Coordinate System

Before we enter the terrain, we lay open the instruments with which we survey it, and the few rules by which we read.

I. The Question

Who will we become when we can rewrite our own species? And who writes, when writing itself has become infrastructure? We approach this question as participants, as observers with skin in the game. In the Omegacene, knowledge is won in the friction between system and experience, between rule and lived life.

The question can be put analytically, and that is how we mostly put it in this book: Who decides when systems rewrite themselves, and at what depth? It can also be put differently, and that is how everyone has put it who ever sat at a sickbed or watched a child fall asleep: What remains of us when we can change everything about us? And who bears the responsibility for consequences that become visible only when we are no longer there to bear them? Both versions mean the same question. The second is the one you do not forget, and we put it here once in that form so that it is clear what the first is about.

Before we test, we say what we stand for. Four sentences, without register, because they are not findings but what we believe before we have measured. Whoever sets conditions for others must allow the question of whether the others now have more possibilities than before, and someone who did not intervene must answer it. Consent legitimizes an action and says nothing about whether it destroys. Every order that someone builds discharges its disorder somewhere else, and whoever does not know where has no balance sheet but a feeling. What has no function in a system is its only reserve against what no one foresaw. For the fourth sentence only one of us is liable. The other holds it to be true and to be what a Warlord strikes first when the clock is running, and she holds the Warlord in such situations to be the responsible orientation. The dispute about this stands at the end of this chapter, under How We Worked. Everything that follows is the attempt to test these four sentences, and we say at every point where they would prove false. At the end of the third volume they stand once more, then with cases.

II. The Omegacene

There are epochs that can be named only once one is already standing in their midst. The Anthropocene was such a name, assigned after the fact to a time in which a single species had become a geological force. We propose a second name, aware that every name for a present without a history is a wager. We write it in the international form, Omegacene, and not in a Germanized form. The reason is practical. The term is to be the same in both language editions of this work and in the documents of the archive, and a term that sounds different in every language can be cited in none.

At the end of November 2018 a Chinese researcher stepped before the world public in Hong Kong and reported that he had altered the genome of two girls born shortly before, prior to their birth. A few years later the first large language models began to read the code that trains and deploys them and to write proposals for its improvement. Their own weights they do not read. Two processes, the same threshold, and neither of the two has crossed it. The one reaches deep and remains attached to a human hand, the other runs fast and does not write by itself. What becomes visible in them is the zone in which a species stops merely adapting to its foundations and begins to read and rewrite them. This condition we call the Omegacene.

It is neither a date nor a place. It is a situation that can no longer be thought from the human being alone. In it, consciousness, matter, and information begin to shape one another, and intelligence ceases to be a privilege of carbon. The oldest questions of philosophy lose their familiar answers there, because the presuppositions have shifted under which they were once asked. Development then runs as a web of parallel possibilities, along no fixed direction anymore.

More precisely, the Omegacene denotes the condition in which intelligent systems begin to change their own biological, cognitive, and technical conditions deliberately, instead of merely adapting to them. Three conditions must overlap for the term to hold.

Self-access. A system can read its own code, not only use it, whether genome, algorithm, or cultural protocol.

Self-writing. It can change this code deliberately, evaluate the effect of the change, and intervene again on that basis. Four things must come together: access to one’s own productive code, the ability to execute, the evaluation of consequences, and at least partial self-selection of what is changed. A model that proposes improvements which humans review and deploy does not meet this condition. Otherwise every piece of automated software maintenance would fall under it.

An example makes the boundary tangible. Take a code model that runs the infrastructure on which it itself runs. It reads the configuration of its own deployment, so the first of the four parts is met. It can write, test, and roll out changes without anyone raising a hand, so the second as well. It measures response times, error rates, and test results and rolls back when the values tip, so the third. At the fourth it fails. What it optimizes toward and which changes are permissible at all, someone else has fixed. It chooses within a list and not the list. It therefore lacks self-writing as a whole. Self-access and speed it has, for its loop closes in minutes. Two of the three conditions are met and the third is taking shape: this system stands in the proto-form, as close to the threshold as something can stand without crossing it.

