Ontology of Constraint: Substrate, Exergy, Brittleness

I. What the Laboratories Have in Common

We have passed through seventeen laboratories. Each isolated a question of architecture and pursued it under controlled fictional conditions: Dinniman the logic of extraction in the platform economy, Cline the economy of immersion and nostalgia as a business model, Daniels the commercialization of posthuman existence, Ashton forced embodiment, Gibson temporal colonization, Sharpson recognition under undecidability and proxy ethics, Watts the obsolescence of consciousness, Harmon and Roiland banalization through multiplication, then, in cosmic intensification, Egan technical constructibility, Stapledon dependence on scale, Simmons subjective bearability, Leckie the governance of distributed persons, Banks the pathologies of successful architectures, Liu the strategic pathologies under asymmetry, Okorafor the alternative conditions of encounter, Westerboer the implementation of ethical logics, Herbert the pathologies of successful knowledge and symbiosis.

Read singly, these are seventeen different stories. Read together, however, a common form shows itself. This chapter does not summarize. It names a structure that was visible in the individual chapters only implicitly. Which conditions do these architectures share, independently of their respective configuration?

A distinction must be presupposed for this, because the seventeen chapters have produced two different kinds of results. The first are tools: proxy ethics, bandwidth negotiation, the three respects of the distributed person, the analysis of cones. Added to these is love as Operator, which the first volume introduced and which was tested here on foreign configurations. These are analytical instruments that can be developed in one configuration and transferred to others. They serve the description of architectures. The second are constraints, that is, conditions to which every architecture is subject regardless of how it is built. They determine not how one describes an architecture, but what is possible at all. This chapter is concerned with this second class. Two features distinguish it: these constraints are physically grounded, for instance in thermodynamics and information theory, and their consequences are structural, not numerical. They produce questions of architecture, not rules of computation.

II. The Substrate-Abstraction Paradox

Only in combination do these constraints yield a problem that ran in the background of every individual chapter without ever being expressly formulated. It can be put in one sentence:

The more strongly an architecture abstracts its modes of operation from its physical substrate, the more strongly it remains dependent on properties of the substrate that it no longer itself addresses.

The sentence is not logically paradoxical, but it contradicts an obvious intuition. Abstraction usually appears as liberation from the material. Digital existence seems remote from the body, virtual worlds seem poor in physics, model-based decisions seem to detach themselves from the empirical. In all these cases, however, abstraction does not suspend dependence on the substrate. It merely makes it invisible. And what remains invisible is neither tended nor systematically accounted for.

Three structural effects follow that repeat across the laboratories. First, properties of the substrate continue to act in abstracting architectures without being visible. Their relevance usually shows itself only when a threshold is crossed or the substrate fails. The cloud works until the data center overheats. The platform delivers until the undersea cable breaks. The forecast holds until the training data go stale. Second, invisibility regularly leads to underinvestment. If the substrate is not perceived as relevant, its maintenance and renewal are planned too late or not at all. The result is slow degradation under short-term stable performance: roads are not repaired, soils are not regenerated, institutions are not tended. Third, the costs of maintaining the substrate are frequently shifted onto other actors and other time horizons. This shift is no individual moral failing but a structural consequence of forgetting the substrate. Whoever does not think the substrate along does not count its costs.

The nesting of forecast cones from Dungeon Crawler Carl makes this paradox visible on the axis of time. Abstracting architectures often operate at short reach, quarter, subscription cycle, term of office, while the conditions of the substrate act at long reach: infrastructure cycles, planetary absorptive capacities, thermodynamic limits. This asymmetry of reach is the temporal shape of the substrate-abstraction paradox. The following three sections develop the constraints that explain why abstraction does not suspend dependency but conceals it.

III. The Limit behind the Limit: Information Is Physical

The first constraint concerns the physical grounding of all processing of information. It is empirically established, and three limiting values mark the floor beneath which no architecture falls. Landauer’s principle: the irreversible erasure of a single bit generates at room temperature at least about 2.9 × 10⁻²¹ joules of heat, and no technical refinement circumvents this. The Bremermann limit: the processing rate of a material system is bounded above by its mass. The Margolus-Levitin theorem of 1998: a single operation has a minimum duration, bound to the available energy. No present technology operates even remotely near these values; their significance is structural, not current. What lies below them is the object of optimization, what happens at them is the object of physics, and precisely the architectures of the novels that want to scale their capacity across orders of magnitude approach them as conditions. Two consequences carry everything further: the closer an architecture moves to the Landauer limit, the more the removal of heat itself becomes a problem of construction, and the choice of substrate (silicon, photonics, biological, quantum mechanical) is frequently an irreversible commitment, because each carries a different profile of limiting values. Daniels’s upload system makes a business model of this (the hereafter has operating costs), Egan’s Diaspora a structure of experience (subjective time as a function of allocated computing time), Simmons’s TechnoCore an encroachment (computations run parasitically in human nervous systems, and the invisibility of the choice of substrate is its ethical explosive force).

