Distributive Capture

Distributive technologies can reconcentrate power through the infrastructure required to operate them at scale. Drawing on Hughes’s (1983) technological momentum and Jasanoff’s (2004) co-production, distributive capture names the point at which distribution-enabling mediation becomes decision-bearing over the terms of participation and difficult for participants to discipline. A narrower concept, residual infrastructure conversion, traces the path: functions necessary for broad participation remain outside the formal core, are carried by intermediary operators, and become governance-relevant because practical participation depends on them.

The argument distinguishes coordination, concentration, governance relevance, and capture rather than folding them into one category. Historical analogs from the printing press, railroads, and the internet motivate the problem and show recurring family resemblance. Ethereum provides the primary contemporary illustration because participant paths in default wallet access, hosted application infrastructure, delegated staking, block construction, and crisis coordination can be reconstructed from public artifacts.

The strongest current novelty claim is jurisdictional and Ethereum-facing. Prior cycles were corrected through institutions with established jurisdiction over identifiable legal persons. Ethereum’s protocol layer is reached more unevenly, leaving protocol-layer correction dependent on concentrated internal coordination even as external correction reaches the intermediary layer more directly.

Keywords: distributive capture, residual infrastructure conversion, technological momentum, co-production, infrastructure studies, large technological systems, blockchain governance, Ethereum, jurisdictional asymmetry, residual formalization, philosophy of technology.

Ethereum presents itself as a system for permissionless participation in computation and settlement. Yet ordinary participation does not occur at the protocol layer in the abstract. It occurs through participant paths that traverse wallets, RPC providers, hosted interfaces, staking intermediaries, relay-builder markets, and crisis-coordination channels. The same infrastructure that makes distributed participation possible can also become decision-bearing over the conditions of that participation.

Ethereum is not historically isolated. Across the printing press, railroads, and the internet, one recurring pattern appears: infrastructure that scaled participation in a previously bottlenecked capacity also became the site through which access and participation were governed. These cases motivate the question rather than settling it. They suggest recurring structural family resemblance, not demonstrated mechanism identity across centuries.

The literature already establishes the background problem. Ferreira, Li, and Nikolowa (2023) demonstrate corporate capture of blockchain governance. Gazi (2025) analyzes blockchain’s decentralization paradox. Brown (2023) measures operational concentration in Ethereum. De Filippi (2020) argues that decentralized infrastructure does not produce decentralization of power. Musiani (2015) reaches an analogous conclusion for internet-based services, showing how decentralized architecture can reproduce centralized governance through the operational requirements of service provision at scale.

What is less specified is the threshold condition. Concentration can reappear without becoming decision-bearing over the terms on which a distributive system is used.

The threshold problem is when infrastructure required to distribute a capacity starts setting the terms of that distribution.

Distributive capture names the point at which distribution-enabling mediation becomes governance-relevant and difficult to discipline. Residual infrastructure conversion names the narrower path by which that happens: functions necessary for ordinary participation remain outside the formal core, are carried by intermediary operators, and can become decision-bearing because practical participation depends on them. Capture is reserved for cases in which that position lets operators shape participation or appropriation in ways participants cannot cheaply bypass, contest, or re-internalize.

The framework distinguishes five recurring pressures: mediation, standardization, operational authority, dependency, and governance leverage. They function here as a diagnostic sequence rather than as a finished law already established across cases.

Jasanoff’s co-production idiom explains why governance is unavoidable in distributive infrastructures at scale. The sequence identifies when a co-produced governance order hardens into dependence and when that dependence becomes governance-relevant for participants. The historical analogs motivate the problem. Ethereum provides the primary contemporary illustration because its participant paths are unusually traceable and its crisis episodes make effective authority unusually visible.

The blockchain-specific claim is jurisdictional. Earlier distributive cycles concentrated through actors that were legally attached to institutions capable of acting on them from outside the technology. Ethereum’s protocol layer is reached more unevenly. External regulation attaches most cleanly to the intermediary layer, while protocol-layer correction stays dependent on concentrated internal coordination. '''

The theoretical claims draw on Hughes’s large technological systems framework, Jasanoff’s co-production idiom, and infrastructure studies on mediation, installed base, and dependence. The empirical illustration is deliberately narrower. Ethereum is reconstructed from public dashboards, official support and product materials, staking and builder-market measurement, outage reporting, and public crisis-process artifacts, and where a thicker contemporary reconstruction is needed the article relies on the companion Ethereum dossier developed in Ethereal. The Bitcoin contrast uses repo artifacts, BIP texts, and the literature on governance, mining pools, and light-client dependence. The historical analogs are comparative and concept-forming rather than presented as equivalent proof across periods.

