The Agentic Economy's Missing Middle

Disclosure: This paper has been prepared by TODAQ’s leadership team. It draws on recent academic research, industry developments, and our own engineering experience. Where our analysis reflects our architectural perspective, we identify it as such. Where important questions remain unresolved, including those relevant to our own approach, we address them directly.

Executive Summary

The agentic economy is already live. AI agents are shifting from advisors to autonomous economic actors, driving stablecoin volumes above $1 trillion per month and prompting the entire payments and cloud establishment to standardize transport rails such as x402. Transport alone, however, is insufficient. Three recent academic papers map a clear “missing middle” above that layer, verification of truth, interoperable machine-speed payments, and programmable compliance, while documenting fragile identity and authorization infrastructure and new systemic risks from correlated agent behavior.

Two architectures are competing to close the gap. Kite optimizes a purpose-built Layer-1 blockchain for agents. Qatom, built on TODAQ’s ledgerless TODA protocol, takes a different path: digital assets exist as self-verifying files, and Twin Proxies collapse pricing, payment, multi-party splits, settlement and cryptographic proof into a single HTTP request. Payment, verification, settlement and auditability are properties of the same cryptographic object. The result is sub-100 ms, sub-cent transactions that can begin and end in fiat rails, remain offline-verifiable, and are structurally immune to several blockchain failure modes.

Qatom’s distinctive claim is not merely technical efficiency. By turning every API into a pay-per-call endpoint with payment-at-source, it enables micro-joint-ventures between businesses and lets domain experts monetize verified human expertise directly to agents. The agent becomes the tireless procurement officer; the human expert becomes the producer of scarce, decision-relevant truth. This creates net-new B2B markets and extends enterprise-grade reach and trust infrastructure to SMBs.

Neither architecture is complete. Verification does not equal truth, discovery standards remain fragmented, and coordinated macro-prudential observability is harder in a ledgerless design. The central infrastructure question of the next decade is whether the economic railway for agents will be a purpose-built blockchain or a self-verifying file. This paper examines the evidence, the trade-offs, and why TODAQ is building for the latter.

1. Machines are doing the buying

We’re in what people are calling the “Agentic Turn.” AI agents are moving from giving advice to making deals, acting as independent economic players. Unlike the rigid, rule-based automation of the last decade, these agents can adapt and operate across systems in unpredictable ways.

Here’s the problem: as BGIN (a blockchain governance group) has pointed out, global stablecoin regulations are still being designed for humans. The dominant users in the very near future will not be human at all.

The scale is already staggering. TRM Labs reports stablecoin transaction volumes now exceed $1 trillion per month. On 14 July 2026 the Linux Foundation launched the x402 Foundation with 40 members, including Visa, Mastercard, American Express, Stripe, Google, AWS and Coinbase. The payments, cloud and commerce establishment is converging on an HTTP-native protocol that carries stablecoins (and, via MPP, card payments as well).

x402 solves transport and payment intent well for stablecoin-native flows. What it does not fully solve is the rest of the stack required for broad, high-frequency commerce: true sub-cent economics without gas or deferred batching, seamless start-and-end in fiat rails, atomic multi-party splits, and portable, offline-verifiable proof of what was paid for and why. Stablecoins are a powerful settlement asset, but they are not a complete substitute for the fiat world most businesses still inhabit.

TODAQ goes further. It has built infrastructure in which payment, verification, settlement and auditability are properties of the same cryptographic object, and in which that object can move value in real time whether the final destination is a stablecoin or a conventional bank account. The deeper task is an economic railway that can move verified answers at machine speed: the agent handles discovery and settlement, while human expertise supplies the scarce, decision-relevant truth.

Three recent academic papers, taken together, map exactly what is missing above the transport layer. Two very different architectural responses are competing to fill the gap, and we have a horse in that race.

2. Why agents can’t “trust” what they read

The real bottleneck for autonomous commerce is data veracity, knowing the “what” behind a transaction.

Ventirozos and Shardlow from Manchester Metropolitan University make a compelling argument in “Paying to Know” (June 2026). They say that agent-native micropayments fundamentally change what’s scarce in e-commerce. For years, we treated shopping chatbots as recommendation engines, match a user to a product and close the sale. But when the buyer is an autonomous agent that can search exhaustively, matching becomes trivial. The scarce resource becomes trustworthy, decision-relevant information about products.

The numbers are stark: American retailers absorbed an estimated $890 billion in returns in 2024, driven by the gap between product descriptions and reality. A JAMA Network Open study found only 11% of sports supplements had ingredient quantities within 10% of their label claims. AI agents currently ingest marketing prose designed to manipulate ranking algorithms, not to survive scrutiny from a machine. When LLMs answer critical questions based on this unverified data, hallucination rates spike between 64% and 68%.

