§20 Silicon Networks and Participants in the Value Internet

Part V still centered human decision-makers: procurement signatures, board votes, bank-counter identity checks. When value-network nodes expand to APIs, sensors, and autonomous Agents, “who may spend, and under what rules” becomes a new question—they hold no passports, do not queue for accounts in business hours, yet already perform large coordination functions on the information Internet. When the Value Internet upgrades from “passing information” to “passing and settling value,” reputation updates on a quarterly clock while machine collaboration runs in milliseconds; whether constraint carriers can shift toward verifiable rules becomes a dimension monetary theory must take head-on.

Section 1. From the Information Internet to the Value Internet

Since the late twentieth century, TCP/IP, HTTP, SMTP, and related protocols have driven the cost of global information exchange near zero: an email can cross continents; a webpage can be fetched by any node; the sender need not care which OS or access provider the receiver uses. The information Internet’s success did not rest on one firm monopolizing all content, but on open protocols that let any node participate under a common syntax—precisely Mr. Hayek’s emphasis that “rules precede particular executors,” realized in engineering: protocols public, participants free to enter or exit, competition at the application layer, the routing layer not locked by licenses.

The Value Internet vision extends the same logic to settleable asset transfer. VTP (Value Transfer Protocol) is designed in Openverse documents as the value layer’s routing protocol: by analogy with email addressing and delivery, value messages can route across chains and asset standards in a unified format, without a custom bridge for every asset pair. At the theory layer one may also compare ISO 20022 financial messaging or HTTP-layer micropayment experiments such as x402 to grasp the general structure of “routing semantics separated from settlement semantics.” VTP does not replace each chain’s consensus and security model; it (if landed as designed) lowers the interoperability cost of “how value is addressed, identified, and delivered.” As HTTP only carries messages and does not vouch for content authenticity, VTP if landed owns the routing semantics of value messages; asset authenticity, issuance rules, and final settlement still depend on on-chain state and smart contracts.

The leap from information to value cannot be completed by adding a payment button to today’s Internet. Information can be copied with near-zero loss; value transfer must solve double-spending, finality, counterparty risk, and compliance identity. Information-Internet participants may browse anonymously; if Value-Internet participants cannot be identified or held accountable, large-scale commercial collaboration is hard to establish. Hence the Value Internet must, atop open-protocol spirit, add verifiable identity, auditable ledgers, and programmable settlement conditions—on-chain verifiable rules and public asset standards thus enter the discussion (protocol-neutral counterparts include already-running standards such as ERC/USDC). Permissioned private-domain stablecoins are already discussed in corporate supply-chain scenes; when the counterparty is a continuously running program rather than a legal person or natural person, whether that toolkit is more natural than fiat accounts must be tested by pilot data, not extrapolated from roadmaps.

Section 2. A Spectrum of Value-Network Participants

Entities in value networks may be roughly classed as carbon-based or silicon-based by “what decides.” Carbon-based participants include natural persons, companies, partnerships, nonprofits, and the like—legally identifiable, able to bear civil liability, open bank accounts, and sign legally recognized electronic signatures. Silicon-based participants are system components without independent legal personality that nonetheless initiate economic acts automatically: API gateways that meter and bill, IoT sensors that collect environmental data and trigger contracts, trading bots that place orders by strategy, and large-language-model Agents that have rapidly entered public view—they receive goals, call tools, interact with other services, and complete multi-step operations without step-by-step human approval.

The division here is by institutional interface, not an ontological claim. A warehouse robot arm remains, in law, its owner’s property; a SaaS platform’s API remains, in contract, an extension of the operator’s obligations. At runtime, however, what initiates payment requests, submits on-chain transactions, and consumes API quotas is often a program, not a human click. Traditional financial infrastructure assumes every payment rests on a traceable natural-person authorization chain: account opening requires in-person or video verification; large transfers require tokens or SMS codes; corporate payments require dual review by finance lead and cashier. Silicon subjects cannot satisfy those rituals at runtime—they have no “presence,” no “self,” and no “emergency counter” outside business hours.

