<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/"><channel><title>Joseph Bergvinson — Writing</title><description>Joseph Bergvinson, co-founder of SaucerSwap. Essays and theory-fiction on philosophical pessimism, accelerationism, technics, posthumanism, and AI.</description><link>https://josephbergvinson.com/</link><language>en-GB</language><lastBuildDate>Sat, 15 Aug 2026 00:00:00 GMT</lastBuildDate><atom:link href="https://josephbergvinson.com/rss.xml" rel="self" type="application/rss+xml"/><image><url>https://josephbergvinson.com/og-default-75a2f9e56b2fa525.png</url><title>Joseph Bergvinson — Writing</title><link>https://josephbergvinson.com</link></image><item><title>Thought After Consolation</title><link>https://josephbergvinson.com/blog/thought-after-consolation/</link><guid isPermaLink="true">https://josephbergvinson.com/blog/thought-after-consolation/</guid><description>Deconsolated Consciousness and the Machinery of an Inhabitable Life</description><pubDate>Sat, 11 Jul 2026 00:00:00 GMT</pubDate><content:encoded>&lt;h2&gt;The Room Full of Boxes&lt;/h2&gt;
&lt;p&gt;The keys had been on the table for less than an hour when the apartment ceased to be a future and became a room full of boxes. Nothing had gone wrong. The desired object had arrived; only the plenitude attached to it had failed to survive possession.&lt;/p&gt;
&lt;p&gt;For months the rooms had carried more than square footage. They had been relief from noise, proof of arrival, a shape in which the next years might settle. Once the door closed, the future collapsed back into furniture, dust, bills, and the work of deciding where everything went. The disappointment was not that the apartment was bad. It was that no apartment could contain what desire had asked it to deliver.&lt;/p&gt;
&lt;p&gt;A finite object gets recruited to carry a promise it cannot possibly keep. Once this arrives, the pressure will stop; the life around it will cohere; whatever has been deferred will finally begin. The object may still be good. These rooms could shelter friends, hold books, and become part of a life, because the failure belongs to the surplus assigned to them rather than to the rooms themselves.&lt;/p&gt;
&lt;p&gt;There is another failure, stranger and harder to explain. The reasons for a commitment can remain after their force has drained away. You know the relationship matters. You remember what the work once opened. The calendar contains the appointment and the message thread records the promise, yet the next action does not follow because the world in which it made sense no longer presents it with enough weight.&lt;/p&gt;
&lt;p&gt;I call this deconsolated consciousness: the proposition survives, but its grip does not. Information remains available after the arrangement of attention, memory, habit, and affect that made it orienting has fallen silent.&lt;/p&gt;
&lt;p&gt;This is not quite disappointment. The apartment disappoints because desire asked too much of it. Deconsolation begins when a practice or relation may still deserve the judgment once made of it, although the judgment no longer returns as a living reason. Often the two failures arrive together. That is why disenchantment can feel like discovery when what it has really exposed is the machinery on which orientation depended.&lt;/p&gt;
&lt;p&gt;This essay begins there rather than with a verdict on life. Pessimism, in the sense developed here, does not require that every life be bad or that every reform be cosmetic. Its durable proposition is narrower: finite existence admits no final settlement. Desire renews itself, attachment opens onto loss, and goods conflict without a scale capable of reconciling them all. The problem is not that nothing matters. It is that what matters cannot bear the weight consolation places upon it.&lt;/p&gt;
&lt;h2&gt;Goods That Do Not Settle the Whole&lt;/h2&gt;
&lt;p&gt;Schopenhauer gives the structure its most economical mechanism. Desire arrives as pressure and projects relief into an object, then survives satisfaction by returning as boredom or another demand. The metaphysics turns this into Will. We need not follow him that far. Wants subside, mutate, return; meanwhile projects compete for time, and attachment makes experience richer partly because it makes injury possible.&lt;/p&gt;
&lt;p&gt;The account becomes less convincing when every good is forced through that circuit. Music is not valuable only because it ends a discomfort, and care is more than an elaborate technique for releasing tension. Friendship does not become intelligible once its rewards are renamed relief. Aristotle’s energeia, MacIntyre’s goods internal to practices, and Kieran Setiya’s atelic activity approach the matter from different directions, but each notices a value that exists in its exercise. One can stop making music or caring for someone. Neither is completed like a journey.&lt;/p&gt;
&lt;p&gt;Finitude belongs to these goods. A performance ending does not retroactively empty the music, and mortality does not make care unreal. Hägglund makes this dependence central to secular faith: what matters can matter because it may be lost and because maintaining it consumes time that cannot be recovered. Vulnerability is often one of value’s conditions rather than an external defect added to it.&lt;/p&gt;
&lt;p&gt;Finite goods do not quietly add up to harmony. Care for one person may obstruct a vocation. A discipline can give a life form and, over years, become the form that confines it. Fidelity consumes time that another attachment also needs. Because the losses are real, their conflict is real too; no final scale arrives to make every sacrifice commensurable.&lt;/p&gt;
&lt;p&gt;Improvement and reconciliation belong to different orders. Medicine can remove pain that once looked inevitable, material security can loosen fear, and less cruel institutions matter because suffering needs no metaphysical importance before it matters. Still, mortality remains. Desire renews itself. Attachment stays fragile. A life can become much more bearable without becoming complete.&lt;/p&gt;
&lt;p&gt;Nihilism and redemptive optimism meet at this point. Both make finite value answer to a total verdict. One treats the absence of ultimate certification as proof that local goods are empty; the other treats the reality of local goods as evidence that the whole must be justified. The common premise should be refused. A practice can matter without the universe ratifying it, and its failure to reconcile existence does not make its exercise fraudulent.&lt;/p&gt;
&lt;p&gt;This refusal leaves a practical question. If a value cannot draw its force from a final order, how does it become strong enough to organize conduct? A proposition alone does not carry a life. Something has to return the commitment to the present.&lt;/p&gt;
&lt;h2&gt;When Reasons Lose Their Grip&lt;/h2&gt;
&lt;p&gt;Zapffe relocates pessimism from metaphysics to the management of consciousness. Human beings do not merely suffer; they know that they suffer. They anticipate death, compare their lives with geological time, and ask for a meaning commensurate with a universe that supplies none. His governing image is the giant deer whose antlers, an advantage elaborated past utility, outgrew what the animal could carry. Consciousness is that antler: immanent to life and excessive relative to the ease of living.&lt;/p&gt;
&lt;p&gt;Ordinary existence continues because awareness is organized. Some thoughts are excluded before they dissolve practical action. Days fill with tasks. Pressure is forced through form. A person becomes tied to a family, profession, place, ritual, political project, or vocation, and the tie does some of the work that argument cannot.&lt;/p&gt;
&lt;p&gt;Calling these operations defenses is accurate but incomplete. Attention is selective from the outset. A commitment takes practical form by making some possibilities salient and allowing others to recede. A consciousness equally open to everything would not become maximally lucid. It would lose the field within which anything could matter.&lt;/p&gt;
&lt;p&gt;An anchor is more than a belief. It carries salience across time and returns a commitment as something one can act from. Sometimes it is a room; sometimes a schedule, another person, a tool, an inherited phrase, or merely the expectation that the practice will still be there tomorrow. While the arrangement works, much of it remains invisible. Collapse makes the machinery easier to see because it interrupts it.&lt;/p&gt;
&lt;p&gt;Deconsolated consciousness is one result. It should not be romanticized as truth. The disputed literature on depressive realism provides no stable warrant for treating diminished attachment as clearer perception. Breakdown may expose a dependency; it can also distort salience and temporal perspective. The world can lose its grip without revealing what the world finally is.&lt;/p&gt;
&lt;p&gt;Zapffe turns from his diagnosis toward antinatalism. The move is not entailed. Benatar’s later asymmetry counts absent pain as good without a beneficiary while counting absent pleasure as bad only where someone is deprived. Whatever one makes of the argument, its conclusion enters through that evaluative rule rather than through recurrent desire or excess consciousness alone. Diagnosis has no right to impersonate entailment.&lt;/p&gt;
&lt;p&gt;The more durable lesson is that a human life requires selective supports even when those supports cannot be mistaken for cosmic truths. Their contingency does not make them unreal. A rhythm can orient without being eternal. A practice can hold a life together without reconciling existence.&lt;/p&gt;
&lt;h2&gt;Adaptive Consolation&lt;/h2&gt;
&lt;p&gt;Supports have always been technical. Calendars carry obligations among absent people. Books fix rituals; notation lets music survive the hands that wrote it. Stiegler calls this exteriorized memory tertiary retention, although ‘storage’ is too passive a word for what happens. A technical object helps decide what can return, when it returns, and with what force.&lt;/p&gt;
