The Stack On Software and Sovereignty

Platforms demand an active conversion between economic and technical systems and their respective limitations. Their initial program may be born of economics, but their execution can push sideways through other models of value, confounding and compressing the political spectrum along with them. Their history bears this out. A working technical definition of platform, in general, may include references to a standards-based technical-economic system that simultaneously distributes interfaces through their remote coordination and centralizes their integrated control through that same coordination.⁴ I will unpack this definition below. What I call platform logics refers first to the abstracted systems logic of platforms (their diagrammatics, economics, geography, and epistemology of transaction) and second to the tendency on the part of some systems and social processes to transform themselves according to the needs of the platforms that might serve and support them, both in advance of their participation with that platform and as a result of that participation. Platforms provide an armature and induce processes to conform to it. The Stack is a platform, or, more accurately, a combination of platforms. Its own governing logics are derived from platform logics, but its geography and geometry are also peculiar, and so while stacks are platforms, not all platforms are stacks, and in fact most platforms are not stacks.

42

Platform and Stack, Model and Machine

While systems that arguably operate as platforms might be found in every culture, where does the concept of platform come from, specifically in relation to the development of modern machines? The etymology of platform refers to a “ plan of action, scheme, design ” and, from the Middle French, platte form, or, literally, a plateau or raised level surface. As Benedict Singleton writes, this conjoined with the plot, which itself first implies a plot of land. Once situated on the platform of the stage, the “ plot ” becomes a more abstract structure that situates characters into the forgone conclusion of its unfolding, even as they suffer the choices that aren ’ t really theirs to make. As Singleton would have it, here the plot is a diagram that ensnares the Users of the platform in its designs. ⁵ By at least 1803, platform takes on more explicitly political meaning, as in a “ statement of party policies. ” All three of these connotations (platform as a plan of action, as a stage for a plot, and as proposed rules governance) are important for understanding The Stack as a platform and for platform sovereignty in general. One is set of instructions, one is a situated place where action is played out according to plan, and one is a framework for a political architecture. Already these connotations are slipping and sliding into one another.

Now consider the word program. Its etymology refers first to a “ public edict, ” and in the early modern era, it also comes to mean variously a plan or scheme, a list of events to be presented, a menu of proposed political ideas, and a way to organize how people will occupy architectural space. Only after World War II does “ to program ” mean “ to write software. ” For architecture, computation, and politics, the “ program ” has central significance as a design problem and governing technique. The triangulation of designed site, designed action, and designed polis traces that of “ plot ” : platform and program overlay one another asymmetrically. For example, an architectural program might be defined as an intended organization of Interfaces in a particular arrangement so as to coordinate social contact and interaction (or prevent it). As a diagrammatic image, an architectural program indexes the significance of that organization. A software program is a set of instructions that a designer gives to computational systems with the intention of coordinating that system ’ s internal and external interfaces in relation to itself, to compatible systems, and to Users. An interfacial image of that program, usually the graphical user interface (GUI), summarizes, reduces, and makes those instructions significant for Users. And clearly today, these two kinds of programs intermingle. In many respects, what society used to ask of architecture — the programmatic organization of social connection and disconnection of populations in space and time — it now (also) asks of software. We will return to that shift more than once in the following chapters, and we will have to question what is or isn ’ t the remaining work of physical architecture in light of this. Among what remains is the active contingency of programs, both hard and soft.

Platform and Stack, Model and Machine

43

A recognition of platforms as a third institutional form, along with states and markets, situates the convergence of its architectonic and computational forms in a more specific and fundamental way. A central argument of this book is that the “ political program ” is not only to be found in the legal consensus (or dissensus) and policy admonitions of traditional “ politics ” but also in machines directly. This is where the global-scale arrangement of planetary-scale computing coheres into The Stack, and how the convergence of the architectural and computational design logics of program directly contributes to that system. For our purposes it is far less important how the machine represents a politics than how “ politics ” physically is that machinic system. The construction of platforms draws in, to varying and contingent degrees, strong connotations of “ design ” (design as in to “ designate, ” and to govern through material intervention) and, in this platforms are plots, and (per Singleton) also diagrams that “ ensnare ” actors in their fatal outcomes ( design as in “ to have designs on something, ” to trap the User just so). At the same time, platforms are not master plans, and in many respects, they are the inverse. Like master plans, they are geared toward the coordination of system Interfaces into particular optimized forms, but unlike them, they do not attempt to fix cause and effect so tightly. Platforms are generative mechanisms — engines that set the terms of participation according to fixed protocols (e.g., technical, discursive, formal protocols). They gain size and strength by mediating unplanned and perhaps even unplannable interactions. This is not to say that a platform ’ s formal neutrality is not strategic; one platform will give structure to its layers and its Users in one way, and another in another way, and so their polities are made. This is precisely how platforms are not just technical models but institutional models as well. Their drawing and programming of worlds in particular ways are means for political composition as surely as drawing a line on a map. However, as opposed to the public rights of citizens in a polis and the private rights of homo economicus in a market, we are severely lacking in robust and practical theory of the political design logic of platforms, even as they remake geopolitics in their image (or demand a different language to describe what the political is now or ever was). What we can know from the outset is that an essential logic of platforms is a reconvergence of architectural, computational, and political connotations of “ program ” back into one: the design logic of platforms is the generative program that is equally all three types at once.