That evidentiary structure sets the bounds of the article. The historical cases motivate the problem and help specify the mechanism’s family resemblance. Ethereum supplies the primary contemporary illustration because participant paths and crisis episodes are unusually visible in public materials. The article therefore argues by concept formation and contemporary institutional reconstruction rather than by claiming one demonstrated causal sequence across every historical period discussed or a fully closed blockchain-class proof.

Thomas Hughes introduced technological momentum in Networks of Power (1983) and refined it across subsequent work (Hughes 1987, 1994). The concept occupies a deliberately constructed middle position between technological determinism and social constructivism, with time as the mediating variable. Young technological systems are socially malleable, shaped by what Hughes calls system builders: inventor-entrepreneurs who solve foundational problems, manager-entrepreneurs who organize operations at scale, and financier-entrepreneurs who solve problems of capitalization and political access during consolidation. As systems mature, they acquire inertial characteristics through the accumulation of physical infrastructure, organizational adaptations, regulatory frameworks, trained personnel, and consumer expectations that collectively resist redirection. The mass of these commitments, combined with the velocity of the system’s growth trajectory, constitutes its momentum. Hughes was explicit that momentum "is not the equivalent of determinism" (Hughes 1994): mature systems can be redirected, but redirection requires overcoming compounding resistance from every component simultaneously.

Large technological systems, in Hughes’s analysis, progress through overlapping phases: invention and development, innovation and competition, consolidation and rationalization, and finally momentum, in which institutional structures assume directive functions that no individual system builder controls. The concept of the reverse salient, borrowed from military strategy, describes subsystems that lag behind the advancing performance frontier of the overall system; reverse salients channel innovation toward resolving problems within the existing trajectory rather than exploring alternatives, reinforcing momentum by directing creative energy toward system optimization rather than system redesign.

Hughes’s electrical systems place immediate pressure on the distinction. They distributed a capacity that had previously been unavailable to most participants, and they concentrated through the infrastructure required for that distribution. If every infrastructure technology both distributes and reconcentrates, the distinction between general momentum and distributive momentum collapses.

The distinction is best treated as a heuristic with observable indicators rather than as a categorical separation. Many infrastructures have distributional consequences. The focus here is the subset of systems publicly organized around widening access to a previously concentrated capacity, where participation depends on a mediating layer and control of that layer shapes access to the widened capacity itself. In Hughes’s electrical systems, the governance arrangements produced by momentum governed service delivery in ways that clearly had distributional consequences. In the cases emphasized here, the redistributive ambition is more explicit and the governance question bears more directly on who reaches the widened capacity, on what terms, and with what share of it. Edwards (2003) argued that all infrastructures channel flows and produce distributional consequences. Distributive capture concerns the narrower class of cases where widening access is part of the system’s own public object and where governance over the mediating layer bears directly on that object.

The behavioral signature is observable in narrower form: the Stationers' order governed a dominant licensed print-distribution regime that the press had widened beyond earlier control, while railroad administration turned pricing, routing, and terminal access into authority over how market access was distributed. This tension between distributive purpose and concentrated governance over the distributed capacity is the phenomenon distributive capture names.

Hughes’s framework has been applied to internet infrastructure and platform systems (the intellectual lineage runs through Bijker, Hughes, and Pinch 1987; Latour 1987), but the specification for distributive systems has not been developed.

Five conditions, each individually familiar from institutional economics and infrastructure studies, decompose the pressure by which distributive systems can move from mediation toward governance-relevant concentration. They function as a diagnostic sequence rather than as a self-executing law.

Distribution at scale requires mediation. A capacity exercised autonomously at small scale (printing a pamphlet, exchanging a message, or settling directly among a small set of counterparties) can only be exercised at the scale the technology makes possible through shared infrastructure that mediates between the individual participant and the distributed capacity. The press required printshops; railroads required track, stations, and scheduling systems; the internet required routing infrastructure and hosting services; Ethereum requires wallets, client software, RPC access, validators, and block-construction infrastructure. Mediation is the enabling condition of distribution at scale.