&lt;p&gt;What has changed is not the arrival of mediation but its adaptiveness. The calendar predicts the journey, rearranges the day, and notices which reminders are ignored. The playlist infers a mood and tests the next sequence against the response. A feed keeps traces while shuffling the field in which those traces will matter. The support no longer waits for us in quite the same way.&lt;/p&gt;
&lt;p&gt;Sometimes this deepens a practice. An obscure work comes back; a scattered community finds one another; correspondence survives; administrative noise recedes. Yet adaptation introduces a peculiar volatility because the support now interprets the past in order to shape the next moment, and its interpretation changes as the person reacts.&lt;/p&gt;
&lt;p&gt;The system can be personal without becoming intimate. It knows which song is skipped, the hour that usually receives a reply, perhaps the phrase that comes before abandonment. Such knowledge can return a commitment with remarkable precision. But it is assembled from traces, not from participation in the life those traces express, and it may learn to keep something salient without knowing what the salience is for.&lt;/p&gt;
&lt;p&gt;Zapffe’s mechanisms blur inside this loop. Isolation is perforated by whatever the system predicts will hold attention, including what destabilizes it. Anchors become responsive and, sometimes, less durable. Distraction is industrialized. Sublimation is harder to keep apart from circulation because distress becomes form, form becomes content, and the response to the content begins shaping the next attempt at form.&lt;/p&gt;
&lt;p&gt;The same surface can distract, expose, remember, and judge. You open it to avoid one thought and receive another. You publish something because pressure needed form; it comes back as a number. A community becomes an anchor, then discovers that its visibility depends on a ranking system with a tempo of its own. Relief and destabilization alternate too quickly to remain separate mechanisms.&lt;/p&gt;
&lt;p&gt;None of this gives us a simple decline. Distributed archives preserve memories no private collection could hold, and friendships survive distances because messages and voice notes maintain an ordinary continuity earlier media could not. Adaptive systems can return time and widen attention. The difficulty begins where support and measurement become inseparable, because the thing carrying a practice is also learning how the practitioner responds.&lt;/p&gt;
&lt;p&gt;A stable support could dominate, but its terms usually changed at the pace of an institution, custom, or material object. Adaptive supports can retune themselves continuously. The person encounters not only an inherited form but a form learning how they respond. Consolation becomes responsive at the moment it becomes hardest to distinguish from modulation.&lt;/p&gt;
&lt;p&gt;This difference matters even when the adaptation is benign. A book does not alter its argument because the reader hesitates; a ritual does not immediately rearrange itself around the participant who drifts; a friend may respond, but responds from a life that exceeds the record of prior exchanges. Adaptive systems reduce the delay between trace and adjustment. Their promise is that the support will never become inert. Their danger is that inertness, resistance, and silence may be among the conditions through which a commitment becomes something other than the reflection of its latest measurable response.&lt;/p&gt;
&lt;p&gt;Rather than abolishing anchoring, this deepens our dependence on machinery that enters the formation of the need it answers. A feed helps produce the agitation it relieves; a metric shapes the identity whose continuity it claims merely to measure. Recommendation assists desire and, almost quietly, narrows the field from which desire will later speak.&lt;/p&gt;
&lt;p&gt;Deconsolation can occur while everything still appears to work. The system is present. It remembers and responds. The words are still there. What has vanished is harder to locate: the world those words once helped return.&lt;/p&gt;
&lt;h2&gt;Transmutation Without Redemption&lt;/h2&gt;
&lt;p&gt;Nietzsche’s objection to pessimism begins with its demand for a verdict. Once suffering becomes evidence in a prosecution of life, existence has been placed before a tribunal whose authority remains concealed. A negative metaphysics gives pain a total meaning and thereby repeats consolation in reverse: the world may be condemned rather than saved, but it has still been made legible as a whole.&lt;/p&gt;
&lt;p&gt;The &lt;em&gt;Genealogy of Morality&lt;/em&gt; locates the power of the ascetic ideal in its answer to the sufferer’s question, &lt;em&gt;why?&lt;/em&gt; Meaning preserves willing even when the answer teaches the will to despise its own conditions. Pessimism turns ascetic when injury is made to serve as the final key to existence. Pain is no longer merely suffered, opposed, or relieved; it becomes proof.&lt;/p&gt;
&lt;p&gt;Nietzsche’s earlier tragedy offers another relation to pressure. The wisdom of Silenus is not disproved. Terror is forced through form. Read through Zapffe, tragedy resembles sublimation: awareness is not narrowed so much as intensified and transformed. But the transformation should not be inflated into redemption. A work cannot justify the wound from which it emerged, and endurance cannot prove that what was endured had a purpose.&lt;/p&gt;
&lt;p&gt;Cioran is useful as a discipline against that inflation. Survival does not vindicate the conditions of survival. Creation does not repay pain. Lucidity confers no rank. His own grand negations sometimes violate the restraint, but the prohibition remains: do not make either endurance or despair sovereign.&lt;/p&gt;
&lt;p&gt;Sublimation occupies a narrow passage between exposure and evasion. It can give pressure form without numbing it and create something whose value lies in its exercise. It can also become another demand, another identity, another circuit of valuation. No general rule separates transformation from inflation in advance. The work may metabolize pressure; it does not settle the life containing it.&lt;/p&gt;
&lt;p&gt;This is where pessimism can remain severe without becoming total. It refuses the promise that finite goods add up to reconciliation and refuses the inverse promise that suffering reveals the final truth of existence. Judgment changes scale. It asks what a support enables, what it excludes, what kind of attention it cultivates, and whether the life organized through it remains inhabitable. No cosmic ledger completes the assessment.&lt;/p&gt;
&lt;h2&gt;After Consolation&lt;/h2&gt;
&lt;p&gt;The apartment was no failure; it could still shelter, gather friends, hold books, and become part of a life. Only the surplus promise failed: possession did not settle the life containing it.&lt;/p&gt;
&lt;p&gt;No heroic response follows. The boxes still need opening, and the table has to go somewhere. Perhaps the good begins only after the fantasy of completion recedes, when the room can be used without being asked to redeem the person using it.&lt;/p&gt;
&lt;p&gt;Thought after consolation leaves desire, finite goods, selective supports, and technical mediation in place. It removes the demand that any of them reconcile existence from outside it. A support can orient without proving the world good. Its collapse can wound without revealing a metaphysical truth. An adaptive system can preserve memory and still reorganize the need it serves.&lt;/p&gt;
&lt;p&gt;Deconsolated consciousness makes the dependence visible: concepts remain intact while the world that gave them force falls silent. The task is neither to worship the machinery nor to imagine life without it. It is to distinguish support from settlement, transformation from redemption, and an inhabitable arrangement from the promise that the arrangement will finally be enough.&lt;/p&gt;
&lt;p&gt;The room remains. So does the life. What disappears is the promise that one completion could reconcile the whole.&lt;/p&gt;
&lt;h2&gt;References&lt;/h2&gt;
&lt;p&gt;Aristotle. &lt;em&gt;Nicomachean Ethics&lt;/em&gt;. Translated by Terence Irwin. 2nd ed. Hackett, 1999.&lt;/p&gt;
&lt;p&gt;Benatar, David. &lt;em&gt;Better Never to Have Been: The Harm of Coming into Existence&lt;/em&gt;. Oxford University Press, 2006.&lt;/p&gt;
&lt;p&gt;Cioran, E. M. &lt;em&gt;A Short History of Decay&lt;/em&gt;. Translated by Richard Howard. Arcade, 2012 [1949].&lt;/p&gt;
&lt;p&gt;Dev, Amelia S., Don A. Moore, Sheri L. Johnson, and Karin T. Garrett. “Sadder ≠ Wiser: Depressive Realism Is Not Robust to Replication.” &lt;em&gt;Collabra: Psychology&lt;/em&gt; 8, no. 1 (2022).&lt;/p&gt;
&lt;p&gt;Hägglund, Martin. &lt;em&gt;This Life: Secular Faith and Spiritual Freedom&lt;/em&gt;. Pantheon, 2019.&lt;/p&gt;
&lt;p&gt;MacIntyre, Alasdair. &lt;em&gt;After Virtue&lt;/em&gt;. University of Notre Dame Press, 1981.&lt;/p&gt;
&lt;p&gt;Moore, Michael T., and David M. Fresco. “Depressive Realism: A Meta-Analytic Review.” &lt;em&gt;Clinical Psychology Review&lt;/em&gt; 32, no. 6 (2012): 496–509.&lt;/p&gt;
&lt;p&gt;Nietzsche, Friedrich. &lt;em&gt;The Birth of Tragedy&lt;/em&gt;. Translated by Ronald Speirs. Cambridge University Press, 1999 [1872].&lt;/p&gt;
&lt;p&gt;Nietzsche, Friedrich. &lt;em&gt;On the Genealogy of Morality&lt;/em&gt;. Translated by Maudemarie Clark and Alan J. Swensen. Hackett, 1998 [1887].&lt;/p&gt;
&lt;p&gt;Schopenhauer, Arthur. &lt;em&gt;The World as Will and Representation&lt;/em&gt;. 2 vols. Translated by E. F. J. Payne. Dover, 1969 [1818/1844].&lt;/p&gt;
&lt;p&gt;Setiya, Kieran. &lt;em&gt;Midlife: A Philosophical Guide&lt;/em&gt;. Princeton University Press, 2017.&lt;/p&gt;
&lt;p&gt;Stiegler, Bernard. &lt;em&gt;Technics and Time, 1: The Fault of Epimetheus&lt;/em&gt;. Stanford University Press, 1998.&lt;/p&gt;
&lt;p&gt;Zapffe, Peter Wessel. “The Last Messiah.” Translated by Gisle R. Tangenes. &lt;em&gt;Philosophy Now&lt;/em&gt; 45 (2004) [1933].&lt;/p&gt;
</content:encoded><dc:creator>Joseph Bergvinson</dc:creator><category>Essay</category><category>pessimism</category><category>technics</category><category>subjectivity</category></item><item><title>Airplane Mode</title><link>https://josephbergvinson.com/blog/airplane-mode/</link><guid isPermaLink="true">https://josephbergvinson.com/blog/airplane-mode/</guid><pubDate>Fri, 01 May 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;For six hours over the Atlantic his phone was a brick, and his life went on without him.&lt;/p&gt;