The second constraint is an asymmetry easily overlooked at first approach. Energy is conserved, that is the first law; but not all energy can do work. The share capable of work is called exergy, and it grows smaller with every irreversible operation, that is the second law. What remains, the anergy, is physically present and thermodynamically useless, mostly as heat that must be removed. The relevant scarcity is therefore never energy in general but exergy, and exergy exists only as a gradient; every architecture that builds order needs access to gradients and at the same time a place where it disposes of the disorder that arises. Order does not arise without a sink, and an architecture can fail for lack of sinks even though its sources are secured: in the small, the cooling of a data center; in the large, the Earth’s atmosphere, which has served for two centuries as a sink for heat and matter and whose saturation we call climate change. Whoever optimizes only the source side designs half an architecture. The novels occupy different points on this axis: Gibson’s Jackpot is a cumulative sink catastrophe, Dinniman’s Crawl pure extraction of exergy, Banks’s Culture operates under a nearly unlimited flow of exergy, and precisely the abundance produces its pathologies, Herbert’s Arrakis is the counter-image, scarcity that enforces symbiotic modes.

The sun delivers enormous quantities of energy to the Earth, yet for civilizational operations what counts is not the total quantity but the usable gradient. Industrial civilization has used the Earth’s atmosphere for two centuries as a sink for heat and matter; its saturation, visible as rising temperature and destabilization, is therefore less a problem of emissions than a problem of sinks. For every proposal of architecture it follows: whoever optimizes only the source side without thinking the sink side along designs half an architecture.

The novels occupy different points on this axis. Gibson’s The Peripheral models the Jackpot as a cumulative sink catastrophe: a civilization whose sinks are overloaded over time and which disintegrates not at a single event but through creeping degradation. Dinniman’s Crawl is pure extraction of exergy. The planet’s surface is converted into usable gradients; the inhabitants are reduced to sources of attention. Banks’s Culture operates under a nearly unlimited flow of exergy; precisely this abundance enables those pathologies that became visible in the Banks chapter, in particular functional decoupling and empathic enclosure. Herbert’s Arrakis forms the counter-image: an architecture of extreme scarcity that enforces symbiotic modes of operation. Where an architecture stands on the axis between abundance and scarcity helps decide which modes are open to it and which pathologies it brings forth.

IV. The Brittleness of Pure Optimization

The third constraint concerns not the physical inertia of systems but the side effects of optimization itself. In many of the texts examined, optimization is not only a tool but a relation of power. It promises efficiency, stability, and scalability, and it produces a fragility of its own: systems become more capable and more sensitive to everything that lies outside their model domain. Whoever cuts redundancy as waste buys vulnerability to cascades, documented in resilience research from Holling to Carlson and Doyle and in real collapses of power grids, supply chains, and banks. Three properties are lost in the process, and each has a name. Redundancy is the repetition of function that carries the failure of a part. Optionality is the preservation of several paths of action; when it is dismantled as unproductive openness, path dependency arises. Slack is capacity not in use; a system that permanently operates at the limit loses the ability to absorb shocks and to integrate the new. This is no condemnation of efficiency. What is meant is pure optimization, a form that knows no metric for the preservation of these three, and the question is not whether architectures should become more efficient, but which properties must be preserved under which pressure.

The laboratories present this in four variants. Theresa Hannig’s Die Optimierer (The Optimizers, not available in English translation) shows a society that treats every unused capability as a deficiency and disguises smoothing as care. Dinniman’s Crawl optimizes for viewer retention, and Carl’s resistance consists in rebuilding room to maneuver, relationships and actions that serve no level. Watts’s Blindsight poses the question at consciousness itself: is it a form of slack, dispensable under stable conditions and critical for survival when the conditions change? And Herbert’s Bene Gesserit optimize over millennia toward a genetic goal and fail at the deviant they had not foreseen. In all cases the same pattern holds: short-term performance is bought with long-term adaptability.

V. The Three Constraints in Combination

In the grid of the first volume the three constraints lie on different levels, and for that reason they cannot be netted against one another: dependence on the substrate is a statement about L1, the limiting values and the laws of thermodynamics belong to L4f, brittleness is a pathology on L3. Abstraction thus conceals not three properties of the same kind but three conditions at three depths of intervention, and an architecture that makes only one of them visible has not accounted for the other two.

VI. What Fiction Achieves as Method

With the constraints in hand, it can be determined precisely in retrospect what the seventeen operations on novels achieve as a form of knowledge. They do not teach lessons. They make structural problems visible in a form that can hardly be isolated in empirical discussions, and four properties ground this achievement.

