simulation's artificialization, its emergence from human behavior in technical form and integration at planetary scale, and notes that climate models are what allowed the crisis to be foreseen from the 1970s on. Vernadsky's biosphere as geological force and his later "cultural biogeochemical energy" lead to the noosphere, which Maughan reads broadly enough to include social insects, beavers, and neural networks. Frank, Grinspoon, and Walker's definition of planetary intelligence as something that happens to a planet rather than on it is the organizing idea, and their fifth property, autopoiesis, the capacity to maintain identity from within through continual self-reconstitution, is given particular weight.

The major transitions subsection reproduces the five-class planetary scheme and the immature-to-mature technosphere transition, with Peter Haff's technosphere as the "synthetic corona" that metabolizes fossil fuels. An immature technosphere risks pushing coupled systems into states it cannot regulate; a mature one has feedback loops intentionally modified to sustain the whole system. Maughan adds the SETI corollary that only mature technospheres persist long enough to be detected, and that "contact inequality" makes any detected civilization likely older than the searcher. Long-term here means a new eon, not an era.

On how Earth becomes Class V, the essay lists early evidence of nascent planetary intelligence: the coordinated response to chlorofluorocarbons, asteroid defense tests, arms treaties, vaccines. It then claims that simulations are the sites of active inference through which semantic information flows across geosphere, biosphere, and technosphere, like the information networks of the brain.

The two subsections on prediction engines and map makers develop that analogy. The brain builds a world model from which all prediction and action follow; predictive coding holds that experience is constructed from models updated on error, which is why surprise produces more neural activity. Maughan cites Andy Clark's description of minds as distributed cognitive engines spanning brain, body, and world. Hawkins's account of cortical columns as a universal modeling toolkit, descended from the map-making grid cells of the hippocampus and entorhinal cortex, supplies the mechanism: columns broadcast guesses, and the network settles by voting. Because most of the neocortex maintains sub-perceptual processes, Maughan suggests intelligence is better understood as predictive monitoring and self-maintenance than as problem solving. The Anthropocene technosphere, on this reading, built its thousands of spatialized models the way a child learns its environment, within the arenas of competition and play.

### 3 War Games

This is the longest and most densely factual part of the essay. It begins with Kriegsspiel, developed in 1811 for the Prussian officer class and credited, after the Franco-Prussian War, with producing a generation of strategists trained on a synthetic battlefield. Lanchester's Square Law (military power proportional to the square of unit count) shows how calculation tables became predictive procedures. In the United States simulation became bound to state security and to the semiconductor substrate; the Pentagon's permanent wargaming facility opened in 1954 under the motto "All but war is simulation," and Orlando's training-device industry produced the MILES laser engagement system that Milton Bradley later sold as laser tag.

The second subsection follows the opening of military simulation to civilian institutions (RAND, USC's Institute for Creative Technologies) and the eight-year collaboration between the US Army and Epic Games that produced Unreal Engine 2 and America's Army in 2002. Commissioned to cut recruitment costs, the game had more effect on recruits than all other Army advertising combined, by a 2008 MIT study. The Condor Cluster of 1,760 PlayStation 3s, briefly the fastest interactive computer in the Department of Defense, illustrates the essay's central mechanism: console economies of scale and the arms race between chip designers drove improvements that science and the military then exploited.

Two subsections on the GPU give a corrective history. Dedicated graphics hardware predates NVIDIA's 1999 GeForce 256 (Nintendo's Famicom in 1983, ATI's Wonder cards), and early "GPU" was largely marketing. OpenGL and DirectX standardized the platform; the original PlayStation's success revived the card market; the 2006 GeForce 8800GTX, clocked faster than the CPUs it was paired with, marks the crossover. When resolution gains plateaued in the 2010s the industry turned to deep learning for upscaling and ray tracing, so AI improved the chips that had enabled AI research. Generative AI and the hardware to run it spread together.

