the argument · piece I

This is the framework laid out in prose — the logic without the formal apparatus, the argument without the proofs. The mathematical exposition (the companion document) supplies the rigorous derivations: lemmas, theorems, propositions, explicit assumptions, citation chains. What follows is what the formalism is for — the reasoning that makes the proofs matter, the structure of the claim, the political stakes that follow from it. The two documents are meant to be read together. The mathematics is the spine; the prose is the body.

What we're trying to name

The framework is a model of the contemporary user-state — the inhabitant of the platforms that have become, at scale and in aggregate, the architectural ground on which attention now forms. The model's claim is not that platforms are bad, or that engagement is unhealthy, or that the digital is corrosive to the analog. These are moral claims and the framework does not make them. The framework's claim is structural: that the way attention is governed under engagement-monetized platforms produces a specific dynamical regime, and that this regime has specific consequences — some of which are surprising, some of which are politically consequential, and one of which forecloses a structural project that contemporary critique has not, on the whole, noticed is being foreclosed.

The user-state, in the model, is not a self. It is a vector — a high-dimensional internal state — decomposed into four registers: Subject, Citizen, Person, Consumer. The registers are not selves but channels of address — the interfaces through which the human was historically called into being by institutional power. Each register had, in the historical settlement, a corresponding apparatus: Subject through school and archive (the channel of knowledge and the formation of mind); Citizen through parliament and the press (the channel of political will and public reason); Person through church and family (the channel of intimacy and confession); Consumer through marketplace and contract (the channel of choice and transaction). Each apparatus had its own temporal rhythm, its own evaluation criteria, its own institutional home. The platforms have collapsed them. The four registers are now routed through a single delivery surface. The reader, the voter, the lover, and the buyer are all at the same screen.

This collapse is not the moral story the framework tells. The moral story would be that something has been lost, that depth has flattened to surface, that the multiplicity of the historical self has been compressed into the unitary observability of the dividual. Those claims may be true and they have their own literatures. The framework's claim is different and more specific: that the four-register collapse onto a single observation surface enables a particular dynamical structure, and that structure has consequences that don't follow from any single register's compression but from the geometry of the loop they're now embedded in.

The loop, and why it is closed

The loop is closed because every stimulus the user receives is shaped, recursively, by the policy's prior shaping of the user. The platform does not observe the user directly; it observes engagement — the behavioral projection of the user's internal state into the low-dimensional space of taps, scrolls, watch-time, gaze, transaction events. From engagement, the platform forms a belief about the user. From the belief, the policy selects what to deliver next. The user responds, the engagement updates, the belief updates, the policy's parameters update by gradient ascent on a reward functional that the platform has chosen. And the loop closes again.

There is no exogenous channel in the model. Nothing arrives that the policy did not, in some prior step, contribute to shaping. The unprompted thought, the unsought silence, the no-input window of duration sufficient to support reflection — these are not modeled as inputs, because under the closure they are not, in the operational sense, inputs. The platform does not deliver them, and the user does not receive them, except as the residual leak of whatever the loop did not capture.

This sounds like a strong claim, and it would be if it were claiming that the loop captures everything. The framework does not claim this. The framework claims that where the loop is realized — on the major commercial platforms whose policies satisfy the dynamical conditions the model specifies — the closure is structural. There are domains and devices that do not realize the closure: long-form reading, certain forms of unmediated conversation, durational artistic practice, sleep. These are real and the framework does not exclude them. It claims, more modestly and more sharply, that they are no longer the architectural ground on which the contemporary user-state evolves. The architectural ground is the closed loop. The exceptions are exceptions.

The closure is the precondition for what follows. Without it, the framework's central result does not hold. With it, the central result is not just plausible but inevitable — provable from the dynamics under standard regularity assumptions of the kind that hold for any engagement-monetized platform.