At a more mechanical level, a platform is also a standardized diagram or technology. Its structure and the paths of interoperability that hold it together can ’ t be considered outside of the regularization and rationalization of how it connects to the outside world. As infrastructure, a platform ’ s regularity is often guaranteed less by laws than by technical protocols, and this is one of several ways that the sovereign decision is built into the platform ’ s interfacial partitions and surfaces. This kind of intrasystemic standardization was essential to the epochal metatechnologies of industrialization and post-Fordism, revolutionizing the manufacture, distribution, and consumption of massive quantities of identical tangible and intangible items. Because protocols are in place to standardize physical and immaterial properties of integral components and discontiguous manufacturing processes — from the width and direction of grooves in a screw, to the costs of stamps and the nomenclature of international postal zones, longitudinal mean times, cryptographic keys for international monetary transfers, stochastic synchronization of data transfers, and so on — the pace and predictability of industrialization could unfold as it did. ⁶ Artificial standardizations become naturalized as if they were always the measure of things. This kind of complementarity between technique and thought is familiar to adepts of Michel Foucault, Max Weber, Friedrich Kittler and Sam Walton. Standardization drives logistics, and logistics in turn enables geopolitical ambition and momentum. Innovations in munitions standardizations, allowing soldiers to quickly disassemble and repair guns on the battlefield with standard parts, contributed for better or worse to American military prowess in the nineteenth century and its ability to defend a hemispherical doctrine posited by a Virginia farmer, James Monroe. We appreciate the role of railroads, telegraphy, and telephony networks as the infrastructure of globalization, and their speed for the acceleration of the modernities of space and time, but perhaps we underappreciate the metastructuring importance of mundane anonymous standards to turn isolated mechanical inventions into infrastructural innovations (e.g., railroad gauges and spike lengths, timetable templates, the semiotics of graphical interface feedback conventions, transmission line materials, arbitrary telegraphic languages, packet-switching protocols, country codes and area codes, the fixed numeration of money itself, and so on). The centrifugal standardization of how individual components interrelate and assemble into higher-order systems, whether physical or informational, is as important as what any part or component may be. This is how platforms can scale up. To engineer systems that coordinate the shuttling of units from one point to another with efficiency, adaptability, and flexibility is to compose within the rules laid down by other systems, larger and smaller, with which interaction is required. If two different systems share common protocols, then the subsystems of one can interoperate with subsystems of another without nec essarily referring to any metasystemic authority. Systems swap material in this way, such that intermodality and intramodality come to enable one another: no standards, no platform; no platform, no Stack

44

Platform and Stack, Model and Machine

The design of protocols, platforms and programs can be as speculative as needed, but the generativity of standards remains. Protocological interoperability works not only to componentize tangible things, but also to represent undetermined relations between things, events, and locations and to provide the means to compose that traffic in advance. In some cases, these are formal notational systems, and the most ingenious are not always the most widely adopted, and sometimes those adopted become so naturalized that they disappear into the fabric. ⁷ By design, systemic standardization is enforced by fixed physical measurement and procedure, and perhaps here most particularly, the paradoxical tendency of platforms to control and decontrol at the same time is most evident. For example, the formal urban grid in a major city is for the most part rigid and inflexible, but precisely because of this linear and universally authoritarian topography, it affords both maximum tumult of dynamic horizontal interchange in the street plan as well as vertical recombinant programmatic complexity in the skyscrapers that pop up in each of its cells (more on this in the City layer chapter). ⁸ Similarly, it is the legal and practical standard size of the humble paper envelope that makes it possible for it to shuttle messages both discrete and discreet; like the urban grid, the envelope ’ s power is in its dumbness. In the 1970s as the world ’ s cities began to more fully merge into the networked hierarchies of today with the widespread standardization of very-large-scale envelopes, made of steel instead of paper, in the form of fixed proportion and attribute shipping containers. Containerization migrated the packet switching from telecommunications onto the transit of physical objects (or perhaps the other way around). It traded the standardized, linear traffic program of the grounded asphalt grid for another, now smoothed into liquid shipping lanes, pacing big packets of objects back and forth across the avenues of oceans.