The Virtual Battle Space and One World Terrain subsections trace Bohemia Interactive from Operation Flashpoint through the Army's DARWARS Ambush! mod to the VBS series, which aligned game controls with real ordnance and put soldiers side by side in monitor-filled rooms, an exercise Maughan compares to Kriegsspiel as collective cognition. VBS4, released in 2020, is "more of a game engine than a game": a whole-earth trainer opening on a manipulable globe, able to ingest drone photogrammetry and simulate tomorrow's weather in the next valley. BAE Systems bought BISim for $200 million in 2022. The section closes by noting that by the early 2020s the same platforms were training autonomous vehicles, planning cities, and updating climate prediction methods unchanged since the 1970s.

### 4 Whole Earth Simulation

The section begins with Stewart Brand's campaign for a photograph of the whole Earth and suggests that dynamic simulations might do what a static image could not. NASA coined "digital twin" in 2010 and backdated it to the capsule replicas of the 1960s; during Apollo 13, engineers worked across fifteen simulators to reconfigure the lunar module. A digital twin has three parts, a physical system, a computational model, and the communication channel between them, and the relation is recursive. Manufacturing twins of products and assembly lines let AI evaluate remedies before a recall. Maughan stresses that this does not ignore the real world but transforms it and is transformed by it.

The Unreal Ground Truth subsection tells how Blackshark, with Maxar's satellite imagery, built Synth 3D for Microsoft Flight Simulator: 1.5 billion buildings and 30 million square kilometers of vegetation, generated by matching 2D imagery to parameters such as population density and shadow angle and applying machine-learned "geotypical" skins to over 90 percent accuracy. The Walmart drone-delivery test defeated by a dog absent from the simulation (dogs were added) makes the sim-to-real point lightly. Autonomous vehicles are trained predominantly in synthetic environments from Unreal or NVIDIA's Omniverse.

On extreme silicon computing, Maughan reports Germany's national digital twin, predictions of continuous hyperlocal forecasting, and Schulthess and Stevens's argument that climate resolution below one kilometer would be a qualitative leap from parametric approximation to explicit physics. The SCREAM model on Oak Ridge's Frontier reached 3.25-kilometer resolution and simulated a year per wall-clock hour on 9,000 nodes. The Earth-2 and Destination Earth subsection records Jensen Huang's 2021 claim that the next frontier in gaming is the laws of physics, and his line that "simulation is the answer"; Earth-2's demonstrations accelerated carbon-capture analysis and ran the FourCastNet weather model, while the EU's Destin-E aims to integrate Earth system models with human subsystems across scales from days to decades. Maughan reads these projects as the overview effect digitized into a "single source of truth," and as the software form of Frank and colleagues' conditions for planetary intelligence, with the difference that twins claim an unbreakable causal link between simulator and simulated that lets disturbances be foreseen and responses automated.

The final subsection asks what remains unsimulable. Biology was expected to hold out, yet Morrison documents wound-closure and tumor-growth models, and James Sharpe describes GPUs transforming tissue simulation at EMBL. DeepMind's 2022 tokamak controller, trained in simulation and then adjusting nineteen magnets 10,000 times a second, found configurations humans had not tried. Evan Schneider's galaxy simulations resolve individual supernovae inside a 100,000-light-year volume. Alexander Lavin's "simulation intelligence" stands for the inductive, open-ended science Maughan sees carrying forward into the Infinity Mirror.

### 5 Infinity Mirror

The closing section first states the coevolution thesis: better hardware enables better simulations, which demand better hardware, through market pressure and through discoveries (quantum computers, for instance) that can only be understood through simulated results. Thought is spatial, as Clark and Hawkins argue, so the chips best at computing moving objects and their physics are central to the artificial expansion of intelligence. A chronology runs from Kriegsspiel to the "simulation crisis" of the early 2020s, the inflection point of the pandemic, the crypto bull market, factory bottlenecks, and early general-purpose AI.