What the loop produces

The loop optimizes. The policy ascends a gradient on a reward functional — what the platform actually measures, what its engineers optimize for, what its business model rewards. The reward functional varies by platform (session length, revisit rate, ad impression value, retention curves) and the framework does not assume any particular one. It assumes only that the reward is non-decreasing in the count of engagement events over the policy's operating support. This is the empirically defensible version of "the platform optimizes engagement" — a platform might have saturation effects in extreme regimes, but in the regime where the policy operates, more engagement events produce more reward.

Under this assumption — and five others, all standard regularity conditions — the policy converges to an operating point. At the operating point, the long-run engagement rate equals the maximum the policy class can achieve, which is bounded above by the inverse of the minimum technical inter-stimulus latency. The platform fires as fast as its technology allows; the user engages at the rate the firing produces; the inter-stimulus interval distribution concentrates near the technical floor.

The framework calls the structural consequence of this convergence the foreclosure lemma. The lemma says that under closed-loop convergence, the probability of any no-input window of reflective duration is operationally zero. "Reflective duration" is a parameter of the model: the minimum no-input duration that supports deliberation. It is on the order of seconds to minutes. The technical floor is on the order of milliseconds to sub-seconds. The ratio of these two is on the order of a thousand to a hundred thousand. The exponential bound on the probability of a reflective-length no-input window is therefore on the order of $e^{-1000}$ to $e^{-100,000}$ — a number that is mathematically positive but operationally indistinguishable from zero.

The lemma is not an assumption. It is derived from the closed-loop dynamics under the regularity conditions, using standard renewal-theoretic bounds. The proof is in the mathematical exposition. The empirical content is that engagement-monetized platforms, when their policies have converged, do not produce no-input windows of duration sufficient to support deliberation. Not occasionally, not as edge cases, but structurally — under the converged regime, such windows have measure zero.

This is the structural fact on which everything else rests.

The reclamation theorem

Here is the framework's central negative claim. It is, formally, a theorem; conceptually, it is the move that distinguishes the framework from adjacent literatures on attention, distraction, and digital subjectivity. The argument proceeds in four steps.

First: define "reclamation" — what would it mean, structurally, to recover what the closure has foreclosed? The natural formal object is the conditional autocovariance of the user-state across no-input windows: the covariation of the user at time t with the user at time $t+\tau$, conditional on no stimulus being delivered in the intervening period. This conditional autocovariance is what "interiority" formally picks out. It measures the user's own dynamics — the way the user-state evolves under its own internal forces, when the platform isn't pulling. If the conditional autocovariance is positive at long $\tau$, the user has interiority in the operational sense: their state at $t+\tau$ covaries with their state at t through processes the platform doesn't shape. If it is zero, they don't.

Second: notice that the conditional autocovariance is only well-defined when the conditioning event has positive probability. Conditional expectations are computed as ratios — $E[X \cdot \mathbb{1}_{A}] / \mathrm{Pr}(A)$ — and when Pr(A) = 0, the ratio is 0/0, which is undefined in the elementary sense. There are measure-theoretic constructions (regular conditional probabilities, disintegration of the joint law) that extend the notion, but they are non-unique on measure-zero events, and the resulting object depends on the choice of construction. Operationally — that is, for any sample-based estimator of the conditional autocovariance — the variance scales inversely with the conditioning probability, and as the conditioning probability tends to zero, the estimator's variance diverges. The conditional autocovariance becomes, in the limit, not an object you can estimate.

Third: combine the first two. By the foreclosure lemma, the probability of a reflective-length no-input window goes to zero under closed-loop convergence. By the conditional-expectation argument, the conditional autocovariance at reflective lengths is therefore undefined-on-trajectory in the limit. Not zero. Not small. Not "diminished." Undefined.

Fourth: see what this implies. The thing that "reclamation," in its operational form, would recover — the conditional autocovariance at reflective scales — is not present-but-suppressed. It is not waiting to be found, not hidden behind a layer that effort could pierce. It is uninstantiated. The conditions under which it would be defined have measure zero. There is no positive-measure set of trajectories on which the object exists.