Mediation requires standardization. Shared infrastructure operates through protocols, formats, and interfaces that standardize the terms of participation. Standardization makes infrastructure sharable; without it, each participant’s infrastructure is private and the distribution fails. Star and Ruhleder (1996) argued that infrastructure is "fundamentally and always a relation, never a thing": it becomes infrastructure only in relation to organized practices. Standardization is the moment when this relation crystallizes: the terms of participation become fixed enough for shared practice to cohere around them. Scott (1998) showed that analogous standardization projects (cadastral surveys, permanent surnames, uniform measurement) function as tools of "administrative legibility" that make populations and territories visible to governance. Scott also documented that legibility projects encounter resistance and frequently produce unintended consequences, an observation that applies to infrastructure standardization: standardization creates governance capacity while also creating the friction and resistance through which that governance is contested. Amazon Web Services achieved infrastructure dominance through proprietary APIs (S3, Lambda, DynamoDB) rather than open protocol standardization, a trajectory the standardization condition must address. The standardization condition specifies that shared infrastructure requires standardized terms of participation; it does not require that the standard be open or publicly governed. AWS’s S3 API has become so dominant that competitors implement S3-compatible interfaces to achieve market viability; it is proprietary in origin and de facto standard in effect. Proprietary standards achieve the same governance effect as open ones, concentrating authority in whoever controls the standard’s evolution, while also restricting the exit pathway that open standards provide through reimplementation.

Maintaining a standard often concentrates operational authority. The actors who manage implementation, resolve disputes, and adapt the standard as conditions change accumulate authority in proportion to the infrastructure’s scale. This accumulation is a condition of the standard’s continued operation. The skills, capital, and institutional relationships required for infrastructure maintenance are scarce relative to the population served, a scarcity rooted in the division of labor that infrastructure operation at scale demands. Dependency becomes analytically important when participants reorganize their own activity around the infrastructure’s continued availability, creating exit costs and adjustment burdens that weaken routine discipline over operators.

The chain does not always complete. Distributive capture is most likely when a set of interruption conditions are jointly weak. Those conditions include: whether participants can cheaply reproduce the distributed capacity outside the governed infrastructure (reproduction cost); whether the infrastructure is modular enough that contributions can be made independently (modularity and granularity, following Benkler 2006); whether the infrastructure’s value is inseparable from its scale (network-effect lock-in); whether the governance arrangement generates revenue that incentivizes its persistence (rent-extraction incentive); and whether participants possess practical tools of contestation or local autonomy (governance tooling). When any interruption condition is strong enough to discipline the governance order through credible departure, independent reproduction, or organized contestation, the chain softens or stops.

The sequence is analytical rather than deterministic. Mediation and standardization identify how a distributed capacity becomes infrastructural. Operational authority and dependency identify where contestable pressure toward leverage emerges. Governance leverage is the discriminating step: a residual function becomes governance-relevant when its operators can shape access, ordering, valuation, continuity, legitimacy, or accountability for participants. Capture should be reserved for the narrower case in which that leverage is difficult to bypass or discipline and can shape the terms on which the distributed capacity is used or appropriated. The interruption conditions matter because they decide whether this pressure hardens, softens, or stays limited to coordination.

Linux kernel development concentrates governance (Torvalds as BDFL, corporate contributors controlling the majority of commits) while Linux usage is distributed, because modularity, granularity, and costless forking keep the interruption conditions strong. SMTP illustrates partial interruption: the protocol layer stays federated because interoperability standards allow independent operators to compete, but the infrastructure layer (spam filtering, deliverability scoring) has reconcentrated where reproduction costs are high and network-effect lock-in is strong. Wikipedia’s editorial governance concentrated (Shaw and Hill 2014; Halfaker et al. 2013), but commercial capture did not follow because Creative Commons licensing, nonprofit legal form, and the absence of advertising revenue kept reproduction cost low and rent-extraction incentive absent (Konieczny 2009; Forte and Bruckman 2008). BitTorrent resisted concentration through protocol evolution (DHTs, peer exchange, magnet links) that maintained high modularity, though its primary pressures were exogenous (copyright litigation, ISP throttling per Mueller and Asghari 2012) rather than endogenous infrastructure capture. The Fediverse (Raman et al. 2019) exhibits infrastructure concentration at the hosting layer (top 5% of instances host ~90.6% of users) while governance contestation tools (defederation, as documented by Theophilos 2024) keep the fifth condition actively contested.

Each counterexample maps to the interruption stack rather than requiring a separate ad hoc exclusion. The sequence hardens where reproduction is costly, network effects lock participants to scale, rent extraction incentivizes governance persistence, and contestation tools are absent or ineffective. Ethereum, the primary contemporary case examined below, displays these pressures across several operational paths, though the threshold into capture must be argued path by path rather than assumed for the system as a whole.