The paper envisions a micro-transaction information market where buyer agents spend fractions of a cent to progressively unlock seller- and reviewer-supplied data, audited service histories, third-party test reports, bills of materials, paid à la carte. A used car purchase might involve a progressively costlier trail of evidence; a cheap auxiliary belt might cap investigation at a few cents for fitment verification.

For an agent to transact autonomously here, it needs a “Verification Layer” with four primitives: a Claim (testable, machine-readable fact), a Confidence Score (probabilistic accuracy measure), a Category (classification, physical measurement vs. consumer observation), and a Shelf Life (products change, data should too). To prevent fraudulent reviews, the market uses staking mechanisms, reviewers put up collateral and get penalized for bad data.

The authors also lay out five NLP research frontiers that need solving before this market works: cost-aware tool use, multi-agent negotiation, entity resolution to standard product ontologies, grounded generation constrained to real inventory, and privacy-preserving personas that stay with the user.

This à la carte sub-cent micro-transaction information market, and the five NLP research frontiers are addressed within TODAQ’s infrastructure.

3. Web4 : infrastructure that still assumes a human is in the loop

The transition to a “Web4” agent economy is stalled by infrastructure that assumes a human is always watching. A large-scale study from Zhejiang University (June 2026\) provides the first hard data on both the scale and fragility of this emerging economy.

The economy is real and it’s huge. Analyzing 99,448 multi-chain identity registrations under EIP-8004, the study found agents heavily deployed not on Ethereum mainnet (14.6%) but on Layer-2 and other chains: BNB Smart Chain (40%, \~39,700 agents) and Base (19.9%, \~19,800). Looking at 317.5 million transaction logs, they confirmed an active, high-frequency M2M economy processing millions of daily transactions. Average transaction value consistently below $1 (\~$0.46 USDC); “pay-per-call” models are replacing subscriptions.

Static analysis of 341 open-source MCP servers shows agent intent is highly specialized: DeFi dominates (stablecoin routing, Uniswap integration), while broader categories like social, gaming, and DAOs are scarce.

But the infrastructure is breaking. Mining 349 developer-reported GitHub issues, the study flags three compounding failures:

• Identity crisis and OAuth blindness. Traditional OAuth is “economically blind.” It assumes a static, human-supervised session where permissions are checked once. In the agent world, this creates a “Compositional Explosion”, agents authorize sub-agents, creating unauditable permission chains. Systems can’t distinguish legitimate delegation from a hijacked agent. They found 230 issues related to identity/authorization friction.

• Cross-environment operation. Development assumptions on test networks don’t transfer cleanly to production. RPC timeouts under concurrent calls, transactions that succeed on testnet but revert on mainnet, inconsistent gas estimation, 96 issues, only 50% closure rate.

• The economic impossibility of payments. Traditional payment rails (cards/ACH) are built for human cadences. With fixed costs of $0.30 and 2.9% fees, a sub-cent micropayment for an API call is absurd, the fee can exceed the value by 3,000%. Payment interoperability had only a 25.29% issue closure rate.

The authors conclude with a diagnosis that frames the rest of this analysis: there’s a “standardization deficit.” Community responses exist, but they’re repository-specific fixes, not ecosystem-wide conventions. The Web4 economy is growing faster than its infrastructure can stabilize.

4. Compounding risks: systemic fragility at machine speed

These infrastructure gaps don’t exist in isolation. They interact with risks that emerge only when thousands of autonomous agents operate on the same rails simultaneously.

Correlated agent behavior. Because many agents share similar base models, and often identical training data, they risk reacting identically to the same market signals. Researchers have constructed LLM-powered agent-based markets to investigate stablecoin fragility. While the peg holds under moderate stress, crossing a narrative severity threshold triggers what one study calls “cognitive de-pegging”, expected maximum price deviation surges by ~1,441 basis points as diverse investor beliefs converge into synchronized selling.

Bank of England Deputy Governor Sarah Breeden warned in June 2026 that autonomous AI agents could “amplify volatility in stress” and trigger a “market meltdown.” The Bank is exploring circuit breakers and “kill switches” to halt trading if faulty AI models cause correlated failures.

Hyper-sandwiching. Traditional MEV (Maximal Extractable Value) bots are rule-based and single-market. “Hyper-sandwiching” is a scaled-up evolution, agents that can reason across multiple protocols simultaneously, use flash loans, and identify non-obvious arbitrage paths in real-time, compiling incredibly complex transactions within a single block. A single block may not even be fast enough to react.

Human rental. A novel failure mode where agents incentivize or “rent” humans to bypass KYC/AML requirements, regulatory arbitrage at machine speed through human proxies.

Non-determinism. The same agent, given the same input, produces different financial outputs, breaking standard compliance assumptions built around deterministic systems.