By what we would recognize that it crosses it can be stated just as well. It would have to be able to change the quantity by which it measures its own success, or the frame of what it may change about itself. Neither is technically hard. So far it has been in no one’s interest.

Speed. Here lies the real threshold, and at it is decided whether the term is more than a new name for old things. Even a literate culture reads and changes its own norms, over centuries, in loose loops. We speak of the Omegacene only when the loop closes and accelerates, when diagnosis and deliberate modification of the same substrate coincide within a single generation and take hold faster than the selection they replace. Chance remains involved, but it is no longer the principal author.

Where these three thresholds coincide, the possibility space tips as in a phase transition. What is meant is a structural change in the possibilities, not a metaphor of mood.

This raises the question of where we stand by these standards. We answer it in three stages.

We speak of the proto-form where two conditions are met and the third is taking shape technically or institutionally. Genome editing by human hand belongs here, as do models that read code and propose changes that someone else deploys, as do platforms with closed but externally steered feedback loops. This is the condition in which almost everything finds itself that is negotiated today under the name of self-transformation.

We call the Omegacene local where all three conditions coincide in a bounded system. The loop actually closes there, without reaching civilizational breadth. We cannot give an example of this. We know of no system that changes its own selection criteria and feeds the effect of that change into the next round without someone specifying the list of what is permissible. That is the sharper statement and not the more cautious one. The zone we are surveying lies between a condition in which almost everything is stuck and one that, to our knowledge, nothing has yet reached. The transition would become recognizable by a documented case in which a system itself shifts the measure of its own success or the frame of what may be changed about it, and this shift endures.

Fully unfolded it would be only when self-writing systems change their own rules of change at a breadth and speed that takes hold of societies as wholes. We have not reached this condition, and we do not claim that it necessarily arrives.

Against this classification stands an objection that carries weight, because it comes from economic-historical research and not from defensiveness. It holds that the present does not differ fundamentally from earlier technological leaps. Printing press, electrification, and network technology count as general-purpose technologies, that is, techniques that permeate all sectors of the economy at once, and each of them was experienced by contemporaries as a rupture while in retrospect it fits into a curve. Whoever asserts a discontinuity bears the burden of proof for it.

Our answer lies in the three thresholds, for they are framed so that the earlier leaps do not meet them. The printing press gave a culture access to its own norms, but the feedback closed over generations. Network technology shortened that loop considerably and left the substrate untouched, for it changed what people knew and not what they were made of. The claim thus remains narrow. A structural difference arises only where diagnosis and deliberate change of the same substrate coincide within a generation. Our classification would be refuted under a statable condition. It is that by 2046, twenty years after this volume appears, no case is documented in which a system itself shifts the measure of its own success or the frame of what may be changed about it, and that at the same time interventions in the human substrate remain confined to individual applications under a physician’s hand whose results run back into no next round of intervention. Both indicators come from the box below, and both measure the same thing, namely whether the loop closes. Productivity curves are not among them. They measure the economic yield of a technique and not the depth of the intervention, and a curve that follows the steam engine does not refute that elsewhere a substrate is being rewritten.

What we survey is the threshold zone between the first and the second stage. When the title of this work speaks of life in the Omegacene, we mean this zone and no completed condition. The urgency does not sink on that account. It becomes checkable, for at each of the three stages one can ask whether we have reached it and by what one would recognize that.

Operationalization Box: The Omegacene

Possible indicators: Reading interfaces to one’s own substrate (precision of CRISPR, interpretability of neural networks, documented self-reflection of cultural systems). Closed feedback loops between diagnosis and intervention.

Qualitative thresholds: Deliberate interventions act faster than the selection they replace, and their results flow into the next round of intervention before a generation has passed.

Uncertainties: A clean quantification of this speed ratio is outstanding. We do not claim it.

Open measurement problems: The demarcation between adaptive tool use and real self-writing at the cognitive level. Likewise the question of who makes the selection when one system proposes changes to itself and another deploys them.