Carbon and silicon nest together. Most silicon economic acts still require a carbon subject to bear ultimate off-chain responsibility: companies answer for erroneous payments by their automation; developers bear civil or criminal liability for contract vulnerabilities they deploy. The question is not whether silicon can fully escape carbon guardianship, but whether—when coordination frequency and grain rise to milliseconds, micropayments, and cross-service depth—requiring every value transfer to rewind to manual carbon authorization will kill at the start much collaboration that should occur. If the Value Internet is to carry a machine economy, it must offer silicon participants value interfaces matched to their runtime character—Agent wallets and machine-readable asset layers are concrete forms of that interface.

Section 3. Friction of Fiat Account Systems for Non-Human Subjects

Modern commercial-bank account systems are products of twentieth-century carbon society and paper identity. Opening an account requires valid ID, proof of address, face recognition or in-person verification; corporate opening requires business licenses, charters, legal-representative authorization, and beneficial-owner disclosure. Those requirements serve AML and KYC goals and are reasonable against natural-person fraud and shell-company laundering. Applied to “opening a settlement account for an inference-service instance deployed in a cloud region,” however, they expose a structural mismatch.

The opening process mismatches silicon runtime from subject identification onward. Bank and payment-institution service agreements usually assume account holders are natural or legal persons, not “a service account in a Kubernetes namespace.” Separate sub-accounts for automation in practice usually hang under the parent company’s subject, with unified credit and reconciliation; the subsystem itself cannot own an independent, programmable off-chain balance view. Result: microservice A consumes B’s API, yet ledger entries consolidate on the parent’s general ledger; internal cost allocation depends on ex post ERP allocation rather than real-time value transfer.

Authorization likewise blocks automation. Online banking and payment APIs commonly require human-triggered second-factor authentication: SMS, tokens, biometrics. An autonomous Agent calling a paid data API at 3 a.m. on a market signal cannot wake the CFO for an SMS code. Some firms bypass real-time authorization with “prefunding + fixed API Key,” but that moves credit risk forward into prepaid risk and cannot support fine per-call, per-volume, or per-second billing—contrary to the economics of cloud-native elastic scaling.

Business hours and clearing cadence form another friction layer. Interbank transfers are limited by central-bank clearing hours and holidays; cross-border wires pass correspondent chains measured in business days. Silicon systems trade in seconds or milliseconds; fiat pipes’ batch clearing differs by orders of magnitude. Firms often compromise with “end-of-day batch settlement,” sacrificing timeliness in high-volatility scenes (compute bidding, on-chain arbitrage, real-time ad auctions) or concentrating risk on a few market makers.

Compliance-report grain closes the problem on the definition of “who.” Regulation requires tracing “who paid whom for what”; in carbon scenes “who” maps to ID numbers or unified social credit codes. In silicon scenes “who” may be hundreds of concurrent instances under one legal person; traditional reports struggle to distinguish “the company” from “the company’s automated agent No. 47.” Without finer programmable identity and auditable logs, compliance departments often ban autonomous payment wholesale, and innovation stays in human-approval sandboxes. For money to perform its medium-of-circulation function, it sometimes must appear in its golden person—Marx put the requirement bluntly in Volume I of Capital¹; on-chain addresses and signatures, in silicon scenes, are the digital counterparts of that same function.

These frictions do not mean fiat systems should open exemption lanes for non-human subjects—money-laundering and fraud risks are real. They mean that, under equally strict auditability, on-chain accounts and smart contracts can supply parallel infrastructure: address as interface, signature as authorization, state as balance, twenty-four hours without interruption, each transfer bearing an immutable timestamp and call trail. The further question: should a firm’s internal silicon components also hold on-chain positions constrained by governance rules?

Section 4. Silicon Coordination in the API Economy

The Application Programming Interface is the most visible form of the silicon economy. Maps, payments, SMS, model inference, logistics tracks—much modern commercial capability is sold as APIs per call or per month. Stripe, Twilio, AWS, and similar platforms build on the chain “developer registers → binds payment method → obtains API Key → program charges”; silicon appears to call, but the settlement node remains a carbon subject’s pre-authorized card or account balance.