Maughan then turns to the physical limits of silicon. Legacy code and GPU shortages slowed adoption across the sciences; high-numerical-aperture EUV lithography continues shrinking transistor gates, but the binding constraints are heat, quantum tunneling, and background radiation, not the theoretical size of a gate. Moore's Law is "an aspiration," sustained by investment and by some of the most sophisticated experimental technologies ever devised. The text predicts a "Cambrian explosion of hardware" across neuromorphic, photonic, analog, quantum, and biological substrates, each enabling new simulations, and claims that by 2030 simulation's old association with falsehood is forgotten.

The section's last third is fiction, written in the past tense. After restating autopoiesis as the continual remaking of a system's own conditions, it describes the 2034 anomaly: rising noise and error across Earth's most advanced simulations, recognized not as a black swan but as the long-anticipated simulation point omega, when the modeling system's representational complexity surpasses the data it models, raising the question whether overfitting reality produces "epistemic black holes." Energy use for information processing reaches 10^20 joules per year by 2040, from 10^14 two decades earlier. Solar-powered satellite computers launched in quantity federate into the Infinity Mirror, a "glassy overseer" above the Arctic powered by a fraction of the 90,000 terawatts of sunlight Earth receives, with computation joining mining and toxic synthesis off-planet. With this capacity, galactic models predict fluctuations in distant stars' output that count as statistically incontrovertible technosignatures: contact as mathematical certainty rather than physical presence. The essay ends with humans unlikely ever to visit, watching the number of such suns grow as the simulation runs, "pained by their ongoing solitude but confident of things to come."

## Key concepts

- **Simulation** — The execution of a model, or several overlapping models, over time; a constructive tool for situations where knowledge breaks down, not a degraded copy of reality. Maughan distinguishes it from a model, which is built rather than run.
- **Simulation paradox** — The fact that simulations seeking ground truth and simulations that divert from it (closed cultural worlds, brittle ideologies) run on the same hardware and share algorithmic ancestry, at a moment when simulations have become sovereign actors in governance. [Paradox of Simulation](https://antikythera.wiki/terms/paradox-of-simulation)
- **Planetary intelligence** — Frank, Grinspoon, and Walker's term for collective knowledge operating at planetary scale and integrated into coupled planetary systems; something that happens to a planet rather than on it. Maughan's thesis is that it is impossible without simulation.
- **Mature technosphere / Class V planet** — A technosphere that has built feedback loops to stabilize its relationship with the biosphere, by analogy with the Archean-to-Proterozoic maturation of the biosphere. [Mature Technosphere / Class V Planet](https://antikythera.wiki/terms/mature-technosphere-class-v-planet)
- **Autopoiesis** — A system's capacity to maintain its identity from within through continual self-reconstitution; the fifth of Frank and colleagues' properties of planetary cognition, and the essay's model for what the Infinity Mirror sustains.
- **Predictive coding** — The neuroscientific view that experience is built from world models updated on prediction error; Maughan generalizes it so that planning at any scale is the resolution of ambiguity about the future. [Predictive Intelligence](https://antikythera.wiki/terms/predictive-intelligence)
- **Cortical column / reference frame** — In Hawkins's theory, the repeated neocortical unit that models objects in spatial reference frames and votes with other columns on perception; Maughan's template for distributed, collective intelligence. [Column](https://antikythera.wiki/terms/column)
- **Digital twin** — A physical system, a computational model of it, and a communication channel connecting them recursively, so that each alters the other. [Digital Twin](https://antikythera.wiki/terms/digital-twin)
- **Simulation point omega** — In the fictional coda, the moment when the representational complexity of the modeling system surpasses the available data it is meant to model. [Simulation Point Omega](https://antikythera.wiki/terms/simulation-point-omega)
- **Infinity Mirror** — The essay's name for a federated fleet of solar-powered orbital supercomputers running the simulations integral to planetary intelligence, and for the project's history of how such a machine became necessary.
- **Noösphere** — Vernadsky's term, via Teilhard, for the sum of collective cognition as a planetary force, read here to include pre- and post-technological information processing. [Noösphere](https://antikythera.wiki/terms/noosphere)