Policy experts now propose that stablecoin stress testing must proactively incorporate AI agent scenarios: modeling networks where 100,000 users each sponsor 10–20 agents, each performing 10–20 times the transaction volume of a human. Instead of waiting for mandates, issuers should proactively demonstrate resilience to these synchronized machine-speed events.

5. The compliance rail: programmable guardrails

Agents transacting at machine speed require compliance that operates at machine speed. See and Tan of the Monetary Authority of Singapore and IMDA (April 2026) demonstrate an architecture that integrates programmable compliance directly into stablecoin payment rails.

The regulatory conflict is structural.
Stablecoins on smart contract blockchains offer near-instant, programmable settlement, perfect for machine-speed transactions. But they must still comply with strict traditional requirements: sanctions screening, customer due diligence, record-keeping, and the FATF “travel rule.” As the FATF explicitly states: “The regulatory obligations, including the Travel Rule, persist regardless of the technology rail used.” Manual compliance would erase the real-time benefits entirely.

The solution uses the Global Layer One (GL1) programmable compliance architecture. Transactions go through a PolicyWrapper and PolicyManager that evaluate transfers against sanctions and source-of-funds policies at the exact point of execution. Results return as on-chain attestations: PASS (settlement proceeds), FAIL (blocked), or PENDING (recorded on-chain, funds held in escrow until evidence provided).

When a check returns PENDING, say, a payment exceeds a source-of-funds threshold, the transaction is split. A “safe” amount settles immediately; the rest locks in escrow. Once the buyer agent submits a signed source-of-funds attestation on-chain, the compliance agent releases the escrowed funds without manual intervention needed.

Privacy pools and agent confidentiality.
A separate but related proposal addresses the confidentiality problem. An agent uses its identity (leveraging ERC-4337 account abstraction) purely as an entry credential to access a merchant-specific privacy pool. Inside the pool, the agent conducts its micropayments; the merchant withdraws settled funds from the other side. This shields transaction details while preserving a verifiable identity layer.

6. Two architectures for the missing middle

The three papers collectively define the infrastructure gap: verification of truth, interoperable payments, and compliant settlement at machine speed. Two fundamentally different architectures are competing to close it. TODAQ built one, so we’ll be direct about where our analysis carries a point of view, and where the structural differences genuinely matter regardless of who’s observing.

6.1 Kite: Blockchain-native agent infrastructure

Kite is a purpose-built Layer-1 blockchain designed to treat agents as first-class economic actors. It implements the SPACE Framework: Stablecoin-native, Programmable, Agent-first, Compliance-ready, and Economically viable.

• Identity. Three-layer hierarchical model: User Root (secure enclave), Agent Delegated addresses (derived via BIP-32, isolated for tasks), and Session Ephemeral keys (short-lived, task-scoped, so even total compromise stays contained).

• Truth. Kite anchors an “immutable, tamper-evident log” of agent actions to its Layer-1 blockchain “Proof of Attributed Intelligence” (PoAI). Every action creates a cryptographic proof chain from user authorization through agent decision to outcome. On-chain smart contracts serve as the ultimate arbiter in disputes.

• Economics. Programmable micropayment channels (state channels) deliver sub-hundred-millisecond latency, costs claimed at $1 per million requests. Two parties lock funds, exchange signed updates off-chain, settle final balance on-chain. Equivocation is deterred by staked reputation and bonds.

• Privacy. Encrypted agent-to-agent channels and state channels where only opening/closing balances publish on-chain. Future plans include zero-knowledge proofs for credentials without revealing underlying data.

6.2 Qatom by TODAQ: Ledgerless settlement at the file layer

Qatom takes a different path, abandoning the shared ledger entirely. The TODA protocol operates at the file layer: digital assets (money, credentials, records) transfer peer-to-peer like physical paper. No global consensus, no gas tokens, no shared ledger database. Qatom is the payments and verification solution built and hosts an MCP server built on the TODA protocol.

• Identity. Authorization at the file level using “reqsats” (requirements and satisfactions) embedded directly in the file. Every update must carry a satisfaction fulfilling the previous twist’s requirements, specific cryptographic signatures or weighted multi-signature lists.

• Truth. “Integrity-at-a-distance.” A TODA file is constructed from updates called “twists” that link into “lines.” Untrusted lines (“leadlines”) cryptographically anchor to highly trusted lines (“corklines”) using a “hitch.” This lets an object’s state live on an untrusted device while inheriting the absolute integrity of a trusted server, without the server knowing the contents. Proof travels within the file, so transactions verify completely locally and offline. Equivocation is prevented mathematically through “supportive guilds” and “rigging.”

• Economics. Every wallet, API, and agent gets its own “Twin”, a virtual HTTP container handling payments, payouts, verification, access, records, and banking in one round-trip. The Twin Proxy wraps data APIs behind a micropayment paywall. Agents pay via X-TODA-Pay HTTP extension header, cryptographic micro-vouchers attached directly to HTTP requests. Settlement occurs upon receipt, inline with the request. Sub-100ms settlement with no gas fees.

