A physical system “computes” if we can form a fixed “representation relation” mapping its states (up to some level of reliability) to those of a “Turing Machine,” such that the physical evolution of the system over time parallels the abstract evolution of the Turing Machine. Alan Turing defined computation in prosaic terms: it was any sequence of steps that could be carried out by a human computer following unambiguous rules to read, write, and erase symbols, given an unlimited ream of paper and a pencil with an eraser. This abstract construct—the idealized, tireless human worker with paper and a pencil—is a Turing Machine and the symbols written on the tape comprise “algorithmic information,” or “differences that make a difference.” Universal computation is achieved when the rules allow any other set of rules to themselves be written down as symbols and interpreted by the machine, such that the computation performed is given by those written-out rules, which we call a “program” or “algorithm.” If a physical system can be mapped to a Universal Turing Machine, then the computation is universal or “Turing-complete,” no matter what the physical system is made of or how it works; this property is known as “platform independence” or “multiple realizability.” There has also been recent theoretical work to extend Turing’s original, fully deterministic form of computation to the more realistic regime of “stochastic computation,” which exhibits some degree of randomness.

Human-AI Interaction Design (HAIID) is the wider field of interaction design specific to human-AI interaction, and distinct from HCI or Human-Computer Interaction. HAIID includes chat based interactions, from Eliza to contemporary chatbots, but also agent design, multimodal interface design, and increasingly human–AI driven robotics interaction, and more. If HCI-driven graphical user interfaces came to be based largely on visual icons and the simulation of manipulable spatial metaphors (inside, next to, on top of, bordered by, etc.) HAIID to date is driven instead by the simulation of manipulable socio-cultural cues and the implication of mutual theory of mind (mirroring, interpreting intention, semantic precision, cooperative goal-seeking, etc.) Because of this, HAIID contends with the power and problem of anthropomorphization instead of skeuomorphism; the natural reference cues are not everyday objects but everyday conversation. The scope of possible HAIID technologies is largely yet to be designed as the primary and necessary affordances and presentation of affordances of multimodal and reasoning models (and what succeeds them) remains unclear. For this reason, what may appear to be fringe socio-cultural dynamics between humans and AI may demand further research as novel directions for future HAIID design.

During the early decades of the 21st century, the infrastructures of governance split more decisively from the symbolic economy of politics. The Stack model of functionally defined hardware/ software layers comprising a modular discontiguous megastructure became a standard infrastructural format for the ongoing irregular design of planetary computation. The actuation of governance became increasingly infused in computational systems of information production, modeling and recursion. Accordingly, the simultaneous shift toward multipolar planetary computation and a more multipolar geopolitics not only track one another—they are the same phenomenon in different guises. The splitting of “Stacks” is sometimes delineated by State boundaries (India Stack, for example) but is usually transnational, continental, or even drawn across oceans by existing technological and political alliances (Oceania and the US Stack, for example). These hemispheres are the scale and domain of hemispherical stacks, wherein “sovereignty” is defined by the semi-exclusive rights to produce data about social, economic, ecological and cultural flows within these ambiguously overlapping and partially interconnected geotechnical domains.

Intelligence is the ability to model yourself, your environment, and other intelligences; it allows you to predict how your world will be affected by you and appear to you, and how you and others will act or react, allowing you to exercise agency in your actions. Modeling involves computation. Life forms are inherently computational, and computation requires free energy. Thus it is generally necessary for life to take action to some degree (hence, be intelligent to some degree) in order to obtain free energy in a dynamic environment. There is strong evolutionary pressure on the development of intelligence among life forms whose survival is affected by interaction with other intelligences, which can lead to increasing dynamism and “intelligence explosions.” These are feedback loops in which intelligence begets more intelligence, due both to cooperative and adversarial interactions. Since parallelism increases computational capability, intelligence, like life itself, tends to be multiscale, with smaller intelligences symbiotically composing themselves into larger intelligences.