That indirect structure works at small scale; when the caller is itself automation and the call chain depth exceeds three layers, cost and liability attribution blur quickly. Example: Agent A calls Agent B’s orchestrated “market research” skill, which in turn calls news and translation APIs. The month-end bill is one aggregated invoice; it cannot answer “which user query triggered which external API fee”—bad for internal pricing, Agent performance review, and abuse detection. Centralized platforms’ “usage dashboards” are ex post statistics, not composable, verifiable settlement credentials.

On-chain micropayments and programmable limits offer another approach: each API call may accompany a tiny on-chain transfer or consume a verifiable “quota token.” Caller address, callee address, timestamp, and parameter hash write to the same ledger; auditors, regulators, and partners can independently replay afterward without trusting a platform’s unilateral CSV export. The economic counterpart paid for an API call remains a service as a special form of product; only the unit of measure shifts from invoice line items to on-chain events. Centralized API markets still have efficiency in cold start, customer support, and dispute handling; when on-chain settlement layers are widely adopted in pilots, metering and settlement may sink into protocol, decoupled from the application layer—as HTTP decoupled from webpage content—to be tested by x402 and similar pilot data, not whitepaper extrapolation.

Sensors and IoT form the other wing of the silicon economy. Smart meters, cold-chain probes, industrial machine-status reports—they do not produce “intent,” yet produce data that can trigger economic consequences. If temperature exceeds limits, an insurance contract should pay; if equipment runs at full load, a bonus should pay the compute supplier per protocol. Realizing such logic with traditional bank accounts requires data platforms to cloud, ERP to aggregate, and finance to initiate payment manually—delay measured in hours. After oracles feed data into smart contracts, payment can trigger in sync with data arrival—provided sensor identity, data signatures, and contract terms were lawfully configured by carbon subjects before deployment. The governance frame of silicon trigger / carbon liability continues the supply-chain lineage of “trade information on-chain, automatic debit at maturity,” only at finer subject grain.

Section 5. Agents as Emerging Economic Participants

The combination of large language models and tool-calling frameworks has moved “Agents” from academic concept into engineering reality. The user states a goal; the Agent decomposes subtasks, calls search, code execution, database queries, third-party payment, and other tools, approaching results over multi-turn interaction. Unlike fixed-rule RPA (robotic process automation), Agents’ action space is more open, so economic effects on the outside world are harder to enumerate ex ante: they may misread prompts, suffer prompt injection from malicious pages, or buy unnecessary services in the name of “completing the task.”

From a value-network view, Agents’ distinctiveness is that they combine silicon execution speed with the fuzzy boundary of carbon mandate. The user says “book me the cheapest flight”; the Agent may call comparison APIs, airline directs, even on-chain stablecoin payment—without step-by-step user confirmation. If the payment pipe remains a linked card account, risk sits with user and issuer; if the Agent holds an on-chain wallet, the risk structure becomes: who custodians the key, who sets spend caps, who recovers erroneous transfers. Agents already exceed “smarter scripts”: they are new participants on the Value Internet that may hold independent on-chain identity and balances; their institutional treatment will decide whether a machine-readable value layer can land.

Enterprise “employee Agents” and consumer “personal Agents” have different liability structures. The former run under employment contracts and information-security policy, with tool whitelists and spend policies set by the firm; the latter engage consumer protection, unauthorized debit disputes, and traditional financial-regulation issues. In either case, fully banning Agents from value transfer abandons automation’s efficiency dividend; fully unleashing them is unlimited risk. The middle path is “governed on-chain proxy accounts”: multisig, daily limits, pausability, auditability—on-chain-native capabilities that map precisely to control instruments fiat systems struggle to delegate cheaply to subroutines.