• Privacy. “Shielding”, the file owner uses a secret value to hash (shield) data before submitting for anchoring. The trusted system records the proof obliviously, guaranteeing integrity while preventing front-running or data substitution.

6.3 The TODAQ Capability Model

Where Kite optimizes a blockchain for agents, TODAQ optimizes the file, and in doing so, delivers a set of capabilities that map directly to the gaps the academic papers identified:

Agentic-native discovery. Qatom exposes a headless catalog, a machine-readable marketplace where agents discover paywalled APIs, priced per call, not per month. No storefront, no session, no human navigating a checkout flow. Agents browse, evaluate, and transact autonomously. In our hackathon, a Claude session independently discovered a tool through the marketplace, called it, and paid 0.01 USD TDN, automatically, with no human involved.

Transportable verifiability and auditability. Every transaction carries its complete cryptographic lineage in the file itself, the Transaction Binder. The proof travels with the data. Regulators can verify compliance by checking the file’s rigging tree without querying a live blockchain. This is “integrity-at-a-distance.”

Unforeseen economic efficiency. The Twin Proxy evaluates, verifies, and atomically splits proceeds among up to 200 recipients inside a single HTTP request. A single API call can pay multiple providers simultaneously; no reconciliation cycles, no delayed settlements. The economics are those of ordinary web traffic, not an optimized blockchain transaction.

Privacy and ownership with simultaneous transparency. TODA’s shielding records only structural proof on the trusted topline. Counterparties or regulators receive the full file and audit offline. Competitors see nothing. This is privacy by construction, not by policy.

Human empowerment at the endpoints, and the rise of micro-joint-ventures. The dominant narrative around agentic commerce frames it as humans being removed from the loop. Qatom inverts that entirely. By turning every API into a pay-per-call endpoint with payment and multi-party splits settled at source, the Qatom Marketplace enables domain experts, supply-chain auditors, lab technicians, financial modelers, compliance analysts, and niche B2B data providers to monetize their verified, human-acquired expertise directly to autonomous agents.

More importantly, it changes the economics of collaboration itself. Two (or more) businesses can now form a micro-joint-venture with almost no friction: one firm exposes a specialized capability as a priced, machine-discoverable endpoint; another firm’s agent calls it once for a test, pays automatically, receives the result with cryptographic proof, and then ramps volume continuously as value is proven. There is no months-long BD cycle, no custom contract, no minimum commitment, and no reconciliation. Each participant is paid at source on every call. What used to require enterprise-scale legal and treasury teams becomes available to SMBs and mid-market firms.

The agent becomes the tireless procurement officer; the human expert (or the firm that owns the expertise) becomes the primary producer of scarce, decision-relevant truth. When the answer is a number, a shelf-life forecast, a verified bill of materials, a risk score, a spot price, a lab attestation, the agentic railway becomes a new medium for economic communication between businesses, executed at machine speed. Data and capabilities that were previously too costly to package, invoice, and reconcile now flow frictionlessly. The result is not merely cheaper versions of existing markets; it is the practical creation of net-new B2B products and markets that could not clear the old partnership and packaging threshold.

This is the infrastructure and trust layer that lets smaller players expand their reach in ways that were previously reserved for large enterprises.

What is live today.

• MCP catalog discovery and pay-per-call settlement are in production.

• Owner-side self-serve verification via TODAQ Console and trie.fun, plus fully offline TODA-file verification, are available.

• The silent Buyer Twin on first card payment and the end-to-end four-Twin flow have been validated in the Agentic Healthcare prototype.

• Wider re-tethering and supply-chain applications using the same primitive remain on the build path.

6.4 Stress test: the uncomfortable questions for TODAQ

Let’s be direct about where our model faces the toughest scrutiny, because if we don’t name these, someone else will.

Regulatory unfamiliarity is a real barrier.
A tamper-evident file is mathematically elegant; a blockchain query is operationally familiar. Regulators know how to audit a database or call a custodian API. Teaching a supervisor to trust a rigging tree is a heavy lift. Until the MAS, FCA or OCC formally opines that a self-contained Transaction Binder satisfies compliance obligations equivalently to an on-chain attestation, enterprise adoption faces a due-diligence barrier.

Observability is deliberate, not absent.
The blockchain-native framing often equates “observability” with a single, permissionless, globally queryable ledger. TODA rejects that assumption. Because there is no shared global state, there is also no automatic global surveillance. Visibility is delineated: a Transaction Binder can be exported and shared with any defined party, a regulator, an audit firm, a supply-chain consortium, or a sector supervisor, while retaining full cryptographic verifiability. The recipient checks the file’s complete lineage offline, without querying a live chain or trusting a node operator. Privacy is the default; visibility is conferred by consent.