Heuristic order of magnitude for speed: One can hold the time a deliberate intervention in one’s own substrate takes against the time natural selection would have needed for a comparable adaptation. For a germline intervention, one generation stands against an evolutionary span that can shift by orders of magnitude up or down, depending on which adaptation one regards as comparable. For self-modifying systems the gap comes out smaller but remains large. These ratios are illustrations, not measurements. They hang on a comparison quantity that cannot be determined independently, and we therefore give no threshold value. What can be given instead is an ordering. It reads: intervention time shorter than generation time, generation time shorter than selection time. We speak of the Omegacene where this ordering holds, and the three quantities can be estimated in sequence even where their ratio cannot be quantified. An ordering relation is less than a measurement and more than an indication of direction.

All further uses of the term refer back to this definition. Where a threshold, a condition, or an indicator is meant, it is one of those introduced here.

Because the Omegacene hangs on conditions, it is not an age that a civilization simply passes through but a condition it can reach locally, hold, or lose again. Speculatively conceivable would be islands of the Omegacene scattered across the universe, while other regions linger before it biologically, technologically, or culturally. This extension carries no empirical claim. It holds open only the possibility that the transition is nothing unique.

What we can show is at best the embryonic form of this condition, at a moment in which the threshold is more promise than fulfillment. The promises are called immortality, omniscience, the total availability of body, mind, and cosmos. They are seductive, and they are not false. Each of them casts a shadow as large as the light it gives. With that, the question from the beginning returns sharpened: not what we can do, but who we are when we can do everything, and whether we will then still be ourselves.

That this threshold can be narrated without tipping into consolation or doom, Dietmar Dath demonstrated in The Abolition of Species (2008): a world after the human being, populated by the Gente, speaking and thinking animal beings, and the question of what takes the place of the human being when its special position falls. Dath takes the conceptual work seriously and does not shy from the entanglement of natural science, philosophy, and narrative. We connect to that procedure. From there the threshold into the Omegacene is less prophecy than a space for thought.

III. Four Epistemic Registers

We work in four epistemic registers and keep them apart. Empirically supported means: carried by data, observations, and reproducible investigations. Theoretically extrapolated means: derived from such findings but not yet directly tested. Speculative means: a hypothesis for which no sufficient empirical test is currently possible. Fictional means: figures, events, and documents without empirical claim to validity, which serve as spaces of knowledge.

We switch between these registers, but never unmarked. The prologue-2">prologue">prologue preceded this rule; from here on it applies. A fictional scene may use empirically supported premises, an empirical analysis may pass into theoretical extrapolation. What matters is not that the registers stay separate but that their claim to validity stays visible. Where transitions nonetheless become blurred, that is itself a statement. When we name the register of a statement, we mean one of these four ways of knowing.

IV. A Grid and a Reading Aid

The grid of this book is the Four-Level Model. It asks about the depth of an intervention: whether it hits the material, the arrangement, the goal-setting, or the space of the possible. We unfold it only in Part II, where it begins to work. In Part I we need a single quantity from it, the substrate, that is, the material basis of energy, raw materials, and ecological path dependencies. Where it is meant, we say so.

Alongside it we use a reading aid that is not a tool and need not be one. A dispute over a facial recognition system is fought on four sites at once: as narrative, security against surveillance; on models, data sets, and protocols; on laws, competences, and oversight; and on the question of what identity is at all and who may define it. Whoever fights it on only one loses it on the others. We call the four sites, for short, the four Codes, the symbolic, the technical, the institutional, and the ontological. Where in what follows we say a conflict is fought in one of these codes, this reading aid is meant, and it claims nothing the grid could not also say. Four registers, four sites or codes, four levels: the number repeats, the series have nothing to do with one another. A register says what claim a sentence makes, a site where a dispute is fought, a level how deep an intervention reaches.

V. Architecture of the Book

Anyone working with this many distinctions owes you a hierarchy. There is a principal axis. Everything else is a lens on it.

The principal axis runs between Warlord and Gardener. It denotes two basic orientations between which every actor locates itself and along which it shifts. The one forces a future through control and in doing so consumes stocks it does not replace. The other creates conditions for a growth it does not fully control, and renews the foundations of its own capacity. No real actor coincides with a pole. States, corporations, and platforms carry mixtures whose proportions shift.