Section 6. VTP and the Lightweight Role of Value Routing

Clarify VTP’s place in the overall architecture first, lest stack layers blur. Openverse is designed as Layer 0—a global decentralized value-exchange layer: at the protocol layer it supplies value addressing, routing, final-settlement anchoring, and cross-domain interoperability, so that various Layer 1 networks (ecosystem app chains, Ethereum, Cosmos chains, consortium chains, and the like) can connect—not a closed Hub/Zone LAN. VTP, the exchange layer’s routing protocol, is intended to own format and delivery of cross-domain value messages; UNS (Universal Name Service) maps human-readable addresses to on-chain identifiers; VRC standards (VRC-10/11/12/20/721 and others) are designed to specify interface semantics for each token class—the layering above follows the Openverse whitepaper vision and still awaits mainnet and cross-domain interoperability tests. One envisaged value flow: a VRC-20 quota token on a Layer 1 app chain routes via VTP message to Layer 0 clearing anchors, then converts to VRC-10 Bitcurrency for cross-ecosystem payment; external Ethereum USDC may likewise enter the same exchange layer via VTP/bridges—in protocol-neutral terms, the same logic maps to “app-chain quota token → value-exchange-layer routing → stable settlement asset,” unbound to a single brand.

VTP does not answer “whether a token is trustworthy”—that is collateral ratios, issuer reputation, and regulation; it answers “how assets in different trust domains find one another.” For silicon participants that distinction is critical: when an Agent calls a contract, it must parse “whether the counterparty asset type is on the allowlist,” not “whether every hop on the whole routing path is honest.” In permissioned deployments, firms may confine VTP routing to their own consortium chains and audited public-chain bridges—drawing activity corridors for silicon employees.

Lightweight VTP access means: one need not dive into every routing-message field, but must grasp the economic meaning—the Value Internet is addressable and cross-domain like email, so silicon subjects can programmatically specify payees and asset types without maintaining a separate bank account and SWIFT code for every counterparty. The “multi-currency Treasury” the CFO faces becomes, at the silicon layer, “multi-asset-type positions”: stable settlement may prefer VRC-10/11 or already-live stables such as USDC/DAI; utility and points use VRC-20; permissions and device slots use VRC-721; time commitments use VRC-13—public/private domain, fungible/non-fungible, instant/term, chosen by counterparty and delivery tense.

Section 7. Boundaries of Carbon Guardianship and Silicon Autonomy

Fully autonomous silicon economic subjects that cannot be emergency-stopped by carbon subjects will not gain legal recognition in the foreseeable period. Regulatory trends worldwide strengthen rather than weaken ultimate beneficial-owner traceability. Hence the lawful path for silicon participation in value networks is almost necessarily limited autonomy under carbon guardianship: boards answer for automated Treasury strategies; individual users answer for Agent wallets they deploy; developers answer for smart contracts they publish.

Technical implementations of limited autonomy include: spend-policy contracts (transfers only to whitelist addresses), timelocks (large amounts delayed 24 hours, guardians may cancel), multisig thresholds (enterprise Agents require dual finance and ops signatures), and upgradeable proxy contracts (migrate to new logic on vulnerability discovery without losing historical audit trails). These mechanisms are already mature on-chain; the hard part is linkage to off-chain law—will courts treat “multisig refused a transfer” as the company’s true intent? Will regulators accept on-chain logs as AML reporting basis? Answers vary by jurisdiction and are evolving slowly.

In The Denationalization of Money, Mr. Hayek envisioned competing currencies that must win holders’ trust; for silicon participants, “trust” first appears as verifiable constraint: holders (or their guardians) can read the rules, predict worst-case loss, and assign liability after accidents. On-chain rules that are public and auditable fit that standard better than black-box API charging—but public is not the same as intelligible. Contract bugs, prompt attacks, and key leaks must be mitigated item by item with limits, whitelists, and guardian pause. Silicon participants will not overnight replace carbon subjects as centers of legal liability, but will add new node types to value networks; if money and asset protocols cannot serve those nodes, the machine economy will remain blocked outside fiat-account doors.

Section 8. The Settlement Debt Left by the Information Internet

Decades of “free” or “ad-subsidized” business models on the information Internet have masked costs that cross-organization settlement should have borne. Search, social, open-source distribution—marginal copy costs near zero—yet cross-firm API calls still often require manual commercial negotiation, annual contracts, and invoice reconciliation. When an Agent tries to auto-compose ten vendors’ capabilities for one task, ten different billing dashboards and ten different KYC thresholds form a barrier harder than technical integration. Commodity circulation differs from one-shot barter: after value changes hands, the settlement chain must keep running—information-layer “free” precisely masks the fact that the value layer has not yet achieved continuous circulation.