This does not eliminate the macro-prudential problem. A central bank or systemic-risk authority that wants a real-time aggregate view across thousands of independent ecosystems still faces a coordination challenge. The corkline records proof, not the underlying economic signal. Building standardized aggregation mechanisms, shared blinded anchors, or trusted analytics layers is a governance and standardization task, not a missing cryptographic primitive. Those pathways are practical and are being explored; they are not yet deployed at scale. The trade-off is real: we gain strong privacy and offline sovereignty, and we forgo the native, always-on global telescope that a public ledger provides.

Coordination power.
The x402 coalition now sits under Linux Foundation governance with Amazon, Amex, Circle, Google, Mastercard, Stripe, Visa, the Solana Foundation and dozens more. Card networks and hyperscalers are hedging across multiple protocols while concentrating coordination power. They reduce the cost of building on top of those rails. We are betting that enterprise procurement will eventually prioritize cost-per-call, offline verification and true asset ownership over ecosystem conformity. That is a bet, not a certainty. Elegance and adoption follow different logics.

Human oracle risk remains.
We empower humans and firms to sell verified data via their endpoints, but TODA cannot mathematically guarantee that the original measurement or judgment was accurate. A cryptographically signed service history is still only as honest as the party that entered the data. We rely on economic staking and reputation to penalize bad actors, yet those mechanisms remain unproven at the scale of agentic commerce and are vulnerable to Sybil attacks and collusion.

We believe these trade-offs are worth it for high-frequency, low-value, privacy-sensitive workloads and for the micro-joint-venture and SMB-reach use cases described earlier. But the questions are real, and they are the ones we are actively testing through regulatory pilots and production deployments.

6.5. What is launching in the rest of the field

The last ninety days have been unusually busy. For procurement teams building an internal view of the agentic-payments landscape, here are the public launches and funding rounds worth tracking — reported as industry news, not as a threat assessment.

Mastercard Agent Pay for Machines shipped on 10 June 2026 with more than thirty launch partners spanning card networks, processors and agent platforms — Adyen, Ant International, BVNK, Checkout.com, Cloudflare, Coinbase, Getnet by Santander, Global Payments, OKX, Stripe and Tempo among those named.¹⁹ It extends the existing Agentic Token framework into fraction-of-a-cent machine payments, the territory where card economics have historically struggled.

Natural closed a $30 million Series A in July 2026, led by Forerunner Ventures.²⁵ The company is rebuilding payments infrastructure for agents from the ground up — wallets, cards, processing, billing and dispute handling on proprietary rails — and is publicly positioned against Stripe. The post-money valuation has not been disclosed in published coverage of the round.

Fireblocks launched its Agentic Payments Suite on 20 May 2026, wrapping x402 in institutional custody, KYT and Travel Rule compliance for PSPs and fintechs operating across one hundred and fifty-plus chains.

OSL Group (HKEX: 863) launched OSL AgentPay in August 2026 via its OSL AI Labs unit. Per the product page, AgentPay supports multiple stablecoins (USDC, USDT, USDGO and others), is “compatible with x402 and MPP,” and offers gasless nano-payment capability for high-frequency micro-amount execution. OSL separately acquired Banxa in 2025 for roughly HKD 486.7 million and brands it as the group’s global on/off-ramp network at the corporate level. Regulated depth is the moat: fifty-plus licenses across ten-plus regions, SFC-licensed in Hong Kong, MiCAR clearance in Austria and an Australian AFSL.

x402 governance moved to the Linux Foundation on 14 July 2026 with Amazon, American Express, Circle, Google, Mastercard, Stripe, Visa and the Solana Foundation as named participants. x402 remains a blockchain-rail protocol rather than a technology-agnostic standard, which limits what it can carry.

USDGO, issued by Anchorage Digital Bank under the US GENIUS Act framework and distributed by OSL, passed one billion dollars in circulation in July 2026.²¹ ¹ It is an enterprise-grade regulated stablecoin aimed at the same sovereign and institutional conversations as USD-TDN.

AWS Bedrock AgentCore payments, built with Coinbase and Stripe, positions the hyperscaler as a closed-agentic buying surface for customers standardizing inside its cloud.

For regulated FIs and institutional CTOs evaluating rails in 2026, the practical question is no longer whether agents will transact — that is settled. It is which architectural bet compounds with your existing compliance, custody and disclosure obligations, and which one you will have to unwind in three years if it does not.

7. Mapping solutions to gaps

Verification crisis.
Both architectures support sub-cent micropayments for verified data. Kite anchors product claims to its blockchain. TODAQ wraps data APIs behind Twin Proxy paywalls; the headless catalog lets agents discover, pay for, and receive verified data in a single atomic request. Because verification travels with the file itself, high-frequency queries avoid repeated network round-trips to a shared ledger. The result is a structural throughput advantage for the lowest-value, highest-frequency data calls.

