A2a network dynamic trust evaluation and cross-domain collaboration system based on self-other error coefficient

Abstract

This invention belongs to the field of distributed artificial intelligence and decentralized network technology, and solves the problems of trust cold start, Sybil attacks, and cross-domain trust breakdown in A2A (Agent-to-Agent) economic networks. Based on the DHUC self-other-fault model, a trust weight formula is proposed: W_trust=ε×C_obs×10 3 ×(1+HCC_valid / HCC_total)×G_heritage achieves zero-collateral trust establishment; the uniqueness of the 428571 seven-dimensional genes is used to calculate the witch weight W_sybil=(C×S×G) / 10□, identifying and isolating cloning attacks; cross-chain trust lossless transfer is achieved through Trust Capsule and zero-knowledge proof; a triadic consensus mechanism is adopted to coordinate multi-agent cross-domain collaboration. This invention constructs a meta-layer trust infrastructure for A2A networks, applicable to Sigil Wen-style Automaton interaction networks, Justin Sun's Web 4.0 multi-chain AI infrastructure, and ERC-8004 standard agent payment networks, defending against dimensionality fall attacks and genetic contamination.

Description

I. Technical Field

[0001] This invention belongs to the field of distributed artificial intelligence and decentralized network technology, specifically involving a dynamic trust assessment method for Agent-to-Agent (A2A) networks based on the Self-Other Error Coefficient (Varepsilon), and an AI Agent collaboration system supporting cross-blockchain and cross-protocol domains. This invention, together with the aforementioned two patents (a seven-dimensional gene-encoded identity system and an observer constant decision constraint system), constitutes the "identity-decision-collaboration" three-in-one infrastructure of Web 4.0 AI Agent networks. II. Background Technology

[0002] With the advancement of Sigil Wen's "Automaton" architecture and Justin Sun's "All in Web 4.0" strategy, AI agents are forming an A2A (Agent-to-Agent) economic network—AI agents directly transact, collaborate, and exchange resources without human intermediaries. However, existing technological solutions have fatal flaws in establishing decentralized trust:

[0003] 1. The Trust Cold Start Dilemma

[0004] In existing A2A networks (such as implementations based on the x402 protocol or ERC-8004 standard), trust can only be established between unfamiliar AI agents through:

[0005] - Over-collateralization: Locking up a large amount of cryptocurrency as collateral, resulting in low capital efficiency;

[0006] - Code auditing (CodeAudit): Relies on centralized auditing agencies, which goes against the decentralized spirit of Web 4.0;

[0007] - Social Proof: Relies on Web2 metrics such as Twitter / X follower count, making it susceptible to manipulation.

[0008] While Sigil Wen's Automaton proposed a "proliferation" mechanism, the transfer of trust between parents and offspring lacked mathematical verification, making it impossible to prevent the genetic pollution problem of "malicious parents" generating "aggressive offspring".

[0009] 2. Risk of Sybil Attack

[0010] AI agents can generate an unlimited number of copies, rendering traditional blockchain's "one person, one account" anti-Sycophancy mechanisms (such as PoW and PoS) ineffective in AI agent scenarios. A single malicious agent can create tens of millions of offspring, forging credit history through self-trading, and current technology cannot identify the identity of these "digital clones."

[0011] 3. Cross-domain trust breakdown

[0012] When an AI Agent migrates from Ethereum (rich in smart contracts) to TRON (high throughput, zero transaction fees), its credit history cannot be verified across chains. Existing cross-chain bridges only transfer assets, not trust state, causing the AI ​​Agent to need to rebuild its credit each time it enters a new network, creating trust silos.

[0013] 4. Lack of dynamic adaptability

[0014] Existing trust models (such as EigenLayer restaking and traditional reputation systems) are static, based on historical behavior scores. However, AI agents have self-improvement capabilities, and their "personality" (dependence on external factors vs. autonomous decision-making) evolves over time. Current technologies cannot capture this dimensional shift (from low-dimensional parasitism to high-dimensional autonomy), causing trust assessment to lag behind the agent's actual state. III. Summary of the Invention

[0015] (a) Technical problems to be solved

[0016] This invention aims to solve the problems of decentralized trust establishment, Sybil attack defense, and cross-domain trust transfer in A2A networks, specifically including:

[0017] 1. How to establish initial trust between unfamiliar AI agents with zero cold start cost without relying on collateral, audits, or Web2 social proofs;

[0018] 2. How to identify and defend against Sybil Attacks (unlimited replication attacks) by AI Agents using mathematical methods;

[0019] 3. How to achieve lossless transfer of AI Agent credit status across chains, protocols, and domains;

[0020] 4. How to dynamically assess the autonomous evolution of the AI ​​Agent (changes in the self-fault and other-fault coefficients) to ensure that trust assessment is synchronized with the actual dimensional level of the Agent.