Platform and Stack, Model and Machine

45

1. How Platforms Work

Platforms centralize and decentralize at once, drawing many actors into a common infrastructure. They distribute some forms of autonomy to the edges of its networks while also standardizing conditions of communications between them. Many of the defining cultural, political, and economic machines of our time operate as platforms (from Google to transnational political theologies). Platforms are formally neutral but remain, each and every one, uniquely “ ideological ” in how they realize particular strategies for organizing their publics. They are identified with neoliberalism (not without reason), but their origins lie as much within the utopian megastructures of 1960s experimental architecture, counterculture cybernetics, Soviet planning schemes, and many other systems of sociotechnical governance, both realized and imagined. Platforms are infrastructural but rely heavily on aesthetic expression and calibration. A platform ’ s systems are composed of interfaces, protocols, visualizable data, and strategic renderings of geography, time, landscapes, and object fields. For stack platforms, they also include a predominant architecture of interoperable layers. Even as the majority of the information they mediate may be machine-to-machine communication (as, for example, today ’ s Internet), the specific evolution of any one platform, in the ecological niche between the human and inhuman, depends on how it frames the world for those who use it. It draws some things in and draws other things out, but foremost a platform is a drawing and framing machine. Our interest, however, is not to critique platforms as aesthetic works but to identify the work that aesthetics does in their development, and through this to clarify how some existing (and potential) platforms are worthy of our critiques.

46

Platform and Stack, Model and Machine

Platforms might be analyzed in many different ways, and another book might make a more thorough contribution to a very needed general theory of platforms than this one can. In order to discuss The Stack as a platform, however, it is necessary to identify some typological characteristics that all platforms might share in common. These would characterize platforms in relation to other technologies (such as individual machines, executable programs, fixed infrastructure, legal mechanisms, or social norms) and in relation to other institutions (such as states, bureaucracies, and corporations). I list here seventeen criteria and qualities of platforms (a nice prime number). The list is by no means final or exhaustive, but taken as a whole, the shape and function of platforms as both technical and political-economic forms are more clearly specified, especially in relation to The Stack. Some of the criteria listed look like basic principles of secondorder cybernetics, others of software application design, and others of any networkssavvy political science. As such, “ platform theory ” is probably less about inventing new attributes from scratch than it is about observing that recognizable common practices already do constitute platforms as an institutional and technical norm at the scale of states and markets:

1. As opposed to other macrogovernance institutions, platforms do not work according to detailed premeditated master plans; rather they set the stage for actions to unfold through ordered emergence. Bureaucracies, by contrast, are systems that are also dependent on strict protocols and interfaces, but they operate by premodeling desired outcomes and then working backward to codify interactions that would guarantee these: means are a function of ends. Platforms begin by fixing equally strict means but are strategically agnostic as to outcomes: ends are a function of means.

1. Platforms are based on a rigorous standardization of the scale, duration, and morphology of their essential components. The simplicity and rigidity of these standards make platforms predictable for their Users, but also allow them to support idiosyncratic uses that platform designers could never predict. The formal politics of platforms is characterized by this apparent paradox between a strict and invariable mechanism (autocracy of means) providing for an emergent heterogeneity of self-directed uses (liberty of ends). The emergent politics of any one platform may largely be a function of how it strategizes the relationship between standards calibration and the perceived self-interests of its stakeholders.

1. This standardization of essential components produces an effect of generative entrenchment by which one platform ’ s early consolidation of systems (formats, protocols, and interfaces) decreases a User’ s opportunity costs to invest more and more transactions into that particular platform, while it increases the costs to translate earlier investments into another platform ’ s (at least partially) incompatible systems. ⁹ The ongoing consolidation of systems and reduction of transaction costs leverages that advantage toward increasing the robustness of that platform ’ s unique requirements.