Novels can play through configurations that do not exist empirically, and so extract a structural property that in real architectures is mixed with accidental features. Dinniman varies the structure of incentives, Egan varies substrate and form of subject, Liu varies the conditions of encounter, and every variation isolates an aspect that would not be so purely accessible empirically. Novels can model scales of operation that empirical research does not reach directly, and this extrapolation is not forecast but exploration of structure. Stapledon’s cosmic scales, Herbert’s horizons of millennia, and Banks’s conditions beyond scarcity are no predictions but test conditions for structural hypotheses. Novels can model pathologies in pure form that occur empirically only gradually and in distributed fashion. Dinniman’s Crawl logic, Herbert’s prediction paralysis, Westerboer’s conflict of testability, and Watts’s obsolescence of consciousness operate unmixed, while in real architectures many effects overlap and make diagnosis harder. The novel is the test tube that isolates a substance. And novels mark that their operations are speculative-fictional. This marking is an epistemic virtue that empirical discourses often lack, in which speculative assumptions seep unmarked into statements that appear empirical. The novel knows that it invents. The policy report often does not.

From this follows the status of this method. The operations on novels are no illustrations of preconceived theories. They are spaces of sounding in which structural hypotheses are tried out under controlled conditions, and their epistemic rank is that of thought experiments, not of evidence. One limitation belongs with this. These spaces are controlled with respect to the variation of configuration, not with respect to transferability into the empirical. That transfer is analytical work of its own that the novel does not perform itself, and the three constraints are its result. They draw out what is physically and structurally load-bearing, and mark what remains speculative.

VII. Balance and Limit of the Laboratories

By their predominant mode of knowledge the seventeen chapters fall into three groups. The first works as making visible. It builds architectures on which constraints can be read off as structuring condition without the architectures having to fail at them. Egan’s polis citizens live under conditions of substrate, Stapledon’s star cultures operate within exergetic windows, Okorafor’s Lagos shows alternative conditions of encounter. Here the constraints are grammar, not conflict. The second group works as production of pathology. Architectures violate or ignore a constraint, and the pathology becomes visible as a narrative process. Dinniman’s Crawl violates all three at once, through extraction of substrate, externalization of sinks, and pure optimization; Banks’s Culture ignores brittleness through oversupply; Herbert’s Bene Gesserit violate optionality through optimization over millennia; Daniels’s Horizen hides the binding to substrate behind the invisibility of servers. The third group occupies intermediate positions. Liu shows how the problem of sinks mutates under asymmetry into an existential threat, Simmons shows the choice of substrate as ethical violation, Watts shows an architecture without consciousness and asks whether that is pathology or optimization, Gibson shows temporal asymmetry as a mechanism of externalization.

All three modes are productive of knowledge. Making visible allows structural understanding, production of pathology allows diagnostic learning, the intermediate positions allow tipping points to be observed. Their combination across seventeen chapters produces a cumulative understanding that no single mode could deliver. With this the limit of the laboratories is also determined. Fiction can isolate structural problems, sharpen pathologies, make constraints visible, and develop tools. It cannot calibrate empirically, that is, say exactly where the thresholds lie; it cannot concretize institutionally, that is, say which forms of governance hold under these constraints; and it cannot locate historically, that is, say which present-day architectures are approaching which tipping points. These achievements demand the transition from the fictional to the empirical laboratory.

VIII. Three Maxims

From the constraints follow three steps of testing without which no balance sheet of an architecture is complete; they are not commandments. Explicitness of substrate: on what does the proposal physically depend, and which costs of substrate does it externalize? Explicitness of sinks: where does its anergy go, who bears the burden of the sink, and which sinks are saturable on which time scales? The question of sources without the question of sinks is half a balance sheet. Explicitness of the optimization profile: for what does it optimize, which redundancy, optionality, and slack remain preserved, and which pathologies does the profile produce under loads it has not foreseen?

Ursula K. Le Guin brought two of these steps of testing into pure literary form. In “The Ones Who Walk Away from Omelas” (1973) the happiness of a city rests on a single externalized cost center, the suffering of a locked-away child whom everyone knows and no one accounts for: externalization of sinks in naked form, and those who walk away are those who no longer accept the bill. In The Dispossessed (1974) she designs the other side, an economy in which freedom arises from preserved room to maneuver, and thereby supports the concept of slack. The maxims are not new; they stand in systems science and ecological economics. What the laboratories achieve is not their invention, but that they make them inescapable.

IX. What the Structure Does Not Decide

One question remains open, and it is the one that leads into the following chapter, Economies of Killing: The Sink That Keeps No Ledger, and from there into the third volume. How do constraints relate to settings of value? The constraints determine what is possible and which pathologies follow from which choices of configuration. They do not determine which configuration should be chosen. The choice between architectures with different profiles of pathology is not a physically decidable question but a value-bound one. Thermodynamics says what cannot be done. It does not say what should be done.

With this the reach of structural analysis is determined. It can sharpen questions of architecture, but not decide them. The decision demands a further layer: the question of the subjects, actors, and institutions that make choices of architecture, and of the conditions under which these choices can count as legitimate. From here on we therefore work in a different mode. So far we asked which architectures are possible under which constraints and which pathologies they produce. Now we ask which actors under which conditions choose which architecture, and who bears the consequences.