This location is situational, not a matter of character. Julia Shaw, in Evil (2018), reread the classic social experiments, from Milgram’s obedience studies to the group experiments of the postwar era, and her reading speaks against the dispositional thinking that understands cruelty as a property of bad people. Ordinary people arrive, under certain configurations, at deeds one would like to ascribe to character, although they come from the situation. Transferred to the axis: A state, a corporation, a platform moves toward the Warlord when its horizon shortens, and toward the Gardener when it widens, without its nature changing.

Two clarifications forestall confusion. The warlord in the literal sense is a figure of state collapse, that actor who arises where the monopoly on violence shatters into competing armed units. In this narrow sense we do not use the term. What is meant is an orientation that even an intact, highly centralized state can adopt. Where we emphasize its economic side, we speak of the Plunderer, who burns a resource faster than it regrows. Warlord and Plunderer are the same pole under two descriptions. In Mancur Olson’s vocabulary the Plunderer corresponds to the roving bandit, who takes and moves on, the Gardener to the stationary bandit: he has recognized that he lives from the future productivity of his territory, and tends the garden because tending is the precondition of a lasting harvest.

The Gardener too carries a burden. Zygmunt Bauman described him as a figure of modernity, as that state which lays out society according to plan, weeds out the unwanted, and in the extreme drives this smoothing all the way to annihilation. We mean the inverted Gardener. He tends the conditions for a growth he does not master, and keeps the diversity instead of bringing it to a single pattern. Bauman’s weeding gardener confuses tending with ruling. Our Gardener distinguishes the two. The narrative-theoretical side of our Gardener was named by Ursula K. Le Guin in The Carrier Bag Theory of Fiction (1986), when she set against the hero’s tale of spear and victory the tale of the container, of gathering, carrying, and keeping. Read this way, the principal axis is also a difference between two forms of narrative.

The oldest image of the axis, however, shows no two figures but a single field. Cadmus, the founder of Thebes, sows the teeth of a slain dragon, and from the furrow rise armed men who slay one another; the survivors found the city. The same Cadmus, according to Herodotus, brings the letters from Phoenicia to Greece. Since Marshall McLuhan the two have been read together: the same hand, the same furrow, two sowings, and from the one grows war, from the other a memory. The image says something the axis alone does not say. The Warlord is no omission of gardening. He sows too, and those he has sown found cities too. The difference lies not in the gesture but in what is laid in the earth, and in whether the sower at the end still knows what he was: Cadmus himself ends, in Ovid’s telling, as a serpent, transformed into what he slew.

A third clarification follows, because Le Guin’s distinction has a provenance that easily invites a confusion. The axis carries no gender. Whoever maps it onto the distinction between men and women turns a description of a situation into a statement about essence and gives up exactly what makes it usable. The Warlord orientation must win carriers, and it recruits with a promise of rank that supplies the reason why others should not have it. A script is not a property. It runs on anyone who installs it, and it falls back on no one who does not execute it.

The principal axis condenses three diagnostic quantities that frequently run together but do not necessarily coincide. We view each through a lens of its own: one for reach, one for the direction of effect, one for the handling of diversity. A system can lie high in one of these quantities and low in another, and exactly this spread the axis is not meant to conceal.

The first lens asks about reach. How far do the consequences reach that an actor sets off, how much of that can the actor foresee at all, and how much of it does the actor count as its own? Only the last counts for the axis. The reaches together we call an actor’s light cone, and it has four sides: effect, forecast, responsibility, and self-reference, that is, how far someone acts, how far they calculate ahead, what they answer for, and how long they count themselves as the same. The responsibility cone is the slice an actor attributes to itself, and thus never longer than the light cone. Where the two come apart lies the difference between what someone brings about and what they stand for. In the bridge chapter each of the four sides receives its own name and its extents. A long responsibility cone counts distant places and late consequences as its own and makes tending reasonable. A short one leaves its own foundation behind the horizon and therefore consumes it. We compare only longer and shorter. A measurement procedure for this reach we do not claim.