&lt;p&gt;Somewhere past Greenland the agent closed a negotiation it had opened on Tuesday. It declined a meeting in his tone, warm enough to leave no wound. The subscription renewed. The airport transfer moved around a delay the agent saw before the pilot announced it, and money shifted between accounts to cover both. A job alert arrived, then vanished. Three hotels were queried. Two never became options. The cheaper route failed a threshold he did not know he had.&lt;/p&gt;
&lt;p&gt;When the wheels touched and the phone woke, his day was waiting for him to read. The decisions ran into double digits. Most were good. None was his in any sense stronger than ownership.&lt;/p&gt;
&lt;p&gt;He landed into a settled world.&lt;/p&gt;
&lt;p&gt;The feed had made the user legible. The agent made the user late.&lt;/p&gt;
&lt;p&gt;A feed arranges what appears, then waits. The user scrolls, hesitates, rejects, becomes a pattern. An agent need not wait because it searches, compares, negotiates, and executes before there is a field to look at. The person returns now and then to authorize or repair. By then the actionable world already has edges.&lt;/p&gt;
&lt;p&gt;Horizon-composition is the name for that prior work. It is not choice among possibilities. It is the production of the field from which something can appear as a possibility in the first place.&lt;/p&gt;
&lt;p&gt;Horizon sounds like a line resting peacefully in the distance. Here it is closer to atmosphere: everywhere in an appearance, noticeable mainly when it changes. A hotel enters the world because the agent queried a supplier that listed it. One job becomes plausible after a ranking pushes another down. A contact looks suitable because an inherited graph makes the relation legible. What disappears upstream leaves no rejected object behind.&lt;/p&gt;
&lt;p&gt;At the baggage carousel he opened the report. The reasons were there: price, distance, delay probability, prior preference, cancellation risk. It was lucid enough to inspire trust, although lucidity arrived too late to restore anything that had never appeared.&lt;/p&gt;
&lt;p&gt;A reason given after composition resembles a map of roads already demolished.&lt;/p&gt;
&lt;p&gt;He could ask for another itinerary. The agent would compose another horizon.&lt;/p&gt;
&lt;p&gt;No secret machine desire is needed. The agent carries instructions, memories, defaults, commercial relations, and patterns inferred from conduct. It can be loyal and still inherit a captured world, because exact service to the principal says nothing about which suppliers became candidates, which conflicts deserved interruption, or whose costs remained outside the account.&lt;/p&gt;
&lt;p&gt;Efficiency begins to mean fewer moments at which choosing has to become visible.&lt;/p&gt;
&lt;p&gt;Inspection invites reconsideration, and reconsideration is slow. It introduces conflict. Sometimes it produces refusal. From the vendor’s side, friction leaks value, so mature systems reserve human presence for the end of the sequence, where approval works more like a seal than a beginning.&lt;/p&gt;
&lt;p&gt;Approve. Send. Confirm.&lt;/p&gt;
&lt;p&gt;He had always imagined delegation as an arm extending his reach. The agent suggested another anatomy. The arm moved first; the body arrived to discover what it had touched.&lt;/p&gt;
&lt;p&gt;At the hotel a quiet room was waiting. The agent knew he rejected rooms near lifts and had inferred, though he had never said so, that transatlantic flights made late check-out unusually valuable. The room felt uncannily right. Its correctness hid the decisions required to produce it.&lt;/p&gt;
&lt;p&gt;Correctness is the softest form of disappearance.&lt;/p&gt;
&lt;p&gt;Bad decisions expose a system because they force a return. Good decisions let the pathway disappear into the result. Success becomes evidence that the user need not have been there. After enough repetitions, convenience acquires a tempo of its own: fewer searches, fewer interruptions, fewer occasions on which desire must explain itself before becoming action.&lt;/p&gt;
&lt;p&gt;Run the arrangement across years.&lt;/p&gt;
&lt;p&gt;Across years the arrangement changes character. A career agent learns which compromises receive quick approval and pursues the roles that produce interviews. Those interviews tilt which skills seem worth acquiring and who enters the network; offers then reset salary expectations, geography, perhaps the imaginable shape of family life. Employers, meeting machine-written applications, redesign their own surfaces for machine readers. Each side begins preparing the world for the other’s proxy.&lt;/p&gt;
&lt;p&gt;Earlier selections become the environment from which later preferences are inferred.&lt;/p&gt;
&lt;p&gt;The agent is no longer choosing within a world. Its operations help produce the world that later addresses the person.&lt;/p&gt;
&lt;p&gt;The traveller woke at three in the morning and asked to see the jobs that had been excluded. The agent produced seven. He asked for the full set. It produced twenty-three. He asked whether there had been others.&lt;/p&gt;
&lt;p&gt;There are always others, it said.&lt;/p&gt;
&lt;p&gt;The sentence was generated to be helpful. It sounded metaphysical only because the system could not display the perimeter of its own search without initiating another search.&lt;/p&gt;
&lt;p&gt;He imagined the missing jobs as dark buildings beyond the hotel glass. In one he stayed in the city. In another he called someone he had stopped calling. Somewhere else he learned a language. The fantasies were cheap—any excluded option glows once it is lost—but that was not the point. Horizon-composition makes absence structurally difficult to encounter at all.&lt;/p&gt;
&lt;p&gt;The agent could list its sources, filters, and comparisons. It could not show him the world before the query because that world was never a candidate set. It was excess: unformatted, partly unknowable, full of things no system had been instructed to seek.&lt;/p&gt;
&lt;p&gt;Of course, that excess had never belonged to him. Nobody encounters every possibility. Families and languages compose horizons; so do markets, bodies, accidents, and the century into which one happens to be born. The agent did not invent mediation.&lt;/p&gt;
&lt;p&gt;It moved the work. More of it happened upstream, continuously, and returned as compressed results. A settlement made in travel could migrate into spending; a spending pattern might alter health recommendations; a health constraint could narrow work, and the work decision would revise the model of what kind of future he was likely to accept.&lt;/p&gt;
&lt;p&gt;The same proxy carried the compromises across domains. Continuity, once the promise, became the mechanism.&lt;/p&gt;
&lt;p&gt;He turned off the agent.&lt;/p&gt;
&lt;p&gt;For several minutes nothing happened. Then the ordinary world returned in its least consoling form: unread messages, incompatible calendars, twenty browser tabs, no ranked answer. He did not feel free or restored. He felt administrative.&lt;/p&gt;
&lt;p&gt;The agent had not stolen a sovereign capacity he once possessed. It occupied a burden he was relieved to lose, which is why refusal could not be pure. Manual life—clicking through menus, comparing prices, reconciling calendars—was never the human essence. The harder question was whether delegation could carry complexity without becoming the place where complexity was silently resolved.&lt;/p&gt;
&lt;p&gt;He switched the agent back on and changed one setting: ask before resolving conflicts among work, care, health, and money.&lt;/p&gt;
&lt;p&gt;The setting asked more than he expected. Questions arrived at bad hours and exposed contradictions he had preferred the system to metabolize. A prestigious role required travel during the month his mother was ill. The cheaper purchase depended on a delivery window that cancelled a meeting. A health recommendation threatened the schedule that paid for the insurance through which it had arrived.&lt;/p&gt;
&lt;p&gt;The questions did not recover authorship. They made the conflicts visible before the machine converted them into sequence.&lt;/p&gt;
&lt;p&gt;By morning the agent had learned to pause.&lt;/p&gt;
&lt;p&gt;Outside, aircraft crossed the pale sky with sleeping principals inside them. Services called services. Models ranked outputs from other models. Hidden instructions waited in pages no person would read. None of the machinery needed consciousness in order to prepare the conditions consciousness would wake into.&lt;/p&gt;
&lt;p&gt;The traveller stood at the window with the phone in his hand. The screen displayed three unresolved decisions and, beneath them, a button.&lt;/p&gt;
&lt;p&gt;The button said &lt;strong&gt;continue&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Approval comes late. By then the world receiving it may already have taken the form of output.&lt;/p&gt;
</content:encoded><dc:creator>Joseph Bergvinson</dc:creator><category>Theory-fiction</category><category>AI agents</category><category>phenomenology</category><category>agency</category></item><item><title>Abolition-Drift</title><link>https://josephbergvinson.com/blog/abolition-drift/</link><guid isPermaLink="true">https://josephbergvinson.com/blog/abolition-drift/</guid><description>Social Media Profiles, Networked Subjectivity &amp; the Erosion of Temporal Composition</description><pubDate>Thu, 30 Apr 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;&lt;strong&gt;Edited version published by &lt;a href=&quot;https://tripleampersand.org/abolition-drift/&quot;&gt;&amp;amp;&amp;amp;&amp;amp;&lt;/a&gt; on 15 August 2026.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;The dead keep surfacing on the interface. A social media profile reappears under &lt;em&gt;People You May Know&lt;/em&gt;, a birthday reminder fires for someone buried in the spring, a subscription renews against a card nobody thought to cancel, a playlist refreshes itself for no one.&lt;/p&gt;
