The Two Tiers Are a Market, Not a Cage

2026-05-12 · 7,262 words · Singular Grit Substack · View on Substack

On the four-layer model, software defaults, and the moral economics of transaction access

Keywords: Bitcoin economics; censorship resistance; transaction costs; two-sided markets; mempool policy; proof-of-work; Coasean bargaining; Hayekian heterogeneity; network topology; BSV.

1. The Concession Examined

There is a particular discipline in refusing to mistake a concession for a victory. The correspondent writing under the name 铭链科技 has granted the topology distinction: the public gossip graph and the miner graph are different objects, and a vertex absent from the path cannot delay traffic upon it. The concession is correct. It is also the smaller of the two questions before us. The larger one—now placed before us with welcome precision—is whether differentiated paths to inclusion constitute a defect, a feature, or merely an unavoidable property of any system in which scarce services are delivered through priced channels.

The correspondent’s framing is this. Layer two of the four-layer model—transaction access—admits multiple delivery mechanisms: public gossip, direct submission to miners, pool infrastructure, commercial services. When public gossip degrades, ordinary users must shift to professional channels. Those channels are not equally accessible. They demand relationships, accounts, fees, sometimes identity. The system therefore exhibits a two-tier censorship resistance: sophisticated actors retain full access; the rest experience degraded service. This is, the correspondent allows, not the same thing as the “home-node sovereignty” claim being demolished. But it is a property of the system worth naming.

The argument has the merit of locating a real asymmetry. It has the defect of mistaking the asymmetry’s character. A market is not a cage. To call differentiated channels “censorship resistance of the second tier” is to import a moral content into a technical description that the description does not bear. The intuition that animates the worry—that some classes of users obtain easier inclusion than others—is the same intuition that animates the complaint that a first-class airline ticket is “first-class transportation” while economy is “second-class transportation,” with the implication that something has been done to the economy passenger. Something has been done. The passenger has been charged less.

The economics of Bitcoin has, by now, accumulated a literature substantial enough that the principal disputes are conducted at the level of which equilibrium is operative rather than whether the system has equilibria at all.1 The correspondent’s letter participates in that mature literature, and the present reply should be read in the same register. The argument I will make in what follows is that the correspondent’s framing collapses three distinct things: the cost of access, the existence of access, and the right of refusal. Censorship resistance is a property of the last of these. Access cost belongs to the first. Existence belongs to the second. Treating the cost of professional channels as itself a kind of censorship is the rhetorical move by which the regulatory state has, for two centuries, transformed the existence of markets into evidence of their failure.

2. What a Market Always Looks Like

Stigler taught us, before most of the contemporary literature had finished its undergraduate education, that information has a price.2 Search is costly. Matching is costly. The market for any good is not a single uniform venue at which all participants meet on identical terms; it is a stratified architecture of channels, each with its own cost structure, each serving the buyer for whom that cost structure is the cheapest.3 The wholesale channel exists because some buyers transact at sufficient scale to amortise the relationship cost. The retail channel exists because some buyers transact at insufficient scale to do so. Neither channel “discriminates.” Both channels price.

The correspondent’s two-tier observation is therefore not a peculiarity of Bitcoin transaction access; it is a structural feature of every market that has ever existed where information is positively costly and capacity is positively scarce. The commodities exchange, the wholesale produce market, the bond syndicate desk, the credit-card processor, the dispatch taxi versus the street hail, the air-freight forwarder versus the postal parcel: all exhibit the same architecture. Direct relationships to capacity-holders serve buyers with the scale to maintain them. Indirect public mechanisms serve buyers without it. Both routes deliver the good. The price differs. The service differs. The participant chooses.

To insist that this structure is something other than what it is—to call it “two-tier resistance” rather than “two-tier pricing”—is to introduce a moral predicate that the system’s behaviour does not warrant. The Wildean inversion is irresistible: those who object to differentiated channels in transaction submission are not objecting to censorship; they are objecting to the existence of price. A transaction that can be submitted, but only through a channel that bills for its delivery, has not been censored. It has been quoted.

This is not an idle distinction. The literature on two-sided platforms has spent two decades documenting that the existence of differentiated user-side and merchant-side pricing is not evidence of market failure but of efficient cross-subsidy.4 Rochet and Tirole demonstrated that the price structure on multi-sided platforms—who pays what for which side of the matching service—is endogenous to the elasticities of the relevant participants and is, under broad conditions, welfare-improving relative to a uniform price.5 Bitcoin is, among other things, a two-sided platform: senders demand inclusion; miners supply it; the price discovers what each side will bear. The correspondent’s “ordinary users” and “sophisticated actors” are precisely the two sides that any two-sided market will partition. To name the partition a censorship asymmetry is to misread an equilibrium for a pathology.