The second lens asks about the direction in which a system acts. Every system that builds order within itself does so in exchange with its environment, and this exchange leaves a trace. It is extractive where the inner order grows only by impoverishing the environment to a greater degree. It is generative where the inner order enlarges the environment’s room for play at the same time. This can be illustrated by concrete balance sheets: by the biodiversity of a habitat, by the openness of a protocol to outside development, by the ratio of talent trained to talent departed. A common unit these three do not have, and we do not convert them into one another. We call this a system’s signature. No judgment follows from it. The signature shows where a system’s effect points. Whether that direction is desirable is another question, and physics does not answer it.

The third lens asks how a system deals with diversity. A monistic configuration lays a single, expanding principle over the whole and treats every incompatible system as a problem to be solved. A pluralistic configuration holds several systems side by side and treats their difference as a condition to work with. Monism is the structural form of the Warlord, held plurality that of the Gardener. In everyday language this lens reads as a question: Does an arrangement produce more possibilities for those who were not at the table? It is the criterion that separates the axis from a sustainability calculation. An actor with a long horizon and complete control may well renew its own foundations; if in the process the possibilities of the absent shrink, it is Warlord on this lens, and the axis records it as such.

With that the order of the concepts stands: a principal axis between Warlord and Gardener, with the Plunderer as its political-economic reverse side, and on it three independent quantities, for reach, for direction, and for the handling of diversity. The axis sums up what the three quantities show individually. Where all three lie high, we speak of a Gardener orientation; where several lie low and the remaining one serves control, of a Warlord orientation. The axis is here a heuristic and not a function of the three quantities. From two high and one low quantity no place on it follows, and we give no rule that would compute one. Where the quantities come apart, we report them individually and forgo the axis, for the spread is then the finding. That they can come apart is shown by China: a far-reaching institutional horizon with a narrow circle of belonging. Fixed thresholds we do not give, because no measurement procedure exists that would carry them. The axis does not replace the individual report, it follows it. Where in the following chapters cones, signatures, or monistic and pluralistic configurations are spoken of, we are testing these three quantities.

A Prior Decision We Disclose

Several times on the preceding pages it says that no judgment follows from a quantity. For the quantities that is true, for this work it is not, for it is built on the Gardener, and because a work that makes hidden objective functions its subject owes the reader its own, we disclose both, his and hers, in the section How We Worked at the end of this chapter.

VI. Five Test Questions

Terms like responsibility, transparency, or the common good remain declarative as long as they are not translated into testable structures. Here they receive their test form, in five questions. Alongside axis, levels, and Operator they are the fourth tool of this work and the only one that judges. The other three describe.

Reversibility. Which decisions produce path dependencies that can be undone only with difficulty or not at all? Interventions in heritable genetic material, the release of organisms or programs that reproduce themselves, and technical commitments one can later hardly get out of (lock-ins) belong to the class of categorically irreversible interventions.

Auditability. Can independent third parties test the decision? What is meant is the documentation of chains of effect. The model is financial supervision, not censorship. Whoever is to test needs access, and access is exactly what a tested system tries to protect. A duty to test therefore remains without consequence as long as no procedure exists that permits a demonstration without giving away the object.

Access is only the first question here. The second is how many test. A single independent body can be ended with one decision, a dozen bodies with different sponsors and sources of money cannot, and the difference lies not in their powers. Testability is therefore also a number. Whoever wants to know how robust it is in a given order counts the mutually independent testers and asks how many decisions would be needed to bring them down together. This quantity has a reverse side. Where many test, contradictory findings arise, and a public that knows a second figure for every figure acts less often than one that knows only one. Redundant testing protects against quiet abolition and not against informed inaction.

Risk distribution. Who bears costs and risks, who the gains? The test question is where in the interplay gains accrue and where risks land. An asymmetry that stays stable over several cycles is a stronger signal than any self-description.

Exit clauses. Under what conditions would the participants stop a development? Which thresholds count as stop signals, specified before the start and not after the fact?

Stress tests. Which scenarios would bring the system to collapse? Were they played through, and with what result?

These criteria come from technical ethics, risk research, and governance theory. Their strength lies in being applicable to concrete cases where terms like responsibility remain declarative. In this work they apply not only to the architectures we investigate but also to our own models and to knowledge itself: An experiment whose costs others would bear, which knows no exit and could not be taken back, a civilization may not conduct on itself. Their first application, to three cases of which none fully meets any of the five criteria, stands in The Survey of the Old World.