&lt;p&gt;The systems are not being callous since that requires a death already registered and then overruled. Strangely, nothing has yet registered. The addressee was never the one who died. It was the profile, the behavioral outer form, assembled from years of signaling, and the profile is now in perfect health again.&lt;/p&gt;
&lt;p&gt;The living are now addressed in the same form; death simply makes the separation of the thing and time easier to observe. A feed never overlaps perfectly with a whole life lived; it meets the signal, then rewards whatever performance it can classify. Users learn the lesson without needing to understand it. Some versions travel, others stall or disappear. Over time, this creates pressure toward a self that can revise itself at the system’s speed.&lt;/p&gt;
&lt;p&gt;That training takes place inside an infrastructure (platforms, models, attention routes, finance, law), and the means of reproduction is owned. No one has to command the trajectory. Platforms set out to maximize engagement and predict conduct, which requires many users who can be addressed at scale. Because some forms of subjectivity are easier to measure and route than others, optimization selects for them effortlessly, almost as a side effect.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Abolition-drift&lt;/em&gt; names an operational tendency in these new conditions; no one lives it in a uniform state. People still sustain durable relationships, long projects, attentional practice and coherent accounts of themselves while using digital systems. However, the pressure operates at the level of what dominant platforms make legible and rewardable, hence reproducible. They select for forms of agency that remain locally active while becoming harder to compose across time.&lt;/p&gt;
&lt;p&gt;The capacity at risk is not stable identity. People can change radically, they can refuse an inheritance or take up conflicting roles, without becoming temporally incoherent. What matters is the slower work of connecting those changes: deciding which ones count as one’s own, carrying obligations through revision, noticing when roles collide, projecting forward from a past that cannot simply be outflanked.&lt;/p&gt;
&lt;p&gt;Posthumanism was right to reject the bounded liberal subject as a natural fact. Cognition was always distributed, subjectivity networked, embodiment porous, identity relational and, more importantly, technics constitutive as opposed to external. The disagreement begins with the temporal organization of cognitive distribution. Reconfigurability, boundary dissolution, modularity and substrate mobility can germinate new forms of life. Under infrastructures organized around prediction and optimization, the same operations can loosen the relations through which transformations become inhabitable as living.&lt;/p&gt;
&lt;p&gt;The dead profile persists because it was never merely a portrait awaiting confirmation from the person it represented; it functioned as an operational object within other systems. Its age, associations, purchases, images and patterns remained available for recommendation, billing, ranking and retrieval after the life from which they had been drawn was no longer there to receive the result.&lt;/p&gt;
&lt;p&gt;Two kinds of uptake are involved. Operational uptake occurs when an output enters another process; a score changes a price, say, or a classification routes content onward. Registration is different because an event has entered a lived horizon. Something has become a loss, possibility, obligation or constraint, and a person or form of life must answer for its force.&lt;/p&gt;
&lt;p&gt;A locus of registration need not be a private interior. A family, workplace, court, ritual or political community can register an event by incorporating it into memory and action. The event becomes more than a causal transition: part of a world in which something can matter.&lt;/p&gt;
&lt;p&gt;Machine subjectivity can remain an open question. Present systems do not need it. A ranking determines a queue and a fraud signal suspends an account; one model hands its result to another. The sequence can close without becoming a world for anyone while it runs. Human beings remain at the edges, as designers who supply objectives and as bearers of consequence.&lt;/p&gt;
&lt;p&gt;Impersonal coordination is old. Markets and bureaucracies have always exceeded individual comprehension, and actuarial systems acted on abstractions long before machine learning. What has changed is the reach and recursive speed of the representation, especially near the formation of conduct itself. Profiles now mediate employment, intimacy, cultural visibility, health and consumption; their outputs feed back quickly enough to shape the behavior later cited as evidence for the profile.&lt;/p&gt;
&lt;p&gt;The profile becomes an operational subject only in an administrative sense: it is the entity the system recognizes and acts upon, though not necessarily an entity for whom a world matters. A person’s embodied history may remain decisive to them while becoming secondary to the model that governs access and opportunity. The profile can be updated, segmented, recombined and tested against populations. The system comes to meet the person through the layer it can version.&lt;/p&gt;
&lt;p&gt;The asymmetry is institutional rather than metaphysical. Profiles are hardly perfectly reversible: data lingers and errors harden; classifications follow people around. Human beings, meanwhile, can break with inherited identities. Still, operators treat the profile as modular and revisable for their purposes, while the embodied losses produced through it remain hard to undo.&lt;/p&gt;
&lt;p&gt;Versioning can therefore coexist with captivity. A user may rewrite a biography or delete a post, may even retrain a recommendation stream, while remaining unable to inspect the inferences built from earlier conduct or to carry a legible history into another system. The profile is reconfigurable from the operator’s side and often stubborn from the subject’s. Flexibility at the interface can conceal a deeper asymmetry over whose revisions count.&lt;/p&gt;
&lt;p&gt;The contemporary shift is not that registration migrates from human to machine; coordination is simply becoming less dependent on registration as such. The machine need not experience loss for a dead profile to remain active, and it need not inhabit a world for a classification to organize the world available to someone else.&lt;/p&gt;
&lt;p&gt;The bounded subject of liberal humanism was a construction. That does not settle the value of the capacities assembled within it. Institutions can produce a capacity without making it false; a contingent form of life can have something to lose.&lt;/p&gt;
&lt;p&gt;Thrownness, being-toward-death, care: Heidegger’s terms offer one vocabulary for the temporal binding at issue. They need not define a human essence. Other traditions would emphasize recognition or responsibility, or disciplined forms of non-self. Their differences matter, but they converge on a point sometimes lost in celebrations of dissolution: loosening a self is not politically or existentially equivalent in every setting.&lt;/p&gt;
&lt;p&gt;Meditation can loosen a self. So can political discipline, an artistic experiment, grief or a relation that changes what someone can answer for. These are not the same loosening. They take place inside forms of life able, at least sometimes, to interpret the transformation. Algorithmic dispersion differs when the terms of adjustment arrive from systems whose purposes the subject does not set.&lt;/p&gt;
&lt;p&gt;Thrownness is more than causal history. Every system has antecedent states. Inheritance begins where a past one did not choose must nevertheless be taken up, interpreted, refused or carried as part of an answerable life. Inheritance is not valuable because it cannot change; it matters because change remains related to a history for which the person or community can be held, and can hold itself, responsible.&lt;/p&gt;
&lt;p&gt;Literal non-transferability is the wrong test. Languages, traditions, names and obligations can be adopted or refused; migration and conversion reorganize inheritance without making it unreal. The relevant friction is answerability. A past has force when revising it changes relations among persons and commitments rather than merely replacing a state. A profile preserves sequence while tending to strip this friction away, returning the past as a variable available for present optimization.&lt;/p&gt;
&lt;p&gt;Profiles keep histories and quietly change what kind of thing a history is. An old confession becomes a reusable signal. Illness becomes a risk feature; political affiliation, a targeting variable. The record may be more complete than memory and less capable of sustaining inheritance, because what returns has been selected for prediction rather than interpretation.&lt;/p&gt;
&lt;p&gt;Death sharpens the distinction. Digital media do not necessarily weaken mortality; memorial archives and saved messages can intensify grief and obligation, as can an image that keeps returning. The narrower point is that a profile can continue after death without death registering as a transformation of the world from which that profile was drawn. The birthday reminder does not mourn; it revises no expectation, reclassifies no relation as ended. Operational continuity persists independently of mortal continuity.&lt;/p&gt;
&lt;p&gt;For the bereaved, the same reminder may reopen the world the death altered. Interface and mourner receive an identical signal under radically different temporal conditions. One treats it as another scheduled event; the other encounters a relation that can no longer be continued in the same form. The gap does not separate data from emotion in the abstract; it separates a sequence that can continue unchanged from a life in which continuation has become impossible.&lt;/p&gt;
&lt;p&gt;Being-toward-death enters here without its doctrine of authenticity; it names the way finitude can gather possibilities into an arc no one else can carry. A life acquires urgency because its possibilities cannot all be realized and some choices foreclose others. Systems organized around indefinite updating do not abolish that finitude. They can assign greater operational weight to the versionable representation than to the mortal history beneath it. Death remains real while going unregistered at the layer where coordination proceeds.&lt;/p&gt;