Where interruption conditions are jointly weak, dependency becomes the medium of governance. The dependencies the infrastructure creates are the mechanism through which participation is conditioned, priced, and controlled. The Stationers' order shaped what moved through the dominant licensed circulation infrastructure. Railroad operators shaped market access through freight pricing, routing, and terminal control. Internet platforms and hosting layers shape visibility, continuity, and monetization through control of the application and service stack.

In distributive systems, momentum accumulates through the same process, but the infrastructure that acquires momentum is the infrastructure through which distribution operates, and the governance arrangements that momentum produces govern the distribution itself. Each condition simultaneously enables distribution and reconcentration: mediation makes participation possible and creates positional advantage; standardization makes infrastructure sharable and concentrates authority over the standard; dependency makes the distributed capacity available at scale and insulates operators from competitive discipline. The same institutional process, viewed from the participant’s position, is distribution; viewed from the operator’s position, it is governance.

Distributive infrastructures do not begin as pure distribution and later acquire governance. Jasanoff’s (2004) idiom of co-production explains why governance is constitutively present from the outset: the architectural choices that make participation possible also determine who controls interfaces, maintenance, compatibility, and access. Co-production is therefore not the mechanism of concentration itself. It is the claim that technical order and governance order are formed together. The conversion account developed above provides the directional content by asking when a residual function within that co-produced order becomes decision-bearing and, in the narrower set of cases, capture.

This distinction matters. Co-production establishes simultaneity, not direction. Which governance order emerges is historically contingent and institutionally specific. At scale, distributive infrastructures always co-produce some governance arrangement because the terms of participation are simultaneously terms of governance. The conversion account then explains why, under weak interruption conditions, that governance arrangement can harden into concentrated residual authority rather than staying federated or democratically contestable.

The historical analogs illustrate this constitutive point. The printing press and the Stationers' governance order were formed through the same process of market construction. As demand for printed material scaled, the capital and coordination requirements of paper supply, binding, licensing, and circulation scaled with it, and the actors who accumulated those capacities also governed what moved through the dominant licensed network. Johns (1998) showed that this order was institutionally constructed rather than technologically predetermined. The pamphlet explosion of the 1640s (Raymond 2003) does not dissolve the point. It shows that governance over licensed distribution coexisted with bypass routes that escaped the governed infrastructure.

Railroad coordination and railroad governance were likewise co-produced. Chandler (1977) showed that continental rail operation required managerial systems capable of scheduling, pricing, and dispute resolution at scale. Those same administrative capacities became the basis of concentrated authority over access to routes, terminals, and pricing. Railroads are a boundary case because service coordination and distributive consequence were tightly joined. Internet protocols and platform governance exhibit the same relationship in later form. TCP/IP distributed communication capacity, but the interfaces, hosting layers, and application ecosystems required to operationalize that openness at scale concentrated governance over visibility, continuity, moderation, and monetization (Lessig 1999, 2006; Wu 2010; Blanke and Pybus 2020).

Scholars of blockchain governance have worked close to this intuition for years without always using Jasanoff’s vocabulary. De Filippi and Wright (2018) argue that code and law are mutually constitutive. Reijers and Coeckelbergh (2018), Schneider (2019), Walch (2019), Werbach (2018), and Zwitter and Hazenberg (2020) each describe, in different terms, the simultaneity of technical architecture and governance order. The move here is to connect that simultaneity to a specific concentrating mechanism and to place the Ethereum case in comparative relation to earlier distributive infrastructures.

Three comparative-historical analogs show recurring structural family resemblance with the mechanism developed here. In the printing press case, the infrastructure that scaled textual reproduction, press capacity, paper supply, and circulation also became the site through which a dominant licensed distribution order was governed. In the railroad case, the infrastructure that scaled market access, track, stations, schedules, and freight coordination also became the site through which pricing, routing, and terminal access were governed. In the internet case, the protocols that distributed communication remained comparatively open while the operational infrastructure required to make that openness usable at scale concentrated into cloud, platform, and interface layers. In each analog, the infrastructure that enabled broader participation also became a site through which participation was shaped, though not in identical form.

These analogs are motivational and comparative rather than probative in the strict sense. They do not establish one identical causal sequence across centuries. They do show that Ethereum is not historically anomalous: distributive systems repeatedly create mediating layers that can become governance-relevant, while bypass routes and correction structures vary across cases. The point of the analogs is to motivate the mechanism, show recurring structural family resemblance, and sharpen the contemporary question the Ethereum case allows examination with greater empirical precision.