The Value Internet will not replace that structure in the short run; it offers an optional settlement infrastructure: within circles willing to adopt on-chain standards, silicon subjects collaborate through unified asset interfaces; within circles still on fiat invoices, carbon subjects continue traditional flows. The two systems coexist long-term, linked by on-/off-ramps, corporate Treasury, and compliant custody—cross-border scenes often need the combination “on-chain speed + off-chain fiat landing.” Under coexistence pressure, where does increment more likely fall? For high-frequency, micro-amount, cross-domain, audit-needed machine-to-machine trades, on-chain protocols’ comparative advantage may amplify as instance counts rise—to be tested by pilot data, not whitepaper extrapolation; for low-frequency, large-ticket, heavily regulated human trades, fiat pipes remain the default for a long time.

Section 9. Verifiability: The Second Emergent Dimension of the Digital Age

In Principles of Economics, Menger argued that money emerges from barter rather than being decreed by legislation or contract in a single stroke: some goods, easier to divide, more portable, and demanded more widely, gain “saleability” advantages and are gradually accepted as a general medium of exchange. Saleability answers which physical properties make something more likely to take on monetary function—the first emergent dimension of classical monetary-origin theory. Chapter 6, Section 4 already operationalizes discipline on a global ledger as three layers of verifiable rules—execution–state–history²: execution verifiable (rules written into contracts then auto-execute), state verifiable (collateral ratios, balances, whitelists queryable in real time), history verifiable (transfer events permanently retained and replayable). This chapter elevates that three-layer structure to a digital-age second emergent dimension parallel to Menger’s “saleability”—verifiability: saleability governs liquidity and acceptance surface; verifiability governs whether rules can be independently rechecked by third parties (including programs). Silicon participants push the latter threshold to millisecond coordination, turning the “state layer” from institutions’ quarterly disclosure into programs’ real-time query.

For silicon participants, the book’s claim may be stated thus: when the participant set of value networks includes non-human economic subjects, the constraint unit of monetary competition must partly shift from “whether the issuer’s reputation is trustworthy” to “whether protocol rules can be independently inspected by any third party (including programs)”—Mr. Hayek wanted to change who issues notes; the silicon age must also change what constrains money. The proposition operationalizes on the three layers “execution–state–history”: $N_1$ rules publicly readable; $N_2$ execution and state independently inspectable; $N_3$ value units routable and exit-able; missing any condition, “popularization” degenerates into slogan or centralized packaging; silicon millisecond coordination amplifies missing $N_2$ into immediate counterparty risk rather than quarterly disputes that can be mediated afterward. The claim is protocol-neutral—unbound to Openverse or any single chain; USDC reserve attestation, DAI collateral-ratio on-chain queries, x402 facilitator receipts, ERC-4337 policy-contract state are different landings on the verifiability spectrum.

Reputation constraints update too slowly for millisecond M2M coordination; verifiable rules convert “discipline” from personal credit into protocol state independently recheckable by third parties (including programs). The shift is partial, not total: facilitator compliance, card-network KYC, and carbon guardianship remain reputation chains, but at the margin, whether silicon subjects can enter the default settlement set increasingly depends on whether rules are on-chain readable and history replayable.

Monetary emergence ≈ saleability ranking × verifiability threshold—the former from Menger’s dispersed selection over physical properties; the latter from three-layer verifiable rules of “execution–state–history.” Szabo (2017) captures the same institutional need as “social scalability”: when participants need not know one another to collaborate, verifiable rules substitute for interpersonal trust costs—silicon Agents push that threshold to millisecond coordination³.

When an Agent calls ten APIs in milliseconds, counterparties cannot rely on “knowing the company’s CFO” to build trust; they must rely on independently recheckable constraints—whether collateral ratios meet stated levels, whether whitelists were tampered with, whether session-key quotas were exceeded. In Menger’s era, verifiability was limited by closed ledgers and human audit costs and could not scale into emergence mechanisms; global ledgers reduce it from institutional privilege to a public technical good—“rules over persons” in silicon scenes means precisely this.