3. The Coasean Architecture of Channels

Coase observed that the firm exists because the transaction cost of organising activity through the price mechanism is, beyond a certain threshold, higher than the transaction cost of organising it through hierarchy.6 A complementary observation, drawn from the same theoretical apparatus and developed in his later work, is that the boundary between channels is itself a price.7 Where the cost of using the public mechanism is high relative to the cost of building a private one, private channels emerge. Where the reverse holds, they do not. The architecture of transaction submission in any monetary system will, accordingly, reflect not the moral preferences of its designers but the actual cost structure of the activity being transacted.

Demsetz extended the analysis to the question of why certain transactions occur in organised exchanges and others through bilateral negotiation, showing that the choice is governed by the relative cost of the alternative institutional arrangements.8 Williamson, building on Coase, identified the conditions—asset specificity, frequency, uncertainty—under which integration economises on transaction costs relative to market exchange.9 Klein, Crawford, and Alchian showed that the appropriability of quasi-rents drives vertical integration where bilateral hold-up risk is high.10 In every case the theoretical apparatus reaches the same conclusion: the architecture of channels is the equilibrium output of cost minimisation, and any insistence that a single uniform channel ought to serve all participants is, on close inspection, an insistence that the costs of the most expensive participant ought to be socialised across the entire system.

Apply this to the correspondent’s worry. A transaction may be submitted via the public gossip relay. The cost of doing so is, today, low—but it is not zero. It includes the cost of running or connecting to a relay node, the cost of maintaining software compatible with the prevailing relay policy, the cost of accepting the latency and the loss rate of an unmanaged network, and the cost of the bandwidth itself. A transaction may also be submitted via a direct API to a miner or pool, which charges a fee, may require an account, and may demand identification. The cost of this route includes the fee, the relationship overhead, and whatever compliance burden is attached. The participant rationally chooses the cheaper route for the particular transaction in question. Some transactions are cheap to route via gossip. Others are not. The architecture serves both.

The correspondent will reply that the second route is “not equally accessible to all participants.” This is true. It is also irrelevant to the censorship question, because equal accessibility is not the criterion of censorship resistance. The criterion is whether a transaction’s inclusion can be prevented by parties other than those who undertake the proof-of-work. A channel that charges a fee does not prevent inclusion; it prices it. A channel that demands an account does not prevent inclusion; it conditions it on the supplier’s willingness to be paid. To prevent inclusion would require that no channel serve the transaction. That is the question the correspondent must answer in the affirmative if the censorship framing is to do any work. So far as I can tell, the answer is no.

4. The Hayekian Heterogeneity of Users

The correspondent’s framing also presumes a population. “Ordinary users” and “sophisticated actors” are taken as classes with stable membership and shared interests, the first deserving of frictionless access and the second tolerating—indeed deserving, by virtue of their sophistication—the friction of professional channels. The presumption deserves examination because it is the bridge by which the two-tier observation becomes a moral complaint.

Hayek’s central methodological insight—that economic activity occurs in a context of dispersed, particular, often tacit knowledge that no central planner can aggregate—applies as cleanly to the question of what users want from a transaction system as to any other allocation problem.11 The user who needs anonymity, irreversibility, and low fee but tolerates long inclusion latency wants one thing. The user who needs same-block confirmation, high reliability, and is indifferent to fee wants another. The user who values the absence of any compliance overhead wants a third. The user who needs proof-of-funds documentation for downstream legal use wants a fourth. There is no single channel that serves all of these at minimum cost. There is no single price that all of them would pay. The market discovers, through the entry of specialised channels, what each class will bear, and channels emerge to serve each willingness.12

To insist on a single uniform channel is, in this light, to insist that the discovery process be foreclosed: to demand that the system serve, at uniform terms, a population whose actual preferences are heterogeneous and whose individual costs of being served on those uniform terms vary by orders of magnitude. The egalitarian intuition is comprehensible. Its economic implication is the suppression of the very mechanism by which the heterogeneous needs of actual users are met. The two-tier architecture is not a regrettable departure from a unified ideal; it is the form a market takes whenever the participants are not interchangeable.

The same point applies to the correspondent’s gesture toward “compliance, fees, and often identity” as a degradation. These are not degradations. They are properties of certain channels, chosen by participants for whom they are not deal-breakers, and avoided by participants for whom they are. A market in which both kinds of channel exist offers strictly more options than a market in which only one does. The complaint that the more demanding channels exist is, on inspection, a complaint that some users have preferences different from those of the complainant.

5. On Software Defaults and the Question of Authority

The correspondent’s strongest empirical claim is that most miners run software with default mempool policy inherited from BTC Core, and that this software coordination shapes the miner’s information environment as surely as it shapes the home node’s. Therefore, the argument runs, the layer-two transaction access surface and the layer-four verification surface are not cleanly separable. The miner’s mempool is the home node’s mempool with a different output.

The premise is partially true and the conclusion does not follow.

It is partially true that miners on the BTC chain predominantly run software derived from Core’s reference implementation with default relay and acceptance policies. It is also true that this is a contingent equilibrium, not a property of proof-of-work. Nothing in the consensus mechanism requires that a miner accept the same transaction set its peers accept; the mechanism requires only that the block, once produced, satisfy the consensus rules and that subsequent miners build upon it. The choice of which transactions to include in the candidate block is, structurally, the miner’s. The default mempool policy is a habit, not a law.