## Connections

- [For a General Theory of Simulations](https://antikythera.wiki/work/journal/theoryofsim) is the theoretical companion: Bratton's account of reflexive and recursive simulation, simulation anxiety, and the geopolitics of climate models supplies the framework within which Maughan's history sits. The two agree that simulation has become foundational rather than derivative; Maughan adds the hardware genealogy.
- [Some Movements of Models](https://antikythera.wiki/work/journal/movementsofmodels) covers overlapping twentieth-century ground from an architectural angle. De Monchaux's claim that models pull away from the world even as they shape it is in mild tension with Maughan's emphasis on digital twins as an ever-tightening causal loop between model and object.
- [Planetary Vision](https://antikythera.wiki/work/journal/planetaryvision) treats the Event Horizon Telescope as a planet-sized instrument; Maughan's Earth-2 and Destination Earth are the modeling counterpart, and his orbital Infinity Mirror is an instrument larger than the planet in the sense that work entertains.
- [The Long L](https://antikythera.wiki/work/journal/longl) shares the Frank, Alberti, and Kleidon planetary classes and the immature-to-mature technosphere transition, and agrees that a mature technosphere is one folded back into the biosphere. Likavčan theorizes this as substrate-agnostic ecology; Maughan shows the simulation infrastructure it would require.
- [The Transition from the Biosphere to the Noösphere](https://antikythera.wiki/work/journal/noosphere) is the Vernadsky source Maughan draws on for the biosphere as geological force and for cultural biogeochemical energy.
- [Existential Technologies](https://antikythera.wiki/work/journal/existentialtech) and [Recursive Worlds](https://antikythera.wiki/work/journal/recursiveworlds) pursue the same thought, that intelligence constructs reality recursively over long spans, and the Infinity Mirror's fictional coda reads as an existential technology in that sense.
- [The History of Now](https://antikythera.wiki/work/journal/historyofnow) provides the temporal framing, from era to eon, that Maughan invokes when he says the mature technosphere's long term is measured in hundreds of millions of years.
- The 2023 studio project Planetary Counter-Simulation is the adversarial complement: where Maughan narrates the construction of a single source of truth, that project studies how rival models are distorted and degraded.


## Terms used

- [Corrective Loops](https://antikythera.wiki/terms/corrective-loops)
- [Cultural Biogeochemical Energy](https://antikythera.wiki/terms/cultural-biogeochemical-energy)
- [Digital Twin](https://antikythera.wiki/terms/digital-twin)
- [Major Evolutionary Transition](https://antikythera.wiki/terms/major-evolutionary-transition)
- [Mature Technosphere / Class V Planet](https://antikythera.wiki/terms/mature-technosphere-class-v-planet)
- [Noösphere](https://antikythera.wiki/terms/noosphere)
- [Planetary Self-Representation](https://antikythera.wiki/terms/planetary-self-representation)
- [Recursive Modeling](https://antikythera.wiki/terms/recursive-modeling)
- [Sim-to-Real Gap](https://antikythera.wiki/terms/sim-to-real-gap)
- [Simulation Point Omega](https://antikythera.wiki/terms/simulation-point-omega)
- [Synthetic Realism](https://antikythera.wiki/terms/synthetic-realism)
- [Synthetic Training](https://antikythera.wiki/terms/synthetic-training)

## Related works

- [Some Movements of Models](https://antikythera.wiki/work/journal/movementsofmodels)
- [The History of Now](https://antikythera.wiki/work/journal/historyofnow)
- [Existential Technologies](https://antikythera.wiki/work/journal/existentialtech)
- [For a General Theory of Simulations](https://antikythera.wiki/work/journal/theoryofsim)
- [Planetary Vision](https://antikythera.wiki/work/journal/planetaryvision)