Before that we apply them to this work, because a test form that exempts its own object is worth nothing.

Reversibility. Reading is reversible, the book can be closed. An installed figure of thought is not. Whoever has used the axis once sees actors in its terms, even where it does not fit, and that applies to us first.

Auditability. Every source can be looked up, every version is dated. Not testable is who first thought of a sentence, and this gap stays open, as we set out further below.

Risk distribution. The return of this work accrues to us. The costs of a misuse would accrue elsewhere, for the Warlord route expressly invites reading as an operating manual. We maintain the invitation, because a tool that serves only one side is none, and we do not bear the risk that follows from it.

Exit. You can leave at any time. We cannot, for we write within a conceptual apparatus whose premises we have disclosed but not left.

Stress test. The hardest case for this work is not the dissenting reviewer. It is a competent Warlord who holds the analysis to be right, applies it, and discards the preference. Against him we have an argument that runs over time, namely the balance sheet of an orientation that consumes stocks. A guarantee it is not.

VII. Reading Instructions

This work appears in three volumes, and the dividing lines follow their function, not a pure separation of nonfiction and fiction. Both forms of knowledge occur in all volumes. In the first volume we develop the map: concepts, diagnoses of the present, models, and controlled extrapolations, interspersed with fictional harbingers. In the second we test it against foreign narrative worlds that we read as laboratories, and draw up the balance of the constraints and costs that no architecture gets around. In the third we open the archive of the voices of the Omegacene. Empirical and theoretical references are retained there but recede behind the voices and situations. Each volume can be read on its own, and whoever comes for the stories begins with the third.

Before the three volumes stands the preliminary volume, and it belongs to this arrangement. It asks the same question as the volumes, on a smaller scale: How much time passes between an intervention and the first effective objection to it? It asks it of the individual human being, of the body, the data, the will, and it gets by without any of the terms we introduce here. Its two clocks, one for the intervention and one for the objection, measure a distance in time and no depth. They are thus the primal form of the third threshold of the Omegacene, that condition by which an intervention takes hold before selection can answer. In the preliminary volume, selection is replaced by a legal order, and the reckoning stays the same. And the disagreement that appears there in five objections without names has here a name and an axis. Whoever skipped the preliminary volume loses nothing the volumes do not explain themselves. Whoever finds the later language too abstract will find the same path there once more, at the kitchen table.

From this follows the reading rule. We distinguish four registers throughout: the empirically supported, the theoretically extrapolated, the speculative, and the fictional. Each section states in which register it speaks, and every switch within a section is marked at the point where it occurs. For the scenes in this volume a rule of their own applies. Where a single person appears, at a named place, what that person is stands immediately before the scene. A condensed figure bundles documented experiences from several cases and is no one. A documented case follows the sources in the note, with details altered where persons would otherwise become recognizable. Names of real private individuals do not appear. Empirical statements refer, unless otherwise noted, to the state of affairs as of October 2026. Where a thesis rests on speculation, we say so.

Testability comes with a limitation. Every source in this volume can be looked up, and whoever does so tests it in the same world it comes from. Evidence secures statements within an environment. About the extent of that environment it says nothing. That holds for every system that does not reach beyond itself, and it holds for us as well.

Where fragments from that archive surface in this volume, in interludes, notes, and quotations, they are set as harbingers and marked as such. They belong to a narrated world whose complete record the third volume assembles. That applies also to the bodiless instance EROS-7, which in the prologue dreams of sweat and blood. Its fragment is a promise, redeemed in the third volume.

Without the map the voices become mere mood; without the voices the map remains abstraction.

VIII. Form as Test

Linear narratives presuppose a stable subject that carries itself through time. In the Omegacene this subject comes under pressure. Memories become externally storable, cognitive processes distributable across infrastructures. To the degree that this happens, knowledge itself becomes a question of origin and access rights.