Biais, Bisière, Bouvard, and Casamatta formalised the observation that a blockchain admits many coordination equilibria; the prevailing one is sustained by the choices of participants, not by any technical necessity.13 The empirical literature on miner extractable value has, over the past five years, made unmistakably clear that direct submission to miners is not an exotic edge case but a substantial fraction of total transaction flow in systems where the relevant fee opportunity exists. Daian et al. documented in 2020 the existence of priority gas auctions and miner-direct order flow at scale on Ethereum, with private mempools serving as the dominant channel for value-sensitive transactions.14 The phenomenon is endogenous. Where the fee opportunity is sufficient to justify the relationship cost, direct submission emerges and dominates. Where it is not, public relay suffices. The correspondent’s “most miners run defaults” is true for the marginal value of a default route, not for transactions where the marginal value of a non-default route exceeds its marginal cost.

This matters for the censorship question in the following way. The correspondent treats software default coordination as a form of pseudo-authority over what transactions reach miners. But software defaults possess authority only to the extent that the alternative is more costly than the default for every participant who might choose it. The moment the alternative is cheaper—and the MEV literature shows it has become cheaper for entire classes of transaction—the default is abandoned, the alternative is adopted, and the coordination equilibrium shifts. To call this arrangement “censorship” is to misuse the word. The arrangement is a behavioural equilibrium subject to revision by anyone with a sufficient incentive to revise it. Censorship would require that no participant possess the incentive or capability to revise; the empirical record demonstrates that both incentive and capability exist and are exercised.

One may, of course, point to the further observation that Core’s policy defaults are non-trivially difficult to alter, that the developer community sustaining them exhibits coordinated preference, and that the social cost of running heterodox policy is non-zero. Each of these observations is correct. None of them transforms a behavioural equilibrium into a structural property of the protocol. The constraint is endogenous to the costs of deviation, and those costs are, in a market system, a function of who is willing to pay how much for the deviation. The correspondent’s empirical point is real. It is also, when properly framed, a description of how markets coordinate, not of how they fail.

6. From Possibility to Practice: The Empirical Map of Direct Submission

The argument advanced thus far rests on a structural claim—that miners are addressable through channels independent of public gossip relay—and the question naturally arises whether the structural claim is satisfied empirically. It is. The matter need not be left as theoretical conjecture; the relevant infrastructure is documented, advertised, and operational, and the configuration is verifiable by direct examination of miner-published interfaces.

On BSV, the principal commercial mining nodes publish documented public APIs for direct transaction submission. TAAL and GorillaPool—the operators that account for substantially the entire active commercial miner set on the network—both run mAPI and the successor ARC interface. The BSV Association’s design description of ARC characterises it as engineered to connect to every mining node on the network, with each ARC instance peering directly with the active commercial mining infrastructure rather than relying on the legacy gossip path.15 GorillaPool’s mAPI endpoint accepts transactions without requiring an API key for ordinary submission and explicitly advertises direct routing to the miner nodes at the centre of the mining network.16 TAAL’s ARC implementation is similarly publicly available, with API documentation supplied through the miner’s developer console.17 The ARC architecture was designed precisely to expose every commercial mining node through a standard, horizontally scalable, public interface; direct submission to the active hash rate is, on the network where the architecture has been adopted, the documented default of its commercial operators, not an exotic exception. The empirical proposition that any user can address any miner on BSV is therefore not a theoretical claim awaiting test. It is the design intent of the deployed infrastructure, satisfied at or near full coverage of the active hash rate.

On BTC, where the historical pattern has been a default mempool policy of considerable conservatism, direct-submission infrastructure has nonetheless emerged as a commercial offering at the level of the publicly-traded mining sector. Marathon Digital Holdings launched Slipstream in February 2024 as a direct Bitcoin transaction submission service, with the explicit purpose of admitting transactions that the default Core mempool policy would refuse. The company’s own description is unambiguous: Bitcoin nodes by default exclude large and non-standard transactions from the mempool even where those transactions adhere to consensus rules; Slipstream permits users to submit such transactions directly to Marathon’s mining pool, with inclusion conditional on protocol compliance and adequate fee.18 The infrastructure thus exists, in the BTC environment, as a publicly-listed company’s productised service offered through a public web terminal—a configuration whose accessibility to any user is verifiable by visiting the address.

The aggregate picture is this. On BSV, substantially all of the active hash rate is addressable through documented public submission interfaces; the design of the standard transaction-processing layer assumes and implements this addressability. On BTC, the publicly-listed mining sector has begun productising direct submission as a standardised commercial service, with the remainder of the hash rate reachable through documented if heterogeneous channels operated by individual pool operators. The structural claim—that miners are independently addressable—is not a theoretical proposition awaiting empirical demonstration. It is the observed configuration of the relevant infrastructure today, and the claim that at least ninety-nine per cent of the active miner set is directly addressable is consistent with the public configuration of that infrastructure as documented by its operators.