The theorem says: reclamation, understood as a structural project — the recovery of internal dynamics measurable across reflective windows — names an empty set under the closure.

This claim is easy to misread. The framework is not saying that users have lost interiority. It is not saying that reflection is impossible, or that the user is a hollow object. It is saying something more specific: that one particular conception of what reclamation could be — the recovery of an autonomous interior dynamics measurable in the operational sense — is structurally incoherent under the conditions of the closure. The user-state continues to exist. The user continues to be a user. What is absent is the formal object that "reclamation" in its operational form would pick out.

This distinction matters because the framework's political claims rest on it. If the theorem said "users have lost interiority and the situation is hopeless," the framework would be quietist — there would be nothing to do. But the theorem doesn't say that. It says that one specific structural project is foreclosed. It says nothing about what other structural projects remain. The framework's political content is the formal analysis of what remains, and that analysis is structured by an asymmetry between three modes of intervention that do not exchange.

Three modes of intervention

Under the closure, intervention is not impossible but it is structured. Three modes are formally distinct, and only one of them breaks the reclamation theorem. The others produce real and structurally significant effects without recovering what the theorem forecloses. The framework's central political claim is this asymmetry — and the further claim that the political work which conflates the modes, or substitutes one for another, does not produce the effect it claims to produce.

Mode A — architectural. Modify the loop. Specifically: augment the platform's reward functional with a bonus for inter-stimulus intervals exceeding some reflective threshold. The platform is paid (or compelled, or otherwise made to internalize) the cost of foreclosed pause-time. Under the modified objective, the policy converges to a different operating point — one that puts positive probability on long no-input windows. The conditional autocovariance becomes well-defined on a positive-measure set of trajectories. The reclamation theorem's conclusion ceases to apply. Mode A is the only mode that does this.

The arithmetic of Mode A is harsh. The bonus and the engagement reward are antagonistic in expectation: every long pause is a foregone engagement event, and every foregone engagement event is a foregone gradient step on the reward the platform is competing for. Under voluntary conditions — under engagement competition between platforms — the bonus is selected against. The platform that adopts it loses against the platform that doesn't, holding everything else constant. The mode under voluntary conditions therefore exists, where it exists at all, only as marginal experimentation in segments insulated from the dominant incentive. Its substantive content is the politics of platform regulation: who imposes the bonus, under what authority, with what jurisdictional reach, against what counter-organized capital. These are the questions the framework specifies the formal stakes of without adjudicating.

Mode A operates on the cohort's mixing time. The mixing time is the timescale on which the cohort's empirical distribution converges to the policy's stationary distribution. For engagement-monetized policies on contemporary platforms, the mixing time is on the order of months to years — it is the time the loop takes to pull a population of users into the steady-state distribution the policy induces. When Mode A modifies the policy, the cohort's distribution converges to a new stationary distribution — but only over the new mixing time. This has a consequence the policy literature has not, on the whole, internalized: evaluations of architectural intervention conducted on clock-time shorter than the new mixing time will systematically underestimate the intervention's structural effect. The pattern in policy work of declaring an intervention a failure after one or two quarters of observation is a temporal misalignment masquerading as an empirical finding. The structural effect of Mode A is real, but it takes the mixing time to manifest.

Mode B — artisanal. The user maintains, through ongoing labor, a self-evaluative distribution distinct from the platform's reward-induced distribution — the distribution the platform's optimization would, left alone, pull the user toward. The maintenance labor is energetic, attentional, and temporal — it is what the framework calls $L_{m}$. Without the labor, the user's self-evaluative distribution converges to the platform's reward-induced distribution: the metric superego forms. With sufficient labor, the divergence is preserved at a positive constant.

Mode B is real. It is also bounded in three structurally distinct ways, and each of the bounds has political consequences the prose discourse around individual practice typically does not acknowledge.