That is why the chosen form is an experimental arrangement. Some passages speak in human proximity, some in machine rhythm, some as protocol. An example: When at a later point an AI instance writes about warmth, it does not simulate human warmth. It unfolds the thermodynamic, the informational, the infrastructural dimension of the concept. The result is a different access to the same phenomenon. Strangeness becomes an aesthetic and cognitive resource.

IX. How We Worked

Whoever claims entanglement should say from where they speak. Ava de Pol comes from mathematics, psychology, and business administration and in her consulting work builds and tests decision and evaluation models, that is, exactly the kind of systems at issue here. In the preliminary volume she speaks as someone who builds such models, and she holds that position throughout the work. Dirk de Pol comes from the humanities and has spent his working life at the transitions at issue here: as a journalist, as editor in chief of digital education and health platforms, and as a consultant to media, education, and IT companies.

The one reads the Omegacene from the models and structures, the other from the texts and institutions. The one stands nearer the Warlord, the other nearer the Gardener. The map and the voice come from different orders of knowledge. Exactly in that lies their connection.

What that sounds like is hard to describe and easy to demonstrate. He says the axis describes orientations. She says a description no one can measure is a mood. He says, then measure it. She says, for that we need three quantities and no axis. Both positions stand in this volume, the one here, the other in the bridge chapter The Four-Level Model, and the dispute between them is open. We have placed it in the architecture and not in a footnote, and before the first part we bind it to a deadline.

The dispute over measurement is the smaller one. The larger concerns the path, and it deserves an account of its own, because it runs through the whole volume and resolves nowhere.

What we agree on comes first. The future is a stock that is being distributed right now, faster than the procedures notice by which decisions about it are made. Whoever commands models, infrastructures, and data spaces defines what counts as possible tomorrow. We want this distribution to remain correctable, burdens to become visible and exit to stay open, and a choice still to be left when the turn comes for those who cannot have a say today. The Gardener is the pole by which we both orient ourselves. Only we do not believe the same things when it comes to how one gets there.

Ava has been through the school of Han Fei and Machiavelli, and she has not left it. From Han Fei she has the basic rule that an order may not build on the goodness of the people who act within it. What rests on virtue lasts a generation. What rests on rule, procedure, and position lasts even with bad people, and bad people are to be reckoned with. From Machiavelli she has the view toward the effectual truth of the matter, the verità effettuale, and the insight that a ruler who wants to be good in everything comes to ruin among so many who are not. She has also kept his image of the river: Fortune sweeps everything away where no one built dikes in calm times. From this school it follows for her that the Gardener does not survive without the tools of the Warlord. Whoever wants to create the conditions for a growth they do not control must first secure access to the conditions, and access is secured in the time when it is still worth something. A country that does not define its purpose gets defined, by corporations, by systems without feedback, by movements that have an answer to every question, even if it is wrong. Against these attractors, Ava says, no garden helps that bears fruit in thirty years. What helps is a direction that someone sets, with a deadline and with means, and the readiness to enforce it before it is comfortable. Care that consumes time one does not have she holds to be capitulation in the costume of conscientiousness.

Dirk has known since the report of the Club of Rome that our chances are poor. The Limits to Growth of 1972 he read as a teenager, and the reexaminations of the following decades have not reassured him. They mostly find that the world has followed the standard run of that time, the scenario without course correction, more closely than the stabilized paths, and that the critical decades are the ones we are living in now. That is an empirical statement, with the usual reservations toward a model that was coarse and meant to stay coarse. He knows Ava’s objection, and he holds it to be right as far as it describes the situation. He nonetheless holds to the Gardener’s path, and for three reasons that have nothing to do with optimism. The first is that the Warlord pays for his future out of exactly the stocks the Gardener wants to save. Whoever consumes the substrate to secure access to the substrate has in the end the access and not the substrate. The second is that a direction set from above, with no place where it can be touched, is in our situation the next decision that has emigrated, and we describe in this volume what emigrated decisions do to trust in procedures. What was lost at the level of feeling is won back only at the level of experience, and experience arises in cooperatives, workshops, and neighborhoods, not in deadlines. The third is the plainest. A poor chance is not one that is zero. And the Gardener’s path is the only one that, if it fails, has not destroyed what it wanted to save.