I emphasise the evidentiary status of these statements honestly. The sources are primary materials: corporate announcements, miner-published API documentation, and industry-association statements. They are not peer-reviewed empirical studies, and I do not present them as such. They are public statements by the operators of the infrastructure regarding the configuration of the infrastructure, and they are subject to direct verification by any reader who chooses to test them. They constitute the appropriate class of evidence for a claim about deployed infrastructure: not an inference from theory, but a description of installed systems whose existence and behaviour are matters of fact rather than of model. The non-default routes exist; they are advertised; they are documented; they are open. The deeper structural reason they suffice is the subject of the section that follows.

7. Reach One Equals Reach All: The Structural Implication of the Miner Subgraph

The correspondent’s two-tier framing rests on a presupposition that has not been examined and that does not survive examination. The framing assumes that access to the miner set is a property requiring access to each miner separately—that the user who has reached one miner has reached only one miner, and that the user who lacks an account or relationship with miner M2 lacks access to M2 specifically. The presupposition is, as a matter of network topology, false. It is the silent assumption on which the entire two-tier narrative depends, and once it is named the narrative does not survive.

Miners maintain a small-world subgraph among themselves. The reasons for this are economic and well-documented. Each miner has a direct incentive to receive blocks from peers as quickly as possible, since mining on a stale tip is a pure loss; each miner has a direct incentive to receive transactions from peers as quickly as possible, since the mempool of any miner is, for fee-bearing transactions of value, a profit opportunity. The miner subgraph is therefore densely interconnected with short path lengths—the canonical signature of small-world topology identified by Watts and Strogatz and shown to emerge robustly under the local optimisation of agents with instrumental reasons to maintain ties.19 Decker and Wattenhofer’s empirical measurements documented the same property in the Bitcoin propagation graph at a scale small enough to permit end-to-end characterisation: efficient propagation along short paths among well-connected nodes.20

The consequence for the two-tier framing is structural and decisive. Let GM denote the subgraph of active mining nodes, and let dM denote its diameter. In a small-world graph, dM scales logarithmically with the number of vertices and is, in practice, small—on the order of a few hops for mining networks of the size relevant to Bitcoin or BSV. Suppose a user submits transaction T to any single miner M1 ∈ GM, and suppose M1 accepts T into its mempool. Within dM propagation steps—milliseconds to seconds in practice—every other miner Mi ∈ GM has received T and may include T in any block it produces. The user did not reach every miner. The user reached one miner. The miner network’s own self-interested propagation reached the rest.

This is not an aspirational property. It is the documented operational behaviour of every functioning mining network in production. Each miner has positive expected revenue from including any fee-bearing transaction it sees; each miner therefore has positive incentive to forward such transactions to its peers; the miner-peer graph is dense; the propagation completes. The correspondent’s premise that the user must reach miners separately, each through its own channel, presupposes a network in which miners do not propagate transactions among themselves—a network that does not exist and could not exist as a functioning proof-of-work system. Propagation among miners is not a courtesy. It is a profit-maximising behaviour endogenous to the economics of mining itself.

The implication for the “open, anonymous, equally accessible channels” requirement is correspondingly decisive. The requirement is not merely costly to satisfy—that point is developed in the section that follows—it is structurally unnecessary. The user does not require access to every miner; the user requires access to any miner. A single open, anonymous, equally accessible channel to any single miner—the GorillaPool mAPI endpoint that requires no API key, the Slipstream public terminal, any one ARC instance on the BSV network—is functionally equivalent, for purposes of confirmability, to N such channels to N miners. The miner subgraph completes the equivalence at no cost to the user and through no action of the user. The correspondent’s two-tier framing, considered against the actual topology, collapses to the observation that some channels are more convenient than others—which is true of every market that has ever existed and which is not a property of censorship resistance under any defensible definition of the term.

Apostolaki, Zohar, and Vanbever’s work on routing attacks sharpens rather than weakens this point.21 The relevant adversary capable of disrupting reach-one-equals-reach-all is not the home-node relayer or the developer community choosing default mempool policy; it is the AS-level routing operator capable of partitioning the miner subgraph itself. The defence is route diversity at the network layer—a property of the infrastructure beneath the protocol, not of the protocol’s social conventions. This is a real and serious adversary worth attending to, and it identifies the correct location at which to direct concern about transaction-access resilience.

8. Layer Two and Layer Four: Overlap Is Not Equivalence

The correspondent grants that layer two and layer four are different. The correspondent then argues that the same software runs on both, that the public gossip graph and the miner graph overlap significantly, and that miners’ transaction access is shaped by the coordination that shapes home-node relay. From which it follows that the separation between the layers is “not as clean” as I have suggested.