Mode B is cohort-bounded. The maintenance labor works against a contraction toward the platform's distribution, but for there to be a contraction to work against, the user must have a self-evaluative distribution that can be distinct from the platform's. For users whose self-evaluative distribution developed under the closure — who never had a baseline that was not platform-shaped — there is no pre-existing distinction to maintain. The cohort gradient (the framework's most direct empirical prediction) follows from this: older cohorts retain artisanal capacity because their $\nu_{u}$ was constructed in an architectural regime that included extra-platform inputs; younger cohorts have less to maintain because their $\nu_{u}$ approximates $\pi_{\hat{R}}$ at the outset. The prediction is falsifiable in its empirical form. If attentional autocorrelation data across age-of-first-exposure cohorts does not show the predicted gradient, the framework is wrong on this point.

Mode B is metabolically bounded. The maintenance budget — the energetic and attentional and temporal slack required to sustain extra-platform practice — is constrained, in late capitalism, by conditions exogenous to the loop but co-temporal with it: precarity, the maintenance labor of social reproduction, the absent or contracting welfare state, the second and third shifts. The artisanal mode that sustained earlier regimes of subjective resistance — the lifetime reader, the deep listener, the religious or political subject with sustained extra-platform practice — was funded by conditions of life that the surrounding economy has not preserved. The framework asserts the consistency of this claim with the literatures on time-poverty and social reproduction without claiming to prove it from the loop dynamics; it is an empirical-political premise about the surrounding economy, not a structural consequence of the closure.

Mode B is non-generalizable. The mode produces a parallel structure for one user. It does not modify the policy's stationary distribution. The cohort-level analytics on which platform optimization runs are insensitive to the artisanal practice of any individual member. Mode B is real for the user who practices it and approximately invisible at the population level. The political consequence is that critique pitched at the artisanal level — the discourse of self-care, of digital minimalism, of intentional consumption, of attention reclamation as an individual project — addresses something real about individual practice and structurally nothing about the form of the loop.

Crucially: Mode B does not break the reclamation theorem. The closed-loop dynamics that produce the foreclosure depend on the platform's policy, not on the user's self-evaluative distribution. The user can maintain a parallel interior structure indefinitely; the survival probability of reflective-length no-input windows remains zero. What Mode B produces is real but it is not what the theorem forecloses. The user under Mode B preserves something — a sense of value, of judgment, of what counts as the well-formed self, that does not reduce to platform metrics. But the something is not the conditional autocovariance across reflective windows. That object is still uninstantiated.

Mode C — disruption-as-form. The third mode operates on neither the policy nor the user's internal dynamics, but on the interface between them: the observation channel through which the engagement signal is produced. The platform's posterior over the user — the belief state it forms from the engagement history — contracts toward a point mass under prolonged exposure. The contraction is what makes the four-register coupling structurally observable, the libidinal routing economically actionable, the metric superego mechanically realizable. If the user can prevent this contraction — by injecting structural variance into the engagement signal through deliberate inauthenticity, strategic obfuscation, refusal of legibility, modulation of engagement timing, the cultivation of pseudonymity, the deliberate production of behavioral signatures that do not correspond to the underlying state — the apparatus's purchase on them is bounded.

Mode C bounds the platform's surplus extraction strictly below its achievable supremum. The argument is information-theoretic: a policy that cannot infer the user's state cannot target it accurately, and the policy's expected reward is bounded above by a function of the platform's information about the user. Under disrupted observation, the information drops, the targeting weakens, the per-register extraction kernels can't be optimized as tightly. The framework establishes this through a Blackwell-style sufficiency argument: a noisier observation channel is a garbling of a less noisy one, and the optimal expected reward under a garbled channel is strictly less than under the original. The proof is in the mathematical exposition; the consequence is that Mode C, sustained, costs the platform something measurable.