If verifiability is indeed the second emergent dimension, then after controlling issuance scale and collateral volatility, stable assets whose collateral and reserve state are real-time auditable on-chain should show systematically shorter median time $T_{\mathrm{recover}}$ to restore peg after depegging than comparable stables whose reserves are only quarterly disclosed and not independently recheckable off-chain—USDC (auditable reserves) versus UST (algorithm not independently verifiable) in 2022–2025, and MakerDAO DAI liquidation data, can serve as preliminary test material⁴. Observable metrics include: $T_{\mathrm{recover}}$ (hours), purchasing-power deviation $\Delta p$ in the 24 hours after a reputation shock, on-chain collateral-ratio query latency $\tau_{\mathrm{read}}$ (seconds). If $T_{\mathrm{recover}}$ shows no significant difference, one should narrow “verifiability determinism” and instead test whether the reputation-shock transmission coefficient $\beta_{\mathrm{rep}}$ (peg deviation amplitude within 24h of issuer negative news) rises as reserve auditability falls.

Section 10. Institutional Experiments in Silicon Settlement, 2024–2026

Over 2024–2026, engineering experiments in silicon settlement have accumulated a set of comparable instances. Frontier schemes have mostly proved “whether automatic payment is possible”; debate has shifted to “whether the constraints payment rests on are verifiable”—x402 and AgentKit cover the former; agentic commerce and M2M/ISO standards still advance at the institutional-reputation layer; DePIN (Akash/Render/Bittensor and others) contests the latter boundary at the stake-slash layer. The tables below first give the Part VI core six-line comparison (main lines the book’s silicon theory must answer head-on); then summarize constraint-carrier differences by stack position; then attach the 2024–2026 frontier panorama. Openverse columns are labeled throughout; the same logic applies equally to ERC/USDC/x402 and the like. The “observable metrics” column gives quantifiable rebuttal metrics within 12–24 months—if measured results diverge from theoretical prediction, narrow rather than extrapolate conclusions.

Core six-line comparison (2024–2026, Part VI theoretical anchors)

Silicon-settlement six-stack comparison (2024–2026)

2024–2026 frontier panorama comparison

Most frontier schemes in the tables are at pilot or draft-standard stage; none alone proves that “the silicon economy has replaced fiat pipes.” Industry already treats non-human economic subjects as new nodes on payment networks; debate has shifted from “whether automatic payment is possible” to “whether the constraints payment rests on are verifiable.” x402 advances HTTP-layer micropayments to engineering usability; AgentKit sinks wallets and on-chain tools into the application layer; Anthropic MCP (2024.11) and Google A2A (2025) standardize Agent tool orchestration but have not yet closed settlement semantics—payment authorization still sits on reputation layers such as AP2/TAP; DePIN advances compute trading onto on-chain order books—constraint carriers for all four remain mainly platform reputation, facilitator compliance, or volatile-token incentives⁵. The differentiating claim of VRC design instances is to bring stable unit of account (10/11), quota layering (20), and permission slots (721) into one auditable standard family—an engineering proposal, not to be equated with already-running pilots such as x402.

At the theory layer, the six stack positions above map to three constraint-carrier classes: reputation (x402 facilitators, AgentKit custody, agentic-commerce card networks, M2M message pipes)—payment can occur; rules hard for third parties to recheck independently; stake (DePIN slash, order-book deposits)—on-chain verifiable; off-chain delivery still weak; parameters (ERC-4337 policy contracts, PCIM collateral ratios, whitelists)—require on-chain state readability to become verifiable rules. The testable core of silicon monetary theory is: as M2M coordination frequency rises, marginal adoption of the parameter and stake layers relative to the reputation layer should rise—if 12–24 month pilot data do not support that, narrow “verifiability determinism” and instead test whether the reputation-shock transmission coefficient $\beta_{\mathrm{rep}}$ rises as reserve/policy auditability falls.

Participants in value networks are no longer only corporate and personal CFOs; they include APIs, sensors, Agents, and robots—programs and devices can likewise enter value transfer under public rules⁶.