The phrase “not as clean” does a great deal of work in the correspondent’s argument and survives less scrutiny than it deserves. Overlap is not equivalence. Two graphs that share most of their vertices are still distinguishable by the edges that are unique to each, and—more importantly—by the functions that those distinguishing edges perform. Two sets of software that share most of their code are still distinguishable by the decision points at which their behaviour diverges, and by the consequences that follow from each divergence.

The relevant question is not whether the layer-two and layer-four populations overlap in software. They do. The relevant question is what each layer’s operators can prevent, and what they cannot. A layer-four verifier can refuse to forward, refuse to acknowledge, refuse to relay—and the consequence for the transaction whose forwarding is refused is that it travels by a different route. The transaction is not refused inclusion; it is refused passage through that particular node. A layer-one operator—a miner exercising proof-of-work—can refuse to include, and the consequence for the transaction whose inclusion is refused is that it must wait for a different miner. The transaction is not refused inclusion; it is refused inclusion by that particular miner. In both cases the system’s structural property is the same: refusal at a node is bypassable; refusal at the network in aggregate is impossible by construction so long as any miner accepts the transaction and is permitted to build on its block.

The “overlap” of software between the two layers is therefore consistent with their structural separation. The same code may run on the relay-only node and on the miner. The same default policy may shape what each chooses to forward. But the miner’s choice to forward, or not to forward, is a different action from the miner’s choice to build a block containing a transaction. The relay node has no equivalent action. It cannot build. It cannot order. Its refusal to relay degrades the convenience of the public path; it does not degrade the existence of the path through the miner that the participant has independently engaged.

This is the asymmetry that the four-layer model was designed to make visible, and the correspondent has, despite the appearance of pressing against it, in fact reinforced it. The fact that miners can run software different from their peers, that they can accept direct submissions outside the default mempool, that they can build blocks containing transactions the default relay would refuse to carry, is the entire structural content of proof-of-work authority over inclusion. The empirical observation that they often do not exercise the option is a comment on the current state of the fee market, not on the location of the authority.

Miners’ strategic choices are not constrained to the defaults the network appears to use; deviation is available, structurally permitted by the consensus mechanism, and—where the value of the option exercised exceeds the cost of exercising it—rational. Cong, He, and Li showed that the pool structure of mining itself emerges as an endogenous response to risk and reward, with the configuration of the industry reflecting the cost-benefit calculus of participants rather than any imposed coordination.22 Huberman, Leshno, and Moallemi formalised the Bitcoin payment system as a queue with fee-driven priority, in which the relevant economic actors are unambiguously the miners pricing inclusion and the senders bidding for it.23 Easley, O’Hara, and Basu documented the empirical evolution of fees and the emergence of a market in inclusion priority.24 Pagnotta has further shown that the security of the chain is itself a market output, jointly determined by the price of the asset, the cost of mining, and the equilibrium hash rate—a finding that anchors the claim that inclusion service is supplied within a market structure that is itself a market output.25 The picture, across this literature, is consistent: inclusion is a service supplied by miners and demanded by senders, priced in a market, with channels emerging to serve each class of buyer. Software defaults are part of the marginal cost structure. They are not the structure itself.

9. User Experience Is Not a Structural Property

The correspondent makes one further move, and it is the most rhetorically powerful one in the letter. Censorship resistance, the argument runs, is a property of the system as experienced by its users, not a property of the graph topology in the abstract. A system in which any transaction can reach a miner, but only some transactions can reach miners through open, anonymous, equally accessible channels, has a different censorship resistance profile from a system in which all transactions can reach miners through such channels.

The move is rhetorically powerful because it appears to be humane. It appears to insist that we attend to the lived experience of the user rather than to dry topological abstractions. It is, however, also the move by which every prior expansion of regulatory authority over financial activity has been justified. The Securities Act’s prospectus requirement was justified by attention to the lived experience of the small investor. The Bank Secrecy Act was justified by attention to the lived experience of the law-enforcement officer. The Patriot Act’s anti-money-laundering provisions were justified by attention to the lived experience of the compliance officer trying to identify a customer. The history of financial regulation is, in significant part, the history of redefinitions of structural properties as user-experience properties, with the implication that authority must be exerted to make the user’s experience match the regulator’s preferences.26

I do not say this to suggest that the correspondent has any such agenda. I say it to suggest that the framing is one we should regard with care, because the framing has been used by parties who do. A structural property of a system—what it can do, what it can refuse, what it can guarantee—is a different kind of thing from a user-experience property—what is convenient, what is cheap, what is familiar. Both are real. Both matter. They are not the same property and they have different governance implications.

Censorship resistance, properly understood, is the structural property that a transaction whose sender is willing to compensate a competent operator cannot be prevented from inclusion by any party that is not also a competent operator. It is a statement about the location of the power of refusal. It is not a statement about the convenience, anonymity, fee level, or compliance posture of any particular channel. To redefine it as a property of user experience is to relocate the analytical centre of the system from the protocol to the user interface, and to make every observation about the user interface—every fee, every account requirement, every onboarding friction—into evidence about the protocol. That is a substitution, not an analysis.