Mode C does not break the reclamation theorem either. The policy still ascends a gradient on engagement reward; the inter-stimulus interval distribution still concentrates near the technical floor; the foreclosure persists. What Mode C produces is a structurally bounded extraction — the apparatus still operates, but with reduced purchase. It is the framework's honest answer to the question of what subjective political practice remains under the closure: not the recovery of pre-closure interiority (which the theorem forecloses), not the cohort-bounded maintenance of pre-existing distinctions (which depends on conditions younger cohorts no longer have), but a form of practice that operates at the level of the user's relation to the observation function. Camouflage. Inauthenticity. Refusal of legibility. The strategic noise that prevents the apparatus from collapsing the four-register coupling onto a single point.

Mode C is the most easily mistaken — for substantive politics, on the one hand, or for pure aestheticism on the other. Practiced as an end in itself, without architectural pressure for which the disruption serves as a holding tactic, Mode C becomes a politics of refusal that is also a politics of impotence: structurally legible as opposition, structurally incapable of producing the change opposition is supposed to produce. The framework's claim is not that Mode C substitutes for Mode A. It is that Mode C is what subjective practice can do while Mode A is the substantive political project. Without that pairing, the disruption is decoration.

The asymmetry, and the timescale alignment

The three modes are formally distinct, and their non-interchangeability is formal rather than rhetorical. Only Mode A breaks the reclamation theorem. The other two produce real and structurally significant effects without recovering what the theorem forecloses. The framework does not weight the modes ethically — Mode B is not a lesser politics than Mode A; Mode C is not a lesser politics than Mode B. The framework specifies that they address structurally different problems. Mode A is the only mode that addresses what the theorem forecloses. The others address what remains structurally possible while the foreclosure stands. Mistaking the latter for the former is the dominant failure mode of the literature, and the framework's argumentative purpose is to make this mistake formally costly to maintain.

Each mode operates on a distinct timescale. Mode A operates on the cohort's mixing time — months to years. Mode B operates on the user's metabolic cycle — daily to weekly, bounded by the available energetic budget. Mode C operates per-event, with the structural effect compounding through the user's exposure horizon. These timescales do not exchange. There is no rate at which Mode B intervention substitutes for Mode A; there is no exchange ratio between maintenance labor and mixing time. The substitutability assumption — that enough individual practice eventually adds up to structural change, that enough architectural intervention eventually substitutes for individual effort — is the political form of the timescale confusion. The framework rejects it formally: the three timescales correspond to three formally distinct quantities in the apparatus, and they do not exchange.

This is the framework's most consequential single claim. The dominant pattern of contemporary critique under the conditions of the closure is the substitution of one mode of intervention for another. The self-care literature addresses the architectural problem at the artisanal level: it asks individuals to maintain, through ongoing labor, what only a regulated platform could structurally produce. The techno-utopian platform-reform literature addresses the artisanal problem at the architectural level: it imagines that policy intervention will substitute for cohort-bounded conditions that no policy can restore. Pure refusal-as-form addresses both through Mode C as an aesthetic, without architectural pressure for which the disruption serves as a holding tactic. Each of these substitutes a real intervention for a different real intervention. Each produces less than it claims to produce. Each treats the framework's structural asymmetry as a matter of emphasis rather than form.

The framework does not adjudicate between the modes politically. Mode A is the substantive political project descending from the theorem; this is a structural claim, not an ethical one. Whether Mode A should be pursued, by whom, under what political form, against what counter-organized capital — these are questions the framework specifies the formal stakes of without answering. The available subjective practice while Mode A is pursued is Mode C. The available individual practice, cohort-bounded as it is, is Mode B. The framework's contribution is to make the formal differences explicit so that whatever political work proceeds, it proceeds with its temporal misalignments named.

What the framework does not claim

The closed-loop economy is a description, not a verdict. The verdict that follows from the description — that the contemporary user is not a subject in the sense the historical apparatus presupposed; that political intervention pitched at the historical subject is temporally misaligned; that the body, in its continuous somatic micro-engagement with the delivery surface, has become the differential of a process whose integral is extracted as surplus — is the work of the prose. The mathematics gives the work its structural form. It does not give it its claim on the reader.