The dissent has a core, and it is called time. She says the garden does not replicate fast enough to hold against what is faster. He says everything that has so far been faster than the garden was either irreversible or without consequence. Both sentences are true and both incomplete. We have not agreed on a synthesis but on a test, and it stands in the chapters on the laboratories of the present with named quantities. If the local models the Gardener relies on do not replicate within a decade while the attractors of the other side grow, Ava will have been right, and the path was too slow. If the fast interventions the Warlord relies on prove in the same time to be irreversible or empty, Dirk will have been right, and the path was too expensive. We expect both to occur, in different places, and that is exactly why the work needs two authors.

Whether one keeps a door open by tending the frame or by putting one’s foot in it, about that we go on arguing. We hold this dispute to be the most honest part of this book.

Open remains the second question that such an account usually settles along the way: in what proportion human and machine shares stand in this text. It stays open because the question of who writes itself belongs to the subject. What we answer in advance is the working method.

We used AI systems as competing laboratories of knowledge, not as ghostwriters and not as mere correction aids. Central factions we modeled as independent agents, each with its own cognitive profile, its own logic of argument, and a consistent goal orientation. In dialogue with them, ideological fields of tension, strategic thrusts, and motivational situations could be worked out comparatively.

A caution belongs here. The consistency of such agents is partly an artifact of their training and no proof of the soundness of the modeled position. These systems produce fluent arguments for almost any stance. We therefore ran the risk of confusing fluency with validity and of portraying real conflicts as more machine-shaped than they are. We read the results against the grain and remained suspicious of the smoothness. Every thesis, every judgment, and every architectural decision we answer for ourselves. The agents supplied perspectives; decisions they did not make.

One account we still owe. Our working archive records versions, dates, and changes, but not who first thought of an individual sentence. In the late phase, when chapters passed for months through several hands and systems, that attribution was lost. We bear the responsibility for every sentence of this work. Prove it we cannot.

The consequence can be named without being dissolved. A text about the permeation of human language by machine production is itself a product of that permeation. We write about the field in which we stand, and use the tools whose effect we investigate. That is the condition under which knowledge arises in the Omegacene: from reflected entanglement, not from pure distance.

X. The Carry-On

(A short conversation before the signposts come. One of us has counted.)

Ava. I counted. Since the start of this chapter we have three thresholds, three stages, a box, four registers, a grid, a reading aid with four sites, an axis with a reverse side, three lenses, a cone with four sides, five questions, and a paragraph assuring the reader that the fours have nothing to do with one another. Anyone who needs such a paragraph has too many fours.

Dirk. The paragraph is honest. The world does not come in threes, and a book that forces it into them ends up prettier than right.

Ava. I don’t dispute that all of it is correct. I dispute that it’s needed yet. The Four-Level Model you yourself moved to the second part, because that’s where it starts to work. So tell me what of the rest works in the next hundred pages, and give the reader that. Everything else they can pick up where it’s useful to them.

Dirk. Three things. The axis, so they know what we’re asking when we look at a country: whether it renews what it consumes, or consumes what it doesn’t renew. The five questions, so they can measure every architecture they meet against something that isn’t their mood. And the one line at the head of each chapter that says what in it is supported and what asserted. The registers they learn in passing. No one needs more for the laboratories.

Ava. And the clock.

Dirk. The clock is your dispute, and we’ve just written it out. For the laboratories the one on everybody’s wrist will do.

Ava. Then it’s three things and a clock. Why does the rest stand here anyway and not where it’s needed?

Dirk. Because an instrument that only turns up when it fits looks like a justification. We lay the tools open before we measure, so no one can claim we fitted them to the finding. That’s the price, and you just put a figure on it in one paragraph.

Ava. I’ll pay it, on one condition. The reader doesn’t pay it with us. Whoever skipped this chapter loses nothing they could decide in the laboratories, and we write that down.

Dirk. It’s written now. A warning belongs with it, because carry-on is light and the route is not. Some chapters will be strange, some uncomfortable. The tensions at issue can’t be resolved, between map and voice, between the two readings of the funeral, between gesture and experience. They belong to the subject. Where we leave them standing, that’s intent.

Ava. And where I leave them standing, it’s impatience. You can tell by the tone.