Platform and Stack, Model and Machine

47

1. Standardized components may also be reprogrammable within constraints by Users, allowing them to innovate new functions for machines that are composed, at least partially, of preexisting platform systems. The systematic reuse of platform systems allows for the development of complex products based on virtual components, reducing development risks, costs, and project duration. For that innovation, the ratio of what is newly introduced by the User versus what is reused from existing platform systems may be extreme in either direction, though neither ratio directly corresponds to the intrinsic novelty of any one innovation ’ s new functions.

1. The design and governance of platforms often relies on formal models to organize, describe, simulate, predict, and instrumentalize the information under its management. Those models may represent a rigorously discrete view of the platform ’ s internal operations, its external environment, or, most likely, some combination of the internal and the external that measures platform performance according to metrics evaluating its outward-facing systems. ¹⁰

1. Platforms ’ mediation of User-input information may result in an increase in the value of that information for the User. Platform network effects absorb and train that information, making it more visible, more structured, and more extensible for the individual User or in relation to other Users who make further use of it, thereby increasing its social value. At the same time, it is likely the platform itself that derives the most significant net profit from these circulations in total. Each time a User interacts with a platform ’ s governing algorithms, it also trains those decision models, however incrementally, to better evaluate subsequent transactions. An economically sustainable platform is one for which the costs of providing systemic mediation are, in the aggregate, less than the total value of input User information for the platform. Platform economics provides then two surpluses: (1) User surplus, in which the information is made more valuable for the User once involved with the platform at little or no direct cost to that User, and (2) platform surplus, that is, the differential value of all User information for the platform is greater than the costs of providing the platform to Users. ¹¹

1. Like centralizing systems, platforms consolidate heterogeneous actors and events into more orderly alliances, but they themselves are not necessarily situated in a true central position in relation to those alliances in the same way that, for example, a master planning committee or federal capitol building would be. Like some decentralized systems, platforms rationalize the self-directed maneuvers of Users without necessarily superimposing predetermined hierarchies onto their interactions. The centralizationversus-decentralization dichotomy may therefore be illusory in many cases (and not in others) in that the choke points where a platform incentivizes commitment and leverages its advantages over other options may be even more widely distributed than all of the Users that it organizes.

48

Platform and Stack, Model and Machine

1. The generic universality of platforms makes them formally open to all Users, human and nonhuman alike. If the User ’ s actions are interoperable with the protocols of the platform, then in principle, it can communicate with its systems and its economies. For this, platforms generate User identities whether they are desired or not. Platforms can provide identities to Users who would otherwise not have access to systems, economies, territories, and infrastructures, such as a person who is not recognized as a political “ citizen ” by a location, but who is nevertheless included in communication by platforms that are agnostic to the legal status of its Users. At the same time, platforms can also name, enumerate, track, and capitalize the identity of Users who would rather remain anonymous. For the former, the required provision of User identity may be seen as an advantage of platforms and for the latter as a disadvantage.

1. Even as platforms guarantee identities to the Users of its systems, for better or worse, they do not provide these evenly or equally. A platform governs one User differently than it does another. An Interface that may open a space for one User also closes it off to another. An interface that may be open for one User at one moment may be closed at another. This differential is a core technique of how platform sovereignties normalize the exceptional reversibility of the partition. What may be an interiorizing partition ( “ enclave ” ) for one User at one moment may be an exteriorizing partition ( “ camp ” ) for another at another moment.

1. An ideal platform architecture is one that produces a strategic minimum of new content into its own communication economy. An ideal platform is like an empty diagram through which Users mediate new and archived information. A search engine, for example, does not produce new Internet content for its Users, but rather structures the value of content that other Users produce. (If medicine were reconceived as a platform, it would obviously provide new critical information to Users, that is, patients and doctors, as well as organize medical knowledge to date, but it would, in principle, focus the point at which new diagnostic or therapeutic expertise is most crucially required and support it with, for example, highly structured patient data and precedents from the literature). ¹² 11. Any structuring component of an ideal platform architecture is replaceable by a new component, and so the platform could, piece by replaced piece, evolve into something entirely different while retaining its essential shape. As in Theseus ’ s paradox, every plank of wood in a mariner ’ s ship is replaced over time by new wood, and yet the new ship occupies the same virtual place as the old ship and so it still is “ Theseus ’ s ship. ” The same operation holds for platform architecture. Any given component (e.g., layer, protocol, interface) could be replaced, inclusive eventually of all components of the platform in its totality.

1. Platforms may respond to User inputs immediately and may draw on archived rules to recursively govern those interactions in real time, or it may act back on those interactions only once some qualitative or cumulative threshold requirement has been met, perhaps by many Users at once. Platforms govern both instantaneously and cumulatively.