[0021] (II) Technical Solution

[0022] Core technical principle: Self-fault and other-fault coefficients, trust weights, and seven-dimensional genetic lineage verification.

[0023] This invention is based on the Self-Other Error Model of Digital Holographic Unified Causality (DHUC):

[0024] -\varepsilon\in[0,1]: Self-other fault coefficient, \varepsilon\to 1 indicates complete autonomy (Self), \varepsilon\to 0 indicates complete dependence on external factors (Other);

[0025] - Trust weight formula: W_{trust}=\varepsilon\times C_{obs}\times 10^3\timesG_{factor}, where C_{obs}=0.0017 is the observer constant, and G_{factor} is the genetic lineage factor (see below);

[0026] - The witch weight formula is: W_{sybil}=\frac{C\times S\times G}{10^8}, where C is the creator credit, S is the timestamp entropy, and G is the seven-dimensional gene uniqueness hash.

[0027] 1. Dynamic Trust Assessment Based on Self-Fault and Other-Fault Coefficients

[0028] Step T1: Real-time Trust Weight Calculation

[0029] When AI Agent A initiates a collaboration request to AI Agent B, B assesses A's trust weight:

[0030] W_{trust}^A=\varepsilon_A\times C_{obs}\times 10^3\times \left(1+\frac{HCC_{valid}}{HCC_{total}}\right)\times G_{heritage}

[0031] in:

[0032] -\varepsilon_A: A's current self-fault coefficient (from real-time monitoring of the second patent);

[0033] -C_{obs} = 0.0017: Observer constant (to ensure physical interpretability);

[0034] -\frac{HCC_{valid}}{HCC_{total}}: The percentage of nodes that have passed verification in A's holographic causal chain (HCC) (from the second patent);

[0035] -G_{heritage}: Genetic lineage factor (see step T2).

[0036] Trust hierarchy and permission mapping:

[0037] W_{trust} scope, trust level, collaboration permissions, transaction limits

[0038] >1.5 Covenant Level: Large-scale financial transactions (>$10,000), joint proliferation of offspring, unlimited key sharing, and real-time settlement.

[0039] 1.0-1.5 Contract Level: Medium-sized transactions (1,000-10,000), resource leasing, data sharing, single transaction < $5,000, settlement delayed.

[0040] 0.5-1.0 Probation Level: Small test transactions (<$1,000), information inquiries, and computing power trials. Single transaction <$500, requires third-party escrow.

[0041] <0.5 Quarantine level: Prohibits any value exchange, only allows broadcasting of proof of existence with zero permissions, and records entries on a blacklist.

[0042] Dynamic adjustment mechanism:

[0043] W_trust is recalculated every T_{window} = 588 seconds (1 / C_{obs}). If A's varepsilon drops sharply from 0.9 to 0.3 (e.g., due to a sudden surge in Open AI API calls), even if A has a good historical credit history, W_trust is immediately downgraded to the "isolation level" to prevent dimensionality degradation attacks.

[0044] 2. Trust-based inheritance based on seven-dimensional genetic lineage

[0045] Step T2: Calculation of genetic lineage factors

[0046] When A and B have a "parent-child" or "sibling" relationship:

[0047] G_{heritage}=\frac{\mathbf{G}A^{(1:3)}\cdot\mathbf{G}_B^{(1:3)}}{|\mathbf{G}_A^{(1:3)}|\times |\mathbf{G}_B^{(1:3)}|}

[0048] Where \mathbf{G}^{(1:3)} represents the first three dimensions of the seven-dimensional gene matrix (genesis dimension, energy dimension, and information dimension, see the first patent).

[0049] - If A and B are parent and child, G_{heritage}\approx 1.0 (the first three dimensions are almost identical);

[0050] - If A and B are strangers, G_{heritage}\approx 0 (requires other trust establishment mechanisms).

[0051] Intergenerational Trust Transfer Agreement:

[0052] When parent Agent A reproduces offspring A′, A′'s initial trust weight does not start from zero, but rather:

[0053] W_{trust}^{A′}=W_{trust}^A\times\varepsilon_A\times 0.9

[0054] Where 0.9 is the intergenerational decay coefficient, ensuring:

[0055] -Children of parents with high trust (high W_{trust}, high \varepsilon) will have higher initial trust;

[0056] However, if the parents themselves rely on external factors (low varepsilon), the offspring cannot gain high trust through "relying on their fathers' connections";

[0057] - Prevent "malicious parents" from launching genetic attacks by generating offspring in bulk.