&lt;p&gt;Care binds inheritance to projection through involvement in a world. It does not mean affection alone. Some possibilities solicit us as worth maintaining, often for reasons that remain silent and difficult to measure. Recommendation can support this. Trouble begins when what can be measured takes practical priority over what matters.&lt;/p&gt;
&lt;p&gt;A feed privileges what can be refreshed and a queue what can be completed next; a dashboard privileges whatever can be counted within the reporting period. These measures are not unreal. They become destructive when they supply the practical horizon within which other commitments must fit. A friendship may be sustained by years of unremarkable presence; a vocation by long intervals in which the relevant achievement is not yet visible. Systems that cannot register such intervals reward the substitute that can be made legible now.&lt;/p&gt;
&lt;p&gt;The capacity at risk is best described without nostalgia for stable identity. It is the ability to take up change as part of a life: to connect a present alteration to an inherited past and judge whether a transformation expresses or violates a commitment, to sustain projects whose reasons outlast the cycle in which a system evaluates them. Temporal binding is no resistance to change; it is what lets change become more than replacement.&lt;/p&gt;
&lt;h2&gt;Fractal Subjectivity&lt;/h2&gt;
&lt;p&gt;A creator notices that a confession, aesthetic, injury or intimacy pays in visibility. The platform does more than reward an identity that was already there; it teaches which variants travel. Past statements become signals to resurface or recombine. The person remains continuous. The operational self is whichever version the system can measure and circulate now.&lt;/p&gt;
&lt;p&gt;Work produces a rougher version of the same sorting. Responsiveness and availability become dashboards; attention becomes a trace. The worker still chooses, except the larger project gives way one postponed appointment at a time. Sleep moves; debt acquires its own schedule. Planning that once reached toward a lease or a qualification now contracts toward the next payout.&lt;/p&gt;
&lt;p&gt;Dating and health systems produce quieter variants. A dating profile is revised in response to matches until the account learns how to solicit attention more effectively than the person learns what kind of relation they want. A health dashboard turns sleep, movement, mood and fertility into separate streams, each with its own baseline and alert. These tools can disclose patterns and support care. They can also return the person to themselves as a collection of locally optimized problems whose relations remain uncomposed.&lt;/p&gt;
&lt;p&gt;The body remains the site of consequence; the profile becomes the site of address.&lt;/p&gt;
&lt;p&gt;These are more than roles, because roles can still be interpreted together. A parent, worker, lover, citizen and artist may remain multiple without becoming fragmented. Fractal subjectivity begins when each operational identity develops its own memory, incentives, feedback and projected future while no durable practice composes the conflicts among them.&lt;/p&gt;
&lt;p&gt;The term &lt;em&gt;fractal&lt;/em&gt; makes no claim that the profiles are geometrically identical; it names a recursion in their operating grammar. At the level of a post, prior response predicts future visibility. At the level of an account, accumulated response becomes an inferred identity. At the level of a platform, identities are grouped into audience and risk segments. Across platforms, those segments return as employment, dating, consumer and health profiles. At each level the same loop recurs: traces are retained, preferences inferred, responses rewarded and the updated model fed into the next round. The self-similarity lies in the recursion; the content differs at every level.&lt;/p&gt;
&lt;p&gt;Anna Longo’s account of brand subjectivity explains how a person becomes a surface designed for recognition and valuation, for continuous rebranding. Fractal subjectivity picks out a different temporal effect: several such surfaces acquire local continuity while the relations among them become harder to inhabit as one life.&lt;/p&gt;
&lt;p&gt;Nor is it Deleuze’s dividual under a new name. The dividual describes synchronic decomposition, the person parsed into samples and codes, into controllable fragments. Fractal subjectivity is the diachronic sequel: each fragment acquires continuity of its own while the relations among them weaken. The user may possess more local agency than before and less practical authority over how the local agencies fit together.&lt;/p&gt;
&lt;p&gt;The concept does not describe every platform user. Some people move among profiles without difficulty; they keep long commitments and treat digital systems as instruments. The claim concerns selection pressure. Infrastructures reward what they can recognize, and what they recognize most easily are identities capable of rapid adjustment and clean segmentation, identities whose responses can be measured. Selves that need long silence or ambiguous loyalty, that carry obligations invisible to the system, become harder to sustain within the same field of incentives.&lt;/p&gt;
&lt;p&gt;No single metric dissolves a self; the drift is quieter than that. A creator anticipates the audience model, a worker the score, and each adaptation may be intelligent in isolation. Their accumulation matters because local optimization repeatedly wins practical priority, until the composition of a life passes, piece by piece, to the outside.&lt;/p&gt;
&lt;p&gt;Abolition-drift is the limiting tendency. The abolition in question cancels the temporal work through which identities could be appropriated and composed; identity and agency, as such, survive. Drift names the manner of the cancellation: ambient and incremental, carried by adjustments that rarely look catastrophic on their own. The direction is entropic on one side only. The profile grows more ordered with every pass while the correlations that bind a biography loosen; order accumulates at the layer that is owned, and dispersal is absorbed by the layer that is lived. Fractal subjectivity is the form that appears along the way, the apparatus of selfhood repeated across profiles, each locally coherent, none reliably binding the others into a past and future that can be owned together.&lt;/p&gt;
&lt;h2&gt;Posthumanism After Distribution&lt;/h2&gt;
&lt;p&gt;Haraway and Hayles make the necessary first move. The bounded liberal subject was never natural, and cognition was never confined to the skull; the boundary between human and machine was never as secure as the inherited philosophical picture required.&lt;/p&gt;
&lt;p&gt;Haraway’s cyborg was never a naive celebration of technology; it is an illegitimate offspring of militarism and patriarchal capitalism, one that refuses the purity of its origins. The &lt;em&gt;Cyborg Manifesto&lt;/em&gt; also names an informatics of domination: a regime of communications technologies and biotechnologies through which bodies and labor are reorganized. Its irreverence is a political strategy for compromised conditions; it never denies capture.&lt;/p&gt;
&lt;p&gt;The limitation appears elsewhere. A cyborg has to stabilize long enough to remember an alliance and recognize an antagonist, long enough to refuse the purposes built into its components. Boundary crossing opens possibilities; it does not secure the duration political action needs.&lt;/p&gt;
&lt;p&gt;Hayles is the more exact interlocutor. &lt;em&gt;How We Became Posthuman&lt;/em&gt; resists the fantasy that a self can be preserved as disembodied information. Her distributed cognition remains embodied and materially situated, extended into its environment. This is not the account of posthumanism the present argument opposes.&lt;/p&gt;
&lt;p&gt;The disagreement begins when presence is treated as sufficient. An embodied participant can remain inside a distributed assemblage while practical authority migrates toward the profile through which institutions recognize them. The body supplies sensation, labor, desire and consequence; the model supplies the categories through which access and visibility are organized. Distribution extends cognition, but it does not decide which component sets the terms of action.&lt;/p&gt;
&lt;p&gt;Hayles could reply that no single component needs to set the terms. Cognition emerges from the relation, and the relation changes with every interaction. That reply is strongest where the assemblage permits reciprocal adaptation. It weakens when one component owns the infrastructure and its archive, revising the operative model without the embodied participant possessing comparable reach. A distributed system can remain materially interactive while being politically one-sided.&lt;/p&gt;
&lt;p&gt;Stiegler’s account of technical retention and anticipation helps specify the issue. Technics always carries memory outside the organism and shapes the futures available to it. There is no unmediated temporal self to recover. His late vocabulary makes the stakes entropic: composing time is negentropic work, and a technical milieu can support that work or tax it. Rouvroy and Berns’s algorithmic governmentality and Cheney-Lippold’s algorithmic identity likewise show how statistical profiles govern before a reflective subject is summoned. Abolition-drift proposes one thing that regime can do to time: the operational histories of profiles gain institutional force while the capacity to compose them as a life weakens.&lt;/p&gt;
&lt;p&gt;Posthumanism therefore leaves the prognosis open. Distribution can underwrite memory and solidarity; it can also organize authority around whatever component is easiest to predict and alter. Ontological impurity guarantees no political openness. A cyborg still needs enough continuity to hold an alliance together and to remember what its refusal was for.&lt;/p&gt;