Platform and Stack, Model and Machine

49

1. Ideal platforms not only act on new interactions according to programmed rules and in relation to archived structured information, but also serve as distributed sensing systems that incentivize the detection of errors (or mere anomalies), which are interpreted by the platform ’ s formal models. In principle, what are interpreted as errors will not only update the model ’ s description of the whole, but will also correct the rules by which future interactions are governed. Ideal platforms also treat anomalies not only as errors but as signals of emergent patterns or norms for which some new positive accommodation may be required.

1. The competition between platforms may occur over new tabula rasa space or over the recomposition of one or more existing systems in accordance with a platform ’ s strategy. To date, many successful platforms are those that provide Users with new capabilities by making their existing systems more efficient. Platforms that organize existing systems and information tend to achieve generative entrenchment more quickly than those that seek to introduce new systems from scratch. Users will make tactical use of some platform interfaces to link some existing systems, and in doing so they are incentivized to incorporate more of their own interests within these systems. Subsequent Users are incentivized to link their systems to benefit from the network effects set in motion by earlier Users, who in turn enjoy increasing network benefits as more User systems are incorporated over time. The platform is able to realize platform surplus value from this generative entrenchment.

1. Platforms link actors, information, and events across multiple spatial and temporal scales at once. Platform ubiquity makes it more robust in relation to some threats, both intrinsic and extrinsic, and more vulnerable in relation to others. A platform ’ s ability to defend one component or even replace it when it is no longer useful can make the whole more resilient, but it can also then leave individual components vulnerable. The integrated architecture of the platform may also allow internal component-to-component feedback loops to cycle out of control, amplifying the destabilization of the whole apparatus.

1. A platform ’ s actual processes may be very different from how they are understood by their Users, who may form mental images of those processes based on their own individual interactions or on how the platform has represented itself to them. Platforms don ’ t look like how they work and don ’ t work like how they look. For example, a User may understand his or her own interactions with the platform according to the content hierarchies of a GUI that bears almost no relation to how the platform actually structures or sees that interaction. Architects of a typical Cloud-based platform may organize the system according to the provision (and strategic throttling) of data through application programming interfaces (APIs) that make many different kinds of platform effects possible, the sources of which may be opaque to the most common Users or even to other components of the system.

50

Platform and Stack, Model and Machine

1. Platform sovereignty may be planned or unplanned, universal or specific, generative or reactive, technologically determined or politically guaranteed. Platform sovereignty is automatic under some circumstances and highly contingent under others, and it may function differently in relation to different components of the platform system. The conditionality of these is a function of how platforms relate to other political, technical, and economic institutions that also manage something (or someone) that is also organized by that platform. When two or more platforms mediate the same thing, site, or person, both making claims on it and providing sovereignty to it, then the two sovereignties generated may be mutually constrained. While one of these forms of sovereignty may be universal in relation to the platform that issues it (always subject to the inversions and reversals noted above), it is also only partial and provisional in relation to other platforms (if it is even recognizable by them at all). These differences may be between how two platforms identify the same thing or between how two different components of the same platform (or different components of different platforms) address that thing. While this multiplication prevents any one User from enjoying unlimited universal sovereign privileges, it also tends to prevent any one platform from capturing all sovereignty-generative components within its whole and monopolizing how sovereignty is made, and for whom and what.

To further outline the platform principle, others can add to and modify this provisional list. Some may want to include, for example, demonetization: how platforms sometimes strip certain things of their scarcity and hence exchange value. Some may focus on how platform design can never account for the accidents that actual platforms bring, but also conclude that well-designed platforms can turn accidents into assets. Some may want to specify how and when a User has rights of exit and entrance from and to platforms. Can you leave, and can you get in? Others may want to explore the organizational logics of technical platforms as exemplified by street grids, punch cards, spreadsheets, circuit boards, and so on. Others may come at it from the other side and ask whether standardization works best when predictable outcomes are desired, whereas customization works best when not, and ask how the generic quality of platforms can and cannot do both at once. The Stack is a machine that becomes a state, but it is also how both become platforms, or at least, as one condition around which their armatures are forced to evolve in relation to platforms. As we will see in the chapters ahead, as platforms like The Stack appropriate technologies of sovereignty previously guaranteed to and by the state, the contemporary coevolution of these organizational forms may be punctuated by new disequilibriums. First, we need to better understand the genealogy of platforms as political models and how they have been deployed (successfully and unsuccessfully) as political machines.

Platform and Stack, Model and Machine

51