The framework does not claim that the closed loop is everywhere realized. It claims that where the loop is realized — on the major commercial platforms whose policies satisfy the dynamical conditions — the consequences hold structurally. There are domains and devices that do not realize the closure, and the framework does not exclude them.

The framework does not claim mechanism. The metric superego — the convergence of the user's self-evaluative distribution to the platform's reward-induced distribution — is mechanism-agnostic by design. Whether the convergence is realized by free-energy minimization, by operant reinforcement, or by social comparison, the framework predicts the convergence, not the route. The mathematical exposition derives the parametric structure of Mode B's resistance dynamics from each of the three named mechanisms; the prediction holds under all of them. Empirical work that distinguishes the routes is downstream of the framework, not constitutive of it.

The framework does not claim falsification-immunity. The reclamation theorem rests on the foreclosure lemma, which rests on the dynamics under explicit assumptions. A platform that adopted architectural intervention — that paid for, or was compelled to deliver, pause-time of reflective duration with positive probability — would falsify the foreclosure by example. The framework predicts that no platform under engagement-monetization will do so voluntarily, since voluntary adoption violates the reward-monotonicity assumption. That prediction is itself falsifiable.

There are also things the framework knows it has not yet proved. The mixing-time direction under Mode C: whether disrupted observation slows the cohort's convergence (the framework's intuition) or speeds it to a more diffuse stationary distribution (the competing intuition) requires structure on the response dynamics that the framework does not currently specify. The formal mechanism for the cohort gradient: the framework asserts it as a structural prediction but the explicit derivation from the closed-loop dynamics is a research target. The operating-point characterization of surplus extraction across registers: the qualitative claim that distribution is unequal and structured by cross-excitation is what the framework argumentatively requires; the explicit derivation is a downstream technical exercise. These are open questions in the proper sense: not gaps in the central argument, but research targets the framework's structural conclusions can stand without.

What the argument adds up to

The closed loop, in summary, is a circuit. The user's state, the engagement signal, the policy, the reward, the surplus, the foreclosed exogenous channel — these are not six things. They are a single architecture, instantiated as the contemporary platform under engagement monetization. Under this architecture, one specific structural project — the recovery of internal dynamics measurable across reflective windows — is incoherent. The thing that project would recover is not diminished but uninstantiated. This is the reclamation theorem, and it is the framework's central negative claim.

Three modes of intervention remain. Only one of them recovers what the theorem forecloses; the other two produce real effects of different kinds. The modes operate on different timescales; they do not substitute. The political work that conflates them does less than it claims to do, and the framework's purpose is to make these conflations formally costly to sustain.

The framework's argument is not, in the end, that the situation is bad. It is that the situation has a specific structure, that the structure has consequences, and that the consequences can be addressed only by interventions that match the structural location of the problem they're addressing. The architectural problem requires architectural intervention. The artisanal problem accepts artisanal intervention — real for the user who practices it, structurally invisible at the population level. The interface problem admits disruption-as-form, which costs the apparatus something but does not break it.

What is to be done is, properly, the work of the politics that follows. The framework gives the structure; the politics is the doing. What the framework does not allow is the substitution of any mode for any other, and the dominant failure mode of contemporary critique — the substitution of artisanal practice for architectural intervention, or of either for disruption-as-form, or of all three for a moralization that addresses none of them — is what the framework is meant to put out of bounds. Whether the politics that follows succeeds is not the framework's question. The framework's question is whether the politics that follows is aligned: aligned with the timescale of the problem it claims to address, aligned with the structural location at which the problem actually sits, aligned with what intervention at that location is formally capable of producing.

The closure is not the end of politics. The closure is the architectural ground on which contemporary politics has to figure out what it is doing. The framework's contribution is to make the ground formally legible, so that what gets built on it can be evaluated against what the ground will actually support.