The Stiglerian observation about regulation applies with full force.27 A vocabulary that converts every cost a user faces into a structural defect of the system is a vocabulary that will be wielded, sooner or later, by parties whose interest in the system’s reform is not the user’s interest. Peltzman extended the analysis to show that regulatory action systematically redistributes wealth from diffuse groups to concentrated ones, with the rhetoric of consumer protection serving as cover.28 Tullock and Krueger documented the rent-seeking dynamics by which interest groups extract value through the political process under the cover of public-interest justification.2930 The “user experience” framing of censorship resistance is precisely the kind of vocabulary that those dynamics will deploy, and is being deployed for, in the political contests now in progress over the future of digital payment infrastructure.

10. The Cost of Equal Access

It is worth pressing on what “open, anonymous, equally accessible channels” would actually require, because the cost of the demand is, on examination, the cost of the system itself.

An “open” channel must be costlessly enterable by any sender. A truly costless channel can sustain no quality, because the cost of the channel’s degradation is borne by parties other than those imposing it: the classic externality problem identified by Coase and resolvable, under positive transaction costs, only through some allocation of property rights that imposes a price.31 An “anonymous” channel must not require identification. Anonymity is itself a service with positive marginal cost: it requires that the channel forgo the information that would otherwise help it price its services efficiently and protect itself against abuse. An “equally accessible” channel must impose identical terms on all comers, which forecloses the discovery, identified by Hayek and elaborated by Kirzner, of what different participants would pay for differentiated service.32 Each of the three properties has a cost. To insist on all three is to insist that the system absorb all three costs as a precondition of being acknowledged as censorship-resistant.

The system does not need to absorb those costs to be censorship-resistant. It needs only that a transaction whose sender is willing to compensate a miner cannot be prevented from inclusion by any non-miner. The market for compensation may take many forms. The market for the relationships through which compensation is exchanged may take many forms. The market for the software that supports those relationships may take many forms. The structural property is preserved across all of them. The user-experience property varies across them—and the variation is, properly understood, the market doing its work.

The economic literature on information has spent more than half a century documenting that the costs of search, matching, and signalling are the principal determinants of how markets organise themselves under positive transaction costs.333435 The architecture of Bitcoin transaction submission is a particular instance of this general theory. The mempool is a search-and-matching mechanism with positive marginal cost of operation; the direct-submission channel is an alternative mechanism with a different cost structure; the commercial relay infrastructure is a third. The participant chooses, the market discovers, and the partition of users across channels reflects the equilibrium of the costs and the demands.

11. Closing: What the Four-Layer Model Was Always For

The four-layer model was not constructed to celebrate Bitcoin as a perfect machine. It was constructed to identify, with such precision as economic and network reasoning permits, which functions the system performs at which layer, and which assertions of authority each layer can sustain. The model was an analytical tool. The correspondent’s letter has done it the considerable honour of taking the tool seriously and asking what it implies. It implies, as the correspondent has correctly noted, that censorship resistance has a location—and that the location is not where the discourse of the past decade has often placed it.

It also implies, as the correspondent’s letter resists conceding, that the differentiated structure of transaction access channels is not a defect in the location of censorship resistance but a property of the markets that surround it. Pricing is not censorship. Conditioning is not refusal. Specialised channels are not exclusion. To call them so is to import into the analysis a moral predicate that the structural facts do not warrant. The deeper point, established in section 7, is that the user does not require reach to every miner; reach to any miner suffices, and the miner subgraph completes the rest.

Bitcoin’s censorship resistance, properly understood, is a structural property of its inclusion mechanism: a transaction whose sender is willing to compensate a competent operator cannot be prevented from inclusion by any non-operator. The convenience of access, the anonymity of channels, the level of fees, the burden of compliance—each is a market property, separable from the structural one, and each is subject to the ordinary discipline of choice and competition. To conflate the two is to invite a category of regulatory intervention whose stated purpose is the protection of the user and whose actual consequence is the foreclosure of the market by which user preferences are served.

The correspondent’s letter is a contribution. It has been received as such. The framework it presses on—the four-layer model—survives the pressing. The two-tier observation it offers is a real observation about a real market. It is not, however, an observation about censorship. It is an observation about price. The discipline of distinguishing the two is the discipline of refusing to surrender our vocabulary to those who would corrupt it.

Notes

Rainer Böhme, Nicolas Christin, Benjamin Edelman, and Tyler Moore, “Bitcoin: Economics, Technology, and Governance,” Journal of Economic Perspectives 29, no. 2 (2015): 213–238.

George J. Stigler, “The Economics of Information,” Journal of Political Economy 69, no. 3 (1961): 213–225.

Harold Demsetz, “The Cost of Transacting,” Quarterly Journal of Economics 82, no. 1 (1968): 33–53.

Jean-Charles Rochet and Jean Tirole, “Platform Competition in Two-Sided Markets,” Journal of the European Economic Association 1, no. 4 (2003): 990–1029.