Life is a complex system that perpetuates itself through time by building itself. Growth, reproduction, maintenance, and healing are all manifestations of this self-construction. Because life is complex (unlike, say, a crystal), self-construction requires a “universal constructor” that can follow directions for self-assembly, which John von Neumann showed to be equivalent to a Universal Turing Machine executing a program. Von Neumann added the proviso that the “tape” on which the program is encoded must itself be copyable, and that the tape must include directions for assembling both whatever does this copying and the universal constructor itself. (In our case, DNA, DNA polymerase, and ribosomes serve these key functions, though none of this was known at the time von Neumann developed his insight.) Life is naturally multiscale—composed of smaller life forms, and composing itself into bigger life forms—because it is at heart computational, and computations can likewise be composed. Thus, if sub-entities contain instructions for building themselves, and enter into a stable symbiosis, then the resulting larger entity also contains instructions for building itself. Because symbioses can only emerge among existing entities, evolution progresses over time from simpler life forms to more complex, composite life forms.

Planetary computation is defined not in terms of information processing but as the aggregate artificial information sensing and processing systems spread discontiguously but in regular networked patterns on, above, and below the Earth’s surface. Planetary computation begins in earnest with the first wide-scale computer networks in the last half of the 20th century, all of which were prefigured by earlier geographic-scale information networks dating at least to the beginning of the industrial era, if not earlier (depending on the scope of operant definition of computation). The emergence of computational technologies that (1) operate at a global scale through the irregular integration of technologies of information producing, transmission, calculation, manipulation, display and / or storage into composite discontinuous assemblages, that (2) allow for computational media to affect planetary systems at the scale of their operations, and / or (3) forms of computational infrastructure capable of relieving fundamental scientific realities about the synchronic and diachronic systemic qualities of Earth’s evolution, condition and future. These three modalities of planetary computation—infrastructural, technological, and epistemic—are often mutually dependent, with the pursuit of once resulting in the accomplishment of another. The architecture of planetary computation.

At any given moment there is a finite amount of available possible computational capacity within the existing aggregate total artificial computational technologies. As of 2025, we estimate this to be roughly 1 zettaFLOPS or 1021 floating point operations per second. For purposes of comparison, 1 zettaFLOPS then equals 1 Planetary Computation Unit (PCU). While the total amount is by definition an abstraction and all but impossible to measure precisely, confident estimates are possible. The single largest category of contributors to this total are the billions of smart phones. Large supercomputers are a small fraction of the whole. It is estimated that over the past several years the total aggregate planetary computation has doubled roughly every 30 months. If this trend were to hold, then in a little over 16 years, there would be 100 times the present compute capacity.

Platform automation is a type of system that refers both to the automation of platforms—the encoding and prescription of biogeochemical, metaphysical, or technical processes—and the platformation of automation, which describes how automation further embeds standardization and amplification of founding conditions into ecosystems and environments. No platforms without automation, no automation without ongoing platform formation. Platform automation evolves via the configurations and cascades of effects of automated processes embedding, encoding, inscribing, and artificially transforming environments in their image through feedback on performance.

Complexity begets complexity by building upon itself. Scaffolding refers to the process by which forms that persist do so by becoming components in yet more complex forms. The resulting assemblages, while new, are also those most “deep in time” according to Assembly Theory in that they contain elemental forms aggregated into novel complex structures. Scaffolding also plays an important role in John Maynard-Smith and Eörs Szathmáry’s theory of evolutionary biology punctuated by “major evolutionary transitions” in which identifiable decisive consolidations of previously distinct entities, differentiation into specialized parts, and increases in energy and information transmitting-processing capacities. A similar principle holds for Brian Arthur’s historical-economic analyses of technological evolution for which new technologies are consolidations of older elements, and may play in a part in what Gilbert Simondon’s terms concretization by which successive iterations of an artifact “condense” multiple formerly separate structures or functions into a single, more integrated whole.

While the term “artificial intelligence” is not incorrect given Antikythera’s more strict definition of artificialization, synthetic intelligence includes additional useful connotations: (1) the synthesis of multiple forms of intelligence acting in temporary concert, (2) the synthetic subtracts the negative connotation of “fake” associated with Aristotle's usage of “artifice”, and (3) an open association with Kant’s notion of “synthetic judgment” for which a concept is combined with another to characterize something important about each. What is recognized as AI may be in itself a form of synthetic intelligence, satisfying all three connotations above and, perhaps more importantly, the interaction between human and machine intelligence can constitute a hybridized form of cognition that may possess capabilities beyond the mere combination of the two.