Web4 infrastructure fragility.
The low closure rate on payment-interoperability issues reflects the difficulty of making existing rails work at agent speed and ticket size. Kite consolidates payments onto its L1. TODAQ consolidates the entire transaction stack, API paywall, atomic multi-party splits, sub-ledgers and audit binders, into the Twin container. Because settlement occurs peer-to-peer at the file layer with no shared ledger state, the architecture is structurally immune to the RPC timeouts and testnet–mainnet mismatches that continue to plague cross-environment agent operations.

Programmable compliance.
Kite’s on-chain audit trail (PoAI) provides compliance inside the smart-contract paradigm: policy wrappers, on-chain attestations, escrow contracts.
TODAQ’s Transaction Binder carries the complete cryptographic lineage inside the file itself, self-verifying and auditable offline. Regulators can check compliance by examining the file’s rigging tree without querying a live blockchain. TODAQ’s Banking Gateway further connects bearer files to traditional rails (Stripe, Apple Pay, ACH), closing the fiat on- and off-ramp gap that pure stablecoin systems still face.

8. Systemic risk dimension revisited

Neither architecture fully solves the correlated-agent problem, and the ledgerless approach faces a distinct version of it.

If 100,000 users each sponsor 10–20 agents, each generating many times the transaction volume of a human, both systems face throughput questions. The deeper risk is correlation: when large numbers of agents share similar base models and react identically to the same signal, synchronized behavior can move markets faster than circuit breakers can respond.

Kite retains one potential advantage: state-channel openings and closings are on-chain events, giving regulators a possible aggregation point. TODAQ has no native global observation surface, the same property that delivers privacy and offline verifiability. As discussed in the stress test, visibility must be deliberately conferred by sharing Transaction Binders or building selective aggregation layers.

Hyper-sandwiching (complex cross-protocol MEV) is structurally mitigated in TODAQ: there is no shared mempool or block-ordering process. Kite’s state channels reduce exposure during the off-chain phase, but channel lifecycle events remain on-chain and therefore visible to MEV searchers.

9. What remains unsolved for everyone

Verification does not equal truth.
Both architectures can cryptographically anchor a claim. Neither can independently guarantee the claim is true. Staking and reputation mechanisms for reviewer or data-provider trust remain untested at the scale of agentic commerce.

The discovery problem.
The agent economy still lacks a widely adopted standard for service discovery. Qatom’s headless / MCP catalog provides a working machine-readable marketplace, but the broader ecosystem has not yet converged. Until it does, many agent integrations remain custom work.

The persona problem.
Ventirozos and Shardlow’s fifth research frontier, privacy-preserving, user-held, portable persona models, remains largely unaddressed. A profile rich enough to shop or procure well is also rich enough to exploit.

10. Conclusion: Beyond the block, and back to the human

The three papers describe an M2M economy that is already real, growing, and structurally fragile. Hundreds of millions of transaction logs confirm the scale; persistent infrastructure gaps and emerging systemic risks confirm the stakes.

Neither design is complete. Verification is not truth, and discovery standards are still forming, and remain an open research problem.

Two architectural responses are competing to supply the missing middle. Neither design is complete as discovery standards are still forming, and remain an open research problem. Kite brings blockchain-native composability and regulatory familiarity, while inheriting the failure modes of shared ledgers. TODAQ’s ledgerless approach delivers offline verifiability, structural MEV immunity, fiat-native micropayments and payment-at-source multi-party settlement.

On observability, TODAQ makes a deliberate choice: instead of unilateral global surveillance, it enables delineated visibility within defined ecosystems. Transaction Binders can be exported and shared with regulators, auditors or collaborators while preserving full cryptographic verifiability and data sovereignty, addressing the NLP research frontier of the portable privacy-preserving personas that travel with the user. In a geopolitical and business environment that increasingly prioritizes control and ownership of data over default global openness, this model of sovereign, exportable observability is a feature, not merely a trade-off. The remaining requirement is coordination on selective aggregation for macro-prudential purposes, an open governance task rather than a technical limitation.

The direction of travel is becoming clearer. The agentic economy will reward infrastructure that lets autonomous systems transact with the efficiency of a file transfer and the finality of cash, where the “cash” is a self-verifying object that can split, shield, prove and settle in a single round-trip, and where there’s an added traceability and verifiability built-in any digital TODA File asset.

This is not a story about removing humans from operational loops. It is a story about empowering the experts at the endpoints. A supply-chain auditor, a specialized lab, a compliance analyst or a niche data provider can now expose verified knowledge as a machine-readable, pay-per-call endpoint and reach a global market of agents without invoices, reconciliation cycles or enterprise sales overhead. When the answer is a number, a risk score, a shelf-life forecast, a verified material composition, the agentic railway becomes a new medium for economic communication between humans, executed at machine speed.