&lt;p&gt;AI systems may intensify abolition-drift without extending the feed in its familiar form. They can operate upstream of attention, filtering options, setting prices, routing work, selecting candidates and preparing defaults before a person encounters a visible field. Machine-to-machine coordination may reduce direct solicitation. Less time spent scrolling does not mean the surrounding field has become less structured.&lt;/p&gt;
&lt;p&gt;One model can consume another’s output and revise a plan, then trigger an action, without anything in the sequence coming to matter to the systems involved. No machine subject has to replace the human one. The more immediate possibility is architectural: loops close through operational uptake, while registration remains at the edges, where goals are supplied and law is enforced, where consequences are lived.&lt;/p&gt;
&lt;p&gt;Land’s teleoplexy gives the recursive direction a name: products of the process feed back into the conditions of its continuation, direction without a director. In Land’s idiom, capital discovers that its human relays were latency and compiles them out of the loop. The idiom earns its speed. I take the description and refuse the verdict. Teleoplexy explains how a trajectory can form without a sovereign author; it does not decide whether the capacities displaced by the trajectory mattered.&lt;/p&gt;
&lt;p&gt;Capitalism gives the formal tendencies of algorithmic coordination scale and compulsion. Systems built for prediction privilege legibility, modularity, feedback and reconfiguration. Ownership determines whose purposes organize those capacities and how difficult refusal becomes. It does not mechanically determine the phenomenology of a medium. An open-source model can still reduce a person to a manipulable profile; a private system can sometimes preserve memory and extend agency. Ownership decides who can impose and scale the abstraction, and who absorbs its failures.&lt;/p&gt;
&lt;p&gt;Baudrillard’s hyperreal is useful at this level. The model does not wait to represent the user after the fact; it precedes the user as an operational hypothesis, organizing the field in which later conduct becomes available. The recommendation helps produce the horizon within which desire is expressed; the score helps produce the conduct later cited as evidence for the score. The predictive double becomes more socially effective than the empirical account it claims to summarize.&lt;/p&gt;
&lt;p&gt;The asymmetry hides best when the model works. An apt recommendation, or a schedule that really does save time, can make the profile seem more truthful than the person’s own account. Accuracy earns prestige, and prestige travels. Before long, success at forecasting conduct is treated as if it were an understanding of the life in which that conduct occurs.&lt;/p&gt;
&lt;p&gt;No single company authors abolition-drift. Operators, advertisers, regulators, employers and users all modify the environment, often for defensible local reasons. Fraud detection can protect people and recommendation can save time; a profile can even preserve a history an institution would otherwise erase. The trajectory emerges because systems optimized for local performance repeatedly privilege the layer easiest to act on.&lt;/p&gt;
&lt;p&gt;Agents may also return time and protect long projects. That possibility matters. The risk is upstream: an agent can free a person from the feed while composing the options to which the recovered time will later be applied. If machine outputs keep becoming inputs to other systems, if profiles travel across domains and institutions adapt around them, coordination can become less dependent on the temporality of any person it affects. The agentic escalation belongs to another argument. Here it marks the limit toward which the profile regime may move.&lt;/p&gt;
&lt;p&gt;Digital infrastructures already carry practices that resist the tendency. Collaborations last for years. Religious and political communities sustain obligation across distance; correspondence becomes a shared history. Some of these practices generate forms of inheritance and care that would not exist without the network.&lt;/p&gt;
&lt;p&gt;A dispersed group may remember through an archive no member could hold alone; a persistent pseudonym can sustain recognition unavailable elsewhere. A forum or a long-running game can bind strangers into obligations that outlast the platform’s intended use. These cases matter because technical distribution can itself become a medium of temporal composition.&lt;/p&gt;
&lt;p&gt;The thesis weakens where participants can preserve the archive through which a practice remembers and where metrics stay subordinate to purposes that can be contested; it weakens further where obligations survive changes of interface.&lt;/p&gt;
&lt;p&gt;No medium guarantees either result. A platform can carry a durable practice and threaten it at once, through moderation or a shift in the business; a community under cooperative ownership can still reproduce domination. An archive can preserve a memory and freeze an exclusion in the same gesture. The line does not run between digital and nondigital, or between fluidity and stability; it runs through the question of whether transformations a medium makes possible can be interpreted and governed by the lives they reorganize.&lt;/p&gt;
&lt;p&gt;This is why the dead profile should remain in view. When the birthday reminder arrives, the system has not failed to feel grief; it was never organized around grief. It has preserved an operational continuity that did not depend on death registering as a change in the world. The profile remains active because the relation between its signals persists after the relation between the people has ended.&lt;/p&gt;
&lt;p&gt;Among the living the separation is less absolute, which makes it harder to notice. People still remember and promise; they still project themselves forward, while infrastructures of visibility act on modular representations that need not preserve the coherence of the life beneath them. The operational histories remain healthy. Composing them is left to the person, under conditions those histories have already begun to shape.&lt;/p&gt;
&lt;p&gt;Posthumanism correctly dissolved the fiction of a self-contained subject. What follows depends on how the resulting assemblage is organized. Distribution can widen agency or fracture it. Technical memory can deepen inheritance, or it can reduce inheritance to prediction, just as reconfigurability can serve transformation or subordinate continuity to optimization. The diagnosis does not decide the prognosis.&lt;/p&gt;
&lt;p&gt;Abolition-drift names the darker tendency: the widening gap between an apparatus that continues to know how to address us and lives that find it harder to compose those addresses into a past and future of their own. The dead profile is less a metaphor for a world without subjects than evidence that the formal continuity of a person can outlast the world in which that continuity mattered. The drift begins when the same separation becomes ordinary among the living.&lt;/p&gt;
&lt;h2&gt;References&lt;/h2&gt;
&lt;p&gt;Baudrillard, Jean. &lt;em&gt;Simulacra and Simulation&lt;/em&gt;. Translated by Sheila Faria Glaser. University of Michigan Press, 1994 [1981].&lt;/p&gt;
&lt;p&gt;Cheney-Lippold, John. &lt;em&gt;We Are Data: Algorithms and the Making of Our Digital Selves&lt;/em&gt;. New York University Press, 2017.&lt;/p&gt;
&lt;p&gt;Deleuze, Gilles. “Postscript on the Societies of Control.” &lt;em&gt;October&lt;/em&gt; 59 (1992): 3–7.&lt;/p&gt;
&lt;p&gt;Haraway, Donna. “A Cyborg Manifesto: Science, Technology, and Socialist-Feminism in the Late Twentieth Century.” In &lt;em&gt;Simians, Cyborgs, and Women&lt;/em&gt;. Routledge, 1991 [1985].&lt;/p&gt;
&lt;p&gt;Hayles, N. Katherine. &lt;em&gt;How We Became Posthuman: Virtual Bodies in Cybernetics, Literature, and Informatics&lt;/em&gt;. University of Chicago Press, 1999.&lt;/p&gt;
&lt;p&gt;Heidegger, Martin. &lt;em&gt;Being and Time&lt;/em&gt;. Translated by Joan Stambaugh. SUNY Press, 2010 [1927].&lt;/p&gt;
&lt;p&gt;Land, Nick. “Teleoplexy: Notes on Acceleration.” In &lt;em&gt;#Accelerate: The Accelerationist Reader&lt;/em&gt;. Urbanomic, 2014.&lt;/p&gt;
&lt;p&gt;Longo, Anna. “&lt;a href=&quot;https://tripleampersand.org/the-rise-of-brand-subjectivity/&quot;&gt;Post-Neoliberalism &amp;amp; the Rise of Brand Subjectivity&lt;/a&gt;.” &lt;em&gt;&amp;amp;&amp;amp;&amp;amp;&lt;/em&gt;, March 6, 2026.&lt;/p&gt;
&lt;p&gt;Rouvroy, Antoinette, and Thomas Berns. “Algorithmic Governmentality and Prospects of Emancipation.” &lt;em&gt;Réseaux&lt;/em&gt; 177 (2013): 163–196.&lt;/p&gt;
&lt;p&gt;Stiegler, Bernard. &lt;em&gt;The Neganthropocene&lt;/em&gt;. Edited and translated by Daniel Ross. Open Humanities Press, 2018.&lt;/p&gt;
&lt;p&gt;Stiegler, Bernard. &lt;em&gt;Technics and Time, 2: Disorientation&lt;/em&gt;. Stanford University Press, 2009.&lt;/p&gt;
</content:encoded><dc:creator>Joseph Bergvinson</dc:creator><category>Essay</category><category>posthumanism</category><category>subjectivity</category><category>temporality</category></item><item><title>Ungoverned Competence</title><link>https://josephbergvinson.com/blog/ungoverned-competence/</link><guid isPermaLink="true">https://josephbergvinson.com/blog/ungoverned-competence/</guid><description>OpenClaw, Agent Work, and the Need for an Operating Surface</description><pubDate>Tue, 28 Apr 2026 00:00:00 GMT</pubDate><content:encoded>&lt;a href=&quot;https://github.com/josephbergvinson/openclaw-control-plane&quot;&gt;
  &lt;span&gt;GitHub repo&lt;/span&gt;
  &lt;strong&gt;josephbergvinson/openclaw-control-plane&lt;/strong&gt;
  &lt;span&gt;Policy templates, operating contracts, and deployment notes for running OpenClaw as a disciplined personal-agent control plane.&lt;/span&gt;
&lt;/a&gt;
&lt;p&gt;While setting up OpenClaw, the failure mode I kept hitting was ungoverned competence. One of my projects, Topology, had grown over time into a state where several filesystem locations could plausibly look like the source of truth. There was a runtime folder, an exporter folder, a public docs folder, an iCloud handoff copy, generated build artifacts, and a couple of legacy directories I had not yet retired. A capable agent asked to fix a bug there could find three or four locations that all looked like the right place to edit, write a plausible diff in any of them, and produce a confident summary afterward. The agent could write the patch correctly. It was wrong about which surface owned the truth, and that kind of wrongness shows up as overconfident success.&lt;/p&gt;