Jean-Charles Rochet and Jean Tirole, “Two-Sided Markets: A Progress Report,” RAND Journal of Economics 37, no. 3 (2006): 645–667.

R. H. Coase, “The Nature of the Firm,” Economica 4, no. 16 (1937): 386–405.

R. H. Coase, “The Problem of Social Cost,” Journal of Law and Economics 3 (1960): 1–44.

Demsetz, “The Cost of Transacting,” 33–53.

Oliver E. Williamson, “Transaction-Cost Economics: The Governance of Contractual Relations,” Journal of Law and Economics 22, no. 2 (1979): 233–261.

Benjamin Klein, Robert G. Crawford, and Armen A. Alchian, “Vertical Integration, Appropriable Rents, and the Competitive Contracting Process,” Journal of Law and Economics 21, no. 2 (1978): 297–326.

F. A. Hayek, “The Use of Knowledge in Society,” American Economic Review 35, no. 4 (1945): 519–530.

Israel M. Kirzner, “Entrepreneurial Discovery and the Competitive Market Process: An Austrian Approach,” Journal of Economic Literature 35, no. 1 (1997): 60–85.

Bruno Biais, Christophe Bisière, Matthieu Bouvard, and Catherine Casamatta, “The Blockchain Folk Theorem,” Review of Financial Studies 32, no. 5 (2019): 1662–1715.

Philip Daian, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach, and Ari Juels, “Flash Boys 2.0: Frontrunning in Decentralized Exchanges, Miner Extractable Value, and Consensus Instability,” in Proceedings of the 2020 IEEE Symposium on Security and Privacy (2020): 910–927.

BSV Association, “Empowering Builders: ARC Service Set to Revolutionise Bitcoin Transaction Processing,” press release, 8 September 2023; see also BSV Association, “Understanding ARC’s Microservices,” 26 June 2024.

GorillaPool, “GorillaPool mAPI: Submit Bitcoin Transactions Directly to GorillaPool Merchant API,” (accessed May 2026).

TAAL, ARC commercial service documentation, accessible via the TAAL developer console (registration required for the commercial endpoint; open-source ARC reference implementation maintained by BSV Blockchain and TAAL). On the open-source character of the ARC reference implementation see BSV Association, “Understanding ARC’s Microservices,” n. 15 above.

Marathon Digital Holdings, Inc., “Marathon Digital Holdings Launches Slipstream,” press release, 22 February 2024.

Duncan J. Watts and Steven H. Strogatz, “Collective Dynamics of ‘Small-World’ Networks,” Nature 393, no. 6684 (1998): 440–442.

Christian Decker and Roger Wattenhofer, “Information Propagation in the Bitcoin Network,” in Proceedings of the IEEE P2P 2013 Conference (2013): 1–10.

Maria Apostolaki, Aviv Zohar, and Laurent Vanbever, “Hijacking Bitcoin: Routing Attacks on Cryptocurrencies,” in Proceedings of the 2017 IEEE Symposium on Security and Privacy (2017): 375–392.

Lin William Cong, Zhiguo He, and Jiasun Li, “Decentralized Mining in Centralized Pools,” Review of Financial Studies 34, no. 3 (2021): 1191–1235.

Gur Huberman, Jacob D. Leshno, and Ciamac Moallemi, “Monopoly without a Monopolist: An Economic Analysis of the Bitcoin Payment System,” Review of Economic Studies 88, no. 6 (2021): 3011–3040.

David Easley, Maureen O’Hara, and Soumya Basu, “From Mining to Markets: The Evolution of Bitcoin Transaction Fees,” Journal of Financial Economics 134, no. 1 (2019): 91–109.

Emiliano Pagnotta, “Decentralizing Money: Bitcoin Prices and Blockchain Security,” Review of Financial Studies 35, no. 2 (2022): 866–907.

George J. Stigler, “The Theory of Economic Regulation,” Bell Journal of Economics and Management Science 2, no. 1 (1971): 3–21.

Stigler, “The Theory of Economic Regulation,” 3–21.

Sam Peltzman, “Toward a More General Theory of Regulation,” Journal of Law and Economics 19, no. 2 (1976): 211–240.

Gordon Tullock, “The Welfare Costs of Tariffs, Monopolies, and Theft,” Western Economic Journal 5, no. 3 (1967): 224–232.

Anne O. Krueger, “The Political Economy of the Rent-Seeking Society,” American Economic Review 64, no. 3 (1974): 291–303.

Coase, “The Problem of Social Cost,” 1–44.

Kirzner, “Entrepreneurial Discovery,” 60–85.

Stigler, “The Economics of Information,” 213–225.

George A. Akerlof, “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Economics 84, no. 3 (1970): 488–500.

Michael Spence, “Job Market Signaling,” Quarterly Journal of Economics 87, no. 3 (1973): 355–374.

Bibliography

Akerlof, George A. “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism.” Quarterly Journal of Economics 84, no. 3 (1970): 488–500.