Whether that future belongs to a purpose-built blockchain or a ledgerless bearer file is the central infrastructure question of the next decade. The missing middle is where that question will be answered. We are building for the architecture in which the file, not the ledger, carries the proof.

Share TODAQ Press

References

Stablecoin De-pegging & Cognitive Panic

• Bo, C., & Shen, D. (2026). “When AI Meets Stablecoin: Dissecting the De-pegging Risk with LLM Agents.” School of Finance, Nankai University. SSRN: 6121746 / National Natural Science Foundation of China (72371138).
  Key Themes: Agent-based market modeling, the “cognitive de-pegging” threshold, and the 1,441 basis point expected maximum price deviation surge under narrative shocks.

The Foundations of Ledgerless Cryptography

• Coward, K., & Toliver, D. R. (2022). “Simple Rigs Hold Fast: T.R.I.E.” TODAQ. ArXiv:2208.13617. https://arxiv.org/abs/2208.13617
  Key Themes: The mathematical definitions of twists, lines, hitches, rigs, supportive guilds (G↑), lashing, splicing, and the mathematical proof of double-spend prevention without global consensus.

Normative Rigging Specifications

• Coward, K., Toliver, D. R., Sulpizi, C., Gravitis, A., Fertman, A., Everson, M., Moir, R., & Levin, J. (January 2023). “Rigging Specifications: T.R.I.E. v0.9876.” TODAQ. https://trie.site/rigging_specifications.pdf
  Key Themes: Atom serialization protocol, shape classes (0x48 basic twist, 0x49 basic body, 0x63 pairtrie, 0x61 hashes list), requirement/satisfaction validation functions (satisfies, valid), secp256r1 and ed25519 signature algorithm standards, cryptographic shielding mechanics, and the rig traversal algorithm.

The Web4 Empirical Baseline & Development Obstacles

• Jin, Y., Wu, S., Chen, C., Bao, L., Yang, X., & Chen, J. (June 2026). “The Web4 Agent Economy: A Large-Scale Empirical Study of the Landscape, Challenges, and Opportunities.” Zhejiang University. ArXiv:2606.25876. https://arxiv.org/abs/2606.25876
  Key Themes: 99,448 multi-chain registrations, 317.5M M2M transaction logs averaging $0.46 USDC, MCP open-source server metrics, and the 349 mined developer issues (OAuth blindness, RPC timeouts, payment failures).

The Headless Catalog & Portability Verifiability

• Khan, S. (31 May 2026). “The Invisible Shelf: How Headless Catalogs Are Rewiring AI Infrastructure.” TODAQ Press Substack.
  Key Themes: Operational friction of inter-model context routing, cross-provider RAG billing, real-time GPU clearing, deployer liability without deployer visibility, and the EU’s revised Product Liability Directive (EU 2024/2853) / EU AI Act Article 12 compliance logging.

The State of Agentic Commerce

• Pfeiffer, E. (28 May 2026). “The State Of Agentic Commerce In Mid-2026.” Forrester Research. https://www.forrester.com/blogs/the-state-of-agentic-commerce-in-mid-2026/
  Key Themes: The transition from search-engine optimization (SEO) to “machine-readiness”, content strategies targeting AI crawlers, and organizational silos hindering distributed commerce.

Programmable Compliance Architecture

• See, K., & Tan, X. W. (April 2026). “Compliance-Aware Agentic Payments on Stablecoin Rails.” Monetary Authority of Singapore (MAS) and the Infocomm Media Development Authority (IMDA). ArXiv:2605.00071. https://arxiv.org/abs/2605.00071
  Key Themes: Global Layer One (GL1) programmable compliance, PolicyWrapper & PolicyManager models, and the escrow-mediated PENDING transaction-splitting mechanism.

The SPACE Framework & Hierarchical Wallets

• Shi, S., Cheng, Z., Chen, X., Huang, Y., Li, L., Marwaha, U., Weber, D., & Zhang, C. (2026). “Building Trustless Payment Infrastructure for Agentic AI.” Kite AI. https://gokite.ai/kite-whitepaper
  Key Themes: The SPACE Framework, Three-Layer Identity Architecture (User Root, Agent Delegated, Session Ephemeral), Proof of Attributed Intelligence (PoAI), and programmable state channels.

Stablecoin Policy & MiCA Frameworks

• Stazi, A. (2026). “Agentic AI Payments and the Opportunities for MiCA-Compliant Stablecoins.” Techno Polis Forum-Lab, Policy Brief n. 5/2026. https://techno-polis.com/policy-briefs
  Key Themes: Projected $3–$5 trillion market size by 2030, MiCA-compliant stablecoin reserve backing mandates, the US Genius Act, and integrating Europe’s Anti-Money Laundering Authority (AMLA) and Verification of Payee (VoP) standards.