Prior analogs also clarify the correction problem. The Stationers' Company, railroad corporations, and later internet platforms were all tethered to identifiable legal persons operating within established jurisdictions. External correction therefore attached to actors that courts, legislatures, and regulators could name and act upon. Whether the same kind of external correction can attach to Ethereum’s protocol layer is the question the following section addresses.

Ethereum provides the primary contemporary illustration because its distributive ambition is explicit, several participant paths are traceable with public artifacts, and its crisis episodes make effective authority unusually visible. The illustration below draws on public dashboards, official documents, public support materials, and existing academic studies rather than on original on-chain collection. The task is to trace one system from distribution-enabling mediation to governance-relevant conversion and, where the threshold is crossed, capture.

Bitcoin is the first serious pressure test for the threshold because it shares the same broad domain of protocol-mediated participation while resisting any simple contrast between purity and failure. Off-chain development hierarchies and technocratic power have long been noted, and empirical work finds meaningful mining-pool concentration and internal pool concentration as well (De Filippi and Loveluck 2016; Wang et al. 2019; Romiti et al. 2019). Bitcoin cannot therefore function as a clean case of absent mediation or absent concentration.

The contrast here is anchored in protocol-native validation, documented light-client designs, mining coordination, and activation materials. It does not claim that exchange-custodial participation is marginal in the wider Bitcoin economy.

The literature does not show that these concentrations collapse into one uncontested governance center. Nabilou (2021) argues that Bitcoin’s governance arrangements have largely preserved the system’s censorship-resistance objective through major crises. Notland, Nowostawski, and Li (2024) find more decentralized behavior in consensus evolution than standard centralization diagnoses would predict. Cong, He, and Li (2021) likewise show that centralized pools do not mechanically eliminate decentralization because miners can diversify across pools and pool managers face endogenous constraints. The activation field constructed around BIP 8, BIP 91, and BIP 148 reinforces the same point: miners matter, but miners do not exhaust constitutional authority.

The comparison therefore belongs at the level of participant paths rather than system-wide slogans. Full-node and signer paths in Bitcoin retain stronger institutionally supported routes for local verification and local custody. Light-client paths remain dependent and privacy-limited, as Kotzer and Rottenstreich (2024) show, so the system does not escape residual mediation. The pooled-miner path is the strongest concentration path because template control, transaction inclusion, and signaling can shape participation conditions on that path, but it does not automatically become generalized governance over ordinary participation across all paths. This makes Bitcoin the clearest current negative-boundary case for the interruption stack: residual mediation is real, but the full-node and signer path remains more locally reproducible and less thickly bundled into one default operational path than the comparable ordinary-participant routes in Ethereum.

Bitcoin therefore disciplines the threshold rather than weakening it. Residual functions and concentration are insufficient. The relevant question is whether a residual layer becomes decision-bearing for an ordinary participant path and is hard to bypass or discipline through credible technical, institutional, or economic alternatives. On the present record, Ethereum crosses that line more clearly in default access and builder-mediated ordering than Bitcoin does across its participant paths. The same contrast weakens the jurisdictional asymmetry: both systems exhibit uneven external reach at the protocol layer, but Ethereum currently routes ordinary participation through thicker residual mediation and therefore makes the asymmetry more governance-relevant.

The argument concerns cases in which broad participation depends on residual infrastructures that become decision-bearing over participant paths. Ethereum makes that problem unusually visible because participant paths in access, staking, block construction, and crisis coordination can be traced with public artifacts.

The conceptual contribution is a thresholded account. Distributed capacity, residual mediating function, conversion path, and capture threshold need to be distinguished if the literature is to say more than that concentration exists. Some residual functions remain coordination devices. Some become governance-relevant. A smaller set become capture because participants cannot effectively bypass or discipline the terms they impose.

The first live prediction object is likewise narrower than a finished system-wide theory. If materially independent client diversification weakens one concentration surface without dissolving residual dependence elsewhere, the framework expects relocation or thinning of the threshold rather than automatic elimination of governance-relevant mediation.

Three lines of work follow. The interruption conditions need sharper operationalization. Cross-system comparison needs to test whether similar residual functions convert in the same way across different architectures. The jurisdictional asymmetry needs longitudinal study as regulation reaches farther into intermediary layers. Those are the right burdens for a theory of distributive capture anchored in residual infrastructure conversion rather than in a universal claim that every distributive system reconcentrates in the same way.