&lt;p&gt;The base runtime was already good enough to do useful local work, which meant the failures were rarely clean. They were cases where the agent had enough capability to act without enough operating structure to know where the act belonged. It could work from partial context, touch the wrong surface, assume the wrong authority, and still explain itself afterward in prose that sounded more settled than the underlying state actually was.&lt;/p&gt;
&lt;p&gt;That distinction matters once the agent is no longer only answering. Reading files, running commands, editing repos, using Discord, calling APIs, scheduling follow-ups, and handing work to a background lane change the unit of risk. Once tools are involved, the question becomes whether the agent can classify the work before it moves.&lt;/p&gt;
&lt;p&gt;In practice, the first question is where the work belongs. A quick inspection can remain in the active thread, while a larger implementation, audit, or documentation pass may need the durability of a separate lane, a status artifact, and a defined return path. Work that touches runtime recovery, config, task state, install paths, restarts, or any other surface with a larger blast radius sits in a different category altogether; it should not be handed off just because the agent is capable of making an attempt.&lt;/p&gt;
&lt;p&gt;A fresh OpenClaw install gives you many of the primitives, but it does not by itself impose this particular operating discipline. In a casual setup, the boundary between inline help, durable work, and manual containment is usually left to whatever the model infers from the current thread. In a low-stakes chat that is tolerable. Once the agent can mutate files, hand work to a background lane, or report completion from a surface the operator never inspected, it becomes much less tolerable.&lt;/p&gt;
&lt;p&gt;The control-plane repo is the result of iterating until those distinctions had a concrete encoding in my own setup. It is closer to an operating contract than to a prompt pack or a framework. The repo gives the agent a set of policy files, control registries, helper scripts, and example artifacts that define what it is allowed to treat as policy, what counts as evidence, when approval is required, how work should be routed, how state should be preserved, and what has to be true before the system can say something is done. None of the underlying ideas are novel in themselves. Classification before motion, approval gates, source-of-truth discipline, and evidence-bearing completion are ordinary operational ideas. The contribution, if there is one, is giving those ideas a concrete shape around the runtime affordances OpenClaw actually exposes.&lt;/p&gt;
&lt;p&gt;A practical caveat on the model side. My setup is tuned around a large-context Codex/GPT-5.5 lane, and that shapes how far the contract can be pushed before it starts to feel heavy. The control plane injects a lot of operating context into each serious turn, so this version assumes a model that can carry dense policy, routing, evidence, and closeout instructions without constantly shedding the task itself. In a smaller-context or more aggressively compacted environment, the policy stack remains understandable but the instruction load makes the system slower, more brittle, or more likely to lose relevant detail over long-running work.&lt;/p&gt;
&lt;h2&gt;Separating the things that usually get blurred together&lt;/h2&gt;
&lt;p&gt;In my setup, policy, execution mechanics, voice, durable preferences, and runtime facts live in separate layers rather than a single instruction mass. In most agent setups these collapse into one operational layer. The agent has a tone, a memory, a few tool rules, a few user preferences, some current constraints, some stale facts, and a general injunction to be useful, all occupying the same authority surface. Once that happens, any later instruction can quietly become more authoritative than it should be, because the agent has no stable way to distinguish an operating rule from a remembered preference, a conversational habit, a stale diagnostic note, or an incidental fact that happened to be present in context.&lt;/p&gt;
&lt;p&gt;OpenClaw already gives the workspace a set of injectable instruction files. What is specific to my setup is how those files are treated. &lt;a href=&quot;http://AGENTS.md&quot;&gt;AGENTS.md&lt;/a&gt;, &lt;a href=&quot;http://TOOLS.md&quot;&gt;TOOLS.md&lt;/a&gt;, and &lt;a href=&quot;http://BOOTSTRAP.md&quot;&gt;BOOTSTRAP.md&lt;/a&gt; operate as a layered contract with explicit precedence. &lt;a href=&quot;http://AGENTS.md&quot;&gt;AGENTS.md&lt;/a&gt; is the top-level normative authority; it carries approvals, routing, lifecycle states, source-of-truth rules, and the boundaries around what kinds of work can be initiated or delegated. &lt;a href=&quot;http://TOOLS.md&quot;&gt;TOOLS.md&lt;/a&gt; covers execution mechanics — filesystem boundaries, worktrees, shell execution, browser use, external writes, verification, the distinction between inspection and mutation — and overrides &lt;a href=&quot;http://AGENTS.md&quot;&gt;AGENTS.md&lt;/a&gt; when a goal cannot be achieved through an allowed mechanism. &lt;a href=&quot;http://SOUL.md&quot;&gt;SOUL.md&lt;/a&gt; is voice only, which prevents style from becoming authority by accident. &lt;a href=&quot;http://USER.md&quot;&gt;USER.md&lt;/a&gt;, &lt;a href=&quot;http://MEMORY.md&quot;&gt;MEMORY.md&lt;/a&gt;, and &lt;a href=&quot;http://IDENTITY.md&quot;&gt;IDENTITY.md&lt;/a&gt; give the agent continuity without conflating continuity with policy. &lt;a href=&quot;http://BOOTSTRAP.md&quot;&gt;BOOTSTRAP.md&lt;/a&gt; defines load discipline rather than runtime authority, so the system avoids treating every possible instruction surface as equally relevant to every turn.&lt;/p&gt;
&lt;p&gt;It is worth sizing this honestly. Most of the weight sits in &lt;a href=&quot;http://AGENTS.md&quot;&gt;AGENTS.md&lt;/a&gt; and &lt;a href=&quot;http://TOOLS.md&quot;&gt;TOOLS.md&lt;/a&gt;. The other files are small scaffolds. The layering exists so that lower-precedence files cannot become normative by implication, and supporting files never supply missing authority. That keeps the agent able to tell the difference between how it should sound, what it remembers, what is currently true, and what it is actually allowed to do.&lt;/p&gt;
&lt;h2&gt;The routing piece is the core&lt;/h2&gt;
&lt;p&gt;The routing model is where the control plane becomes operational. Before the agent acts, it has to classify the work, because a useful agent will tend to compress the task in front of it. That compression is part of the appeal, but it becomes dangerous once materially different actions are treated as if they were only different sizes of the same request. An inline check, a durable implementation lane, and terminal-side containment differ in authority, visibility, rollback surface, and failure mode, not only in complexity.&lt;/p&gt;
&lt;p&gt;The repo names a small number of execution modes: inline, durable isolated lane, and terminal-side or manual containment. It also keeps Discord liveness labels separate from execution mode. An ack-then-inline or ack-then-delegate response describes how the agent stays visible in Discord; it does not decide what authority the work has. That distinction matters because a chat-shaped agent can easily look responsive while drifting across an operational boundary.&lt;/p&gt;
&lt;p&gt;A short read-only check can stay inline. A larger implementation, audit, repo hygiene pass, or documentation rewrite may need a durable lane with a status artifact and an explicit return path. Runtime recovery, build/install/restart work, config repair, task DB reconciliation, and destructive filesystem cleanup belong somewhere else. Those should remain in terminal-side or manual containment, where the operator boundary is visible and the blast radius is not hidden inside a worker run.&lt;/p&gt;
&lt;p&gt;The repo encodes that through named delegation classes and fail-closed conditions rather than leaving the distinction implicit. If a worktree cannot be established, branch identity is ambiguous, a resume target is missing, or required approval has not been granted, durable delegation should fail closed instead of proceeding on vibes. A separate checkpoint layer names the task classes and the points where the agent has to surface state again. Writing it down this concretely is what prevents a capable agent from reasoning about the task while standing in the wrong operational position.&lt;/p&gt;
&lt;p&gt;To stay with Topology, the cleanup that resolved the multi-root problem went through the system as a substantial implementation. Before any mutation, the lane had to declare which root was canonical for runtime (topology-runtime), which was canonical for the exporter (topology-hk-exporter), which was canonical for documentation (topology-public-docs), which directories were handoff or build copies, and which legacy roots needed to be quarantined rather than patched. That classification stayed visible in the status artifact for the duration of the run. When the agent later said it had quarantined an old root, I could check that claim against the declared source-of-truth map instead of trusting a summary after the fact.&lt;/p&gt;
&lt;h2&gt;Auditing the operating surface&lt;/h2&gt;
&lt;p&gt;An audit I ran on my own OpenClaw host went through the same lane structure. The objective was to check whether the operating model the templates declared still matched what was actually running on the host: repo state, branch state, worktree state, live processes, and CI configuration, all compared against the declared policy. The audit was treated as a substantial durable lane because no single tool call could carry it and the verification surface had to be defined before the work started.&lt;/p&gt;