Apostolaki, Maria, Aviv Zohar, and Laurent Vanbever. “Hijacking Bitcoin: Routing Attacks on Cryptocurrencies.” In Proceedings of the 2017 IEEE Symposium on Security and Privacy, 375–392. 2017.

Biais, Bruno, Christophe Bisière, Matthieu Bouvard, and Catherine Casamatta. “The Blockchain Folk Theorem.” Review of Financial Studies 32, no. 5 (2019): 1662–1715.

Böhme, Rainer, Nicolas Christin, Benjamin Edelman, and Tyler Moore. “Bitcoin: Economics, Technology, and Governance.” Journal of Economic Perspectives 29, no. 2 (2015): 213–238.

Coase, R. H. “The Nature of the Firm.” Economica 4, no. 16 (1937): 386–405.

Coase, R. H. “The Problem of Social Cost.” Journal of Law and Economics 3 (1960): 1–44.

Cong, Lin William, Zhiguo He, and Jiasun Li. “Decentralized Mining in Centralized Pools.” Review of Financial Studies 34, no. 3 (2021): 1191–1235.

Daian, Philip, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach, and Ari Juels. “Flash Boys 2.0: Frontrunning in Decentralized Exchanges, Miner Extractable Value, and Consensus Instability.” In Proceedings of the 2020 IEEE Symposium on Security and Privacy, 910–927. 2020.

Decker, Christian, and Roger Wattenhofer. “Information Propagation in the Bitcoin Network.” In Proceedings of the IEEE P2P 2013 Conference, 1–10. 2013.

Demsetz, Harold. “The Cost of Transacting.” Quarterly Journal of Economics 82, no. 1 (1968): 33–53.

Easley, David, Maureen O’Hara, and Soumya Basu. “From Mining to Markets: The Evolution of Bitcoin Transaction Fees.” Journal of Financial Economics 134, no. 1 (2019): 91–109.

Hayek, F. A. “The Use of Knowledge in Society.” American Economic Review 35, no. 4 (1945): 519–530.

Huberman, Gur, Jacob D. Leshno, and Ciamac Moallemi. “Monopoly without a Monopolist: An Economic Analysis of the Bitcoin Payment System.” Review of Economic Studies 88, no. 6 (2021): 3011–3040.

Kirzner, Israel M. “Entrepreneurial Discovery and the Competitive Market Process: An Austrian Approach.” Journal of Economic Literature 35, no. 1 (1997): 60–85.

Klein, Benjamin, Robert G. Crawford, and Armen A. Alchian. “Vertical Integration, Appropriable Rents, and the Competitive Contracting Process.” Journal of Law and Economics 21, no. 2 (1978): 297–326.

Krueger, Anne O. “The Political Economy of the Rent-Seeking Society.” American Economic Review 64, no. 3 (1974): 291–303.

Pagnotta, Emiliano. “Decentralizing Money: Bitcoin Prices and Blockchain Security.” Review of Financial Studies 35, no. 2 (2022): 866–907.

Peltzman, Sam. “Toward a More General Theory of Regulation.” Journal of Law and Economics 19, no. 2 (1976): 211–240.

Rochet, Jean-Charles, and Jean Tirole. “Platform Competition in Two-Sided Markets.” Journal of the European Economic Association 1, no. 4 (2003): 990–1029.

Rochet, Jean-Charles, and Jean Tirole. “Two-Sided Markets: A Progress Report.” RAND Journal of Economics 37, no. 3 (2006): 645–667.

Spence, Michael. “Job Market Signaling.” Quarterly Journal of Economics 87, no. 3 (1973): 355–374.

Stigler, George J. “The Economics of Information.” Journal of Political Economy 69, no. 3 (1961): 213–225.

Stigler, George J. “The Theory of Economic Regulation.” Bell Journal of Economics and Management Science 2, no. 1 (1971): 3–21.

Tullock, Gordon. “The Welfare Costs of Tariffs, Monopolies, and Theft.” Western Economic Journal 5, no. 3 (1967): 224–232.

Watts, Duncan J., and Steven H. Strogatz. “Collective Dynamics of ‘Small-World’ Networks.” Nature 393, no. 6684 (1998): 440–442.

Williamson, Oliver E. “Transaction-Cost Economics: The Governance of Contractual Relations.” Journal of Law and Economics 22, no. 2 (1979): 233–261.

Primary Sources (Non-Peer-Reviewed, Cited as Documentary Evidence Only)

BSV Association. “Empowering Builders: ARC Service Set to Revolutionise Bitcoin Transaction Processing.” Press release, 8 September 2023.

BSV Association. “Understanding ARC’s Microservices.” 26 June 2024.

GorillaPool. “GorillaPool mAPI: Submit Bitcoin Transactions Directly to GorillaPool Merchant API.” Accessed May 2026.

Marathon Digital Holdings, Inc. “Marathon Digital Holdings Launches Slipstream.” Press release, 22 February 2024.

← Back to Substack Archive