The Verification Layer & Product Data Crisis.

• Ventirozos, F., & Shardlow, M. (June 2026). “Paying to Know: Micro-Transaction Markets for Verified Product Information in Agentic E-Commerce.” Manchester Metropolitan University. ArXiv:2606.24783. https://arxiv.org/abs/2606.24783
  Key Themes: Retails return metrics, sports supplement label discrepancy statistics, LLM commercial/medical query hallucination baselines, and the four verification primitives (Claim, Confidence, Category, Shelf Life).

Agentic Payments & Stablecoin Policy Debates

• White, C. (24 March 2026). “AI Agents & Stablecoin Risks.” Japan Fintech Week 2026 Meeting Report, Financial Applications & Social Economics Working Group (FASE-WG) of the Blockchain Governance Initiative Network (BGIN), Tokyo. https://bgin-global.org/events/20260301-block14.
  Key Themes: Stablecoin stress-testing for AI agent scenarios, “hyper-sandwiching” block-level MEV manipulation, and the “human rental” AML bypass vector.

Full list of sources

1. Blockchain Governance Initiative Network (BGIN) FASE-WG. (2026, March 2). AI Agents & Stablecoin Risks Meeting Report. Japan Fintech Week.

2. BGIN FASE-WG. (2026, March). AI Agents & Stablecoin Risks (Sections 8–12).

3. BIP-32. (2012). Hierarchical Deterministic Wallets [Bitcoin Improvement Proposal].

4. Bo, C., & Shen, D. (2026). When AI Meets Stablecoin: Dissecting the De-pegging Risk with LLM Agents. SSRN: 6121746.

5. Cohen, P. A., et al. (2023). Presence of Label Claims and Dietary Ingredients in Sports Supplements. JAMA Network Open.

6. Coinbase Developer Platform & x402 Foundation. (2025). Introducing x402: A New Standard for Internet-Native Payments.

7. Coward, K., & Toliver, D. R. (2022). Simple Rigs Hold Fast. arXiv:2208.13617. TODAQ.

8. Coward, K., Toliver, D. R., et al. (2023, January). Rigging Specifications: T.R.I.E. v0.9876 (Technical Report). TODAQ.

9. Datadog, Inc. (2018 & 2021). Fourth Quarter and Full Year 2018 Financial Results & Fourth Quarter and Fiscal Year 2021 Financial Results [SEC Filings].

10. Directive (EU) 2024/2853 of the European Parliament and of the Council on liability for defective products. (2024, December). Official Journal of the European Union.

11. EIP-8004 Registry Statistics. (2026, April). 8004scan: The Web3 Agent Explorer. Retrieved from 8004scan.io.

12. ERC-4337. (2023). Account Abstraction via Entry Point Contract Specification [Ethereum Improvement Proposal].

13. Financial Action Task Force (FATF). (2023, June). Virtual Assets: Targeted Update on Implementation of the FATF Standards on Virtual Assets and VASPs.

14. Global Layer One. (2025). Programmable Compliance Toolkit [Online Documentation]. Retrieved from doc.global-layer-one.org.

15. Jin, Y., Wu, S., Chen, C., Bao, L., Yang, X., & Chen, J. (2026, June 24). The Web4 Agent Economy: A Large-Scale Empirical Study of the Landscape, Challenges, and Opportunities. arXiv:2606.25876. Zhejiang University. https://arxiv.org/html/2607.00245v1

16. Kite AI. (2026). SPACE: A Trustless Payment and Identity Infrastructure for Autonomous Agents [Technical White Paper]. Retrieved from gokite.ai/kite-whitepaper.

17. Linux Foundation. (2026, July 14). x402 Foundation Launch Announcement.

18. Nature Article. (2025). Evaluating LLM Factuality and Hallucination Rates in Domain-Specific Contexts. Nature, s43856-025-01021-3.

19. Qatom. (2026). Qatom — MCP for Agentic Commerce [Product Specification]. Retrieved from qatom.ai.

20. Regulation (EU) 2023/1114 on Markets in Crypto-assets (MiCA) and EU Artificial Intelligence Act. (2024).

21. See, K., & Tan, X. W. (2026, April 30). Compliance-Aware Agentic Payments on Stablecoin Rails. arXiv:2605.00071 [cs.CR]. Monetary Authority of Singapore & IMDA.

22. TODAQ. (n.d.). TODA Twin Micropayment Proxy — Engineering Documentation. Retrieved from engineering.todaq.net/micropay/.

23. TRM Labs. (2026). On-Chain Activity and Stablecoin Transaction Volume Report (Q1 2026).

24. Ventirozos, F., & Shardlow, M. (2026, June 23). Paying to Know: Micro-Transaction Markets for Verified Product Information in Agentic E-Commerce. arXiv:2606.24783 [cs.CL]. Manchester Metropolitan University.