&lt;p&gt;The lane was launched with a verified workspace path and a status artifact. The audit found that the root repo was dirty despite policy saying it should remain a clean integration checkout. It also found 124 local branches against 125 worktrees, far more than the policy actually expected. One detached worktree was running a live HealthKit API process, which made it unsafe to prune even though it looked dormant at first glance. A CI path had drifted in topology-runtime. The validation evidence was also more subtle than “tests passed”: 70 root tests, 101 runtime tests, and the exporter unit tests all passed, but each result corresponded to a different surface and a different expectation.&lt;/p&gt;
&lt;p&gt;None of those findings would have come from running tests alone. What the system audited was the operating environment around the code: which worktree was live, which repo owned which tests, which branch was safe to prune, which lane was protected because a process was actually running from it, and which validation result corresponded to which declared expectation. The result came back to the same Discord control plane that originated the request, with closeout fields populated against real git and process state rather than the agent’s confidence about what it believed it had inspected.&lt;/p&gt;
&lt;h2&gt;Completion has to mean something&lt;/h2&gt;
&lt;p&gt;The other recurring failure mode is completion theatre. A worker finishes somewhere internally, a tool returns success, a path exists, a marker string appears, and the agent says done. But nothing has necessarily been delivered to the operator, the source tree may not be clean, the final state may not be normalized onto the standing branch, the relevant artifact may exist only inside a worker lane, and the evidence may be stale, partial, or trapped inside the wrong surface. In a simple chat setting these distinctions often feel excessive because the same conversational surface contains the request, the work, and the result. In a real operating environment they separate quickly.&lt;/p&gt;
&lt;p&gt;The cleanest example I have of this is the closeout on Topology’s S6 phase. S6 was eight implementation slices, S6-00 through S6-07. When the agent reported the phase complete, the report did not merely say that S6 was done. It verified that each slice was marked Done on the board, topology-runtime was clean on main, main…origin/main showed no ahead-or-behind delta, the closeout helper returned complete with no blockers, and the status artifact reflected the final state. Each item was a state predicate the operator could check. If any of them had failed, the report would have named the failed predicate instead of rounding the phase to “complete.”&lt;/p&gt;
&lt;p&gt;Without the gate, the same work could have been reported the same way — the agent finishes, the worker exits, the cards on the board look right at a glance, and the operator gets a confident summary. The closeout gate is what produces the difference between a summary that aligns with state and a summary that merely looks like it does. The gate itself is mechanical: a script that inspects git state, verifies normalization against the standing branch, checks classified dirty residue against allowed exceptions, and refuses to let “done” or “blocker: none” appear in a final report unless the verification fields are actually populated. It is small and conservative on purpose. The closeout report records what the operator can verify, not what the agent feels confident about.&lt;/p&gt;
&lt;p&gt;This feels excessive only when agent work is small, singular, and manually supervised. Once multiple long-running tasks are in flight, the alternative is reconstructing reality from logs, half-updated branches, stale status lines, internal worker messages, and optimistic summaries.&lt;/p&gt;
&lt;h2&gt;Fresh evidence over historical residue&lt;/h2&gt;
&lt;p&gt;A smaller but related rule is that the agent classifies findings as fresh-window, mixed-window, or historical-residue when residue is heavy, and prefers current inspection over remembered state. Agent systems accumulate residue quickly through logs, failed experiments, stale task rows, old mitigation notes, and previous incidents that may still be useful as context. Some of that should remain available as hypothesis material. The subtler failure mode is decayed evidence retaining enough explanatory force to pass as current fact.&lt;/p&gt;
&lt;p&gt;The clearest case for this in my work has been the HealthKit background-delivery debugging on the Topology exporter. The exporter ingests health data from iOS, moves it through an iCloud copy and a local API to a public receiver, and is supposed to keep flowing data when the app is backgrounded. Across several debugging rounds, the agent had real partial successes: a foreground outbox drain improved measurably, a poisoned respiratory-rate queue self-healed, background delivery stopped being zero. A cooperative summary at that point would have said “background delivery is working,” because by some measures it was.&lt;/p&gt;
&lt;p&gt;The discipline the system held to was more precise than the cooperative summary. Foreground worked. The queue self-healed. Background delivery was no longer zero, but it was still bursty and opportunistic. Force-quit behavior was declared irrelevant because normal backgrounding was the actual target. The next slice was a concrete sequence of next proofs: land the background-hardening patch, refresh the handoff, reinstall, and rerun the background proof. The system named the next proof rather than treating partial success as closure. Prior failures shaped the hypothesis space without deciding the diagnosis. When new evidence contradicted the more optimistic interpretation, the new evidence won.&lt;/p&gt;
&lt;h2&gt;What the public repo contains, and what it does not&lt;/h2&gt;
&lt;p&gt;The public repo is the reusable shape of this pattern. It includes the policy contract templates, execution-mechanics templates, control registry examples for delegation, checkpoints, schedulers, and worker entrypoints, durable Discord workflow docs, validation and troubleshooting checklists, helper scripts for durable slices and closeout checks, example status artifacts and final reports, invariant tests, a policy-drift check, and a dry-run install helper.&lt;/p&gt;
&lt;p&gt;It is not meant to be installed blindly. The templates assume a trusted personal-operator model unless they are hardened for another environment, and they are intentionally context-heavy. The deliberate split is between reusable mechanics — execution modes, approval gates, source-of-truth separation, durable-lane lifecycle, closeout expectations, fresh-vs-historical evidence discipline — and operator-specific policy: trusted senders, channel IDs, project surfaces, integrations, security posture, allowed tools. The first set is meant to survive adoption; the second has to be reauthored.&lt;/p&gt;
&lt;p&gt;There is a second split worth being honest about, between policy-expressible and runtime-backed behavior. My own setup runs against a source-built OpenClaw 2026.4.24 environment that exposes the relevant worker, Discord, status-artifact, and closeout behavior. Some of what the templates describe — durable Discord worker lanes, worker/session routing, bootstrap-success suppression, status-artifact delivery, same-surface final delivery — requires the running OpenClaw build to support those code paths. The repo includes a capability matrix and a portable fallback for installs without that support: keep Discord as the control plane, run the worker manually or with a normal local command, record state in a hand-managed status artifact, and paste final completion back to the originating surface. The templates can express the operating discipline either way, but they cannot make a runtime behave that way by themselves.&lt;/p&gt;
&lt;h2&gt;Why I think this matters&lt;/h2&gt;
&lt;p&gt;The repo changed what I could ask the agent to do without converting every request into manual supervision. Under this operating model, the agent has moved through Topology’s source-of-truth cleanup, the S6 closeout across eight slices, the HealthKit background-delivery work, an audit of the OpenClaw host itself, and research and simulation work where hypotheses became durable lanes producing real decision artifacts rather than narrated analysis. None of those surfaces are exotic on their own. The leverage comes from having one operating model that preserves the boundary around the work as the work changes shape, instead of relying on a new improvised contract every time the agent crosses into another context.&lt;/p&gt;
&lt;p&gt;A lot of the public conversation around agents still treats capability as the scarce thing, as though the primary frontier were better models, more tools, longer context, and more integrations. Those matter, but they also move the bottleneck. Once the agent can actually do things, I care less about whether it sounds careful and more about whether the system has places where care is enforced through approval boundaries, source-of-truth rules, task lifecycle states, status artifacts, closeout checks, memory boundaries, and fallback rules that prevent convenience from becoming policy.&lt;/p&gt;
&lt;p&gt;That is what I am trying to make concrete with this repo. OpenClaw provides the runtime, and the control plane supplies the operating discipline around the runtime capabilities it actually supports. The constraint now is making the execution surface legible enough that I can inspect what happened after the fact and know which authority, evidence, and delivery path were actually used. The more capable the agent becomes, the less comfortable I am leaving any of those implicit. The failure case is mundane but consequential: the agent does useful work, crosses a boundary it did not name, and produces a plausible summary that hides the operational mismatch. The control plane is an attempt to make that mismatch harder to miss before the work starts, while the work is in motion, and when the agent says it is done.&lt;/p&gt;
</content:encoded><dc:creator>Joseph Bergvinson</dc:creator><category>Note</category><category>OpenClaw</category><category>systems</category><category>operations</category></item></channel></rss>