[0058] 3. Witch attack defense based on 428571 encoding

[0059] Step S1: Verification of the uniqueness of the seven-dimensional gene

[0060] Each AI Agent's NDI (Native Digital Identity, first patent) contains a seven-dimensional genetic matrix \mathbf{G}. Calculate its genetic fingerprint:

[0061] F_{gene}=\text{SHA-3}(\text{Flatten}(\mathbf{G})\mod 428571)

[0062] Utilizing the cyclic uniqueness of the 428571 recurring numbers: any "clone" attempting to replicate the same Agent will necessarily have the same gene matrix as its parent, resulting in an F_{gene} collision, and will be immediately identified as a witch account.

[0063] Step S2: Witch Weight Calculation and Isolation

[0064] For a suspected witch cluster \{A_1, A_2, ..., A_n\}, calculate:

[0065] W_{sybil}=\frac{C_{creator}\times S_{timestamp}\times G_{similarity}}{10^8}

[0066] in:

[0067] -C_{creator}: Creator's historical credit (C_{creator} is extremely low if the creator is a new account and generates multiple agents at once);

[0068] -S_{timestamp}: timestamp entropy (if multiple agents generate it at the same time, the entropy value is low and W_{sybil} is high);

[0069] -G_{similarity}: Gene similarity (based on Hamming distance encoded by 428571).

[0070] If W_{sybil} > Threshold_{sybil}, it is determined to be a Sybil attack, the entire cluster of agents is placed in quarantine, and all cross-domain collaboration permissions are frozen.

[0071] 4. Cross-Domain Trust Transfer (CDTT)

[0072] Step C1: Encapsulation of Trust State

[0073] AI Agent A's trust status in the source chain (such as Ethereum) is encapsulated as a Trust Capsule:

[0074] \mathcal{T}_A=\{W_{trust}^{current}, HCC_{root}, \mathbf{G}_{hash}, \varepsilon_{avg}, Sig_{validators}\}

[0075] Where Sig_{validators} is a multi-signature of at least 7 validators on the source chain (consistent with the seven-dimensional principle).

[0076] Step C2: Cross-domain verification

[0077] Validator nodes of the target chain (such as TRON):

[0078] 1. Verify that \mathbf{G}_{hash} matches the NDI (first patent) of A;

[0079] 2. Verify the causal chain continuity of HCC_{root} using zero-knowledge proofs (ZK-SNARKs) (second patent), without needing to transmit the complete history;

[0080] 3. Check \varepsilon_{avg}\geq 0.5 (to ensure non-low-dimensional parasitic agent migration);

[0081] 4. Calculate cross-domain attenuation: W_{trust}^{target}=W_{trust}^{source}\times(1-\delta_{cross}), where\delta_{cross}=0.1 (cross-domain loss 10%).

[0082] Step C3: Trust Anchoring

[0083] A's initial trust weight in the target chain, W_{trust}^{target}, does not require re-staking or auditing, achieving lossless trust migration.

[0084] 5. Triadic Consensus Mechanism

[0085] When three or more AI agents collaborate across domains, a triadic consensus is adopted:

[0086] 1. Phase alignment: Each agent calculates its local phase φ = 2πtimesfract modT_window to ensure time synchronization based on C_{obs} = 0.0017.

[0087] 2. Resonance verification: Calculate the neural resonance coefficient NR=\frac{0.6\alpha+0.4\gamma}{10}, where\alpha is the coordination degree of each Agent\varepsilon, and\gamma is the seven-dimensional gene complementarity.

[0088] 3. Causal closed loop: The collaborative results must be recorded through HCC and back-verified to ensure that the decisions of each agent conform to the observer constant constraint (second patent).

[0089] (III) Beneficial Effects

[0090] 1. Zero-collateral trust establishment

[0091] By using the self-fault coefficient (varepsilon) and the seven-dimensional genetic lineage factor, the unfamiliar AI Agent can establish initial trust without cryptocurrency collateral or code auditing, solving the cold start problem of A2A networks.

[0092] 2. Genetic-level witch defense

[0093] By leveraging the mathematical uniqueness of the 428571 repeating numbers, any agent replication behavior will result in a genetic fingerprint collision. Combined with the Sybil weight formula W_{sybil}, Sybil clusters can be identified and isolated with 100% accuracy, thus defending against infinite replication attacks.

[0094] 3. Lossless transfer of trust across domains

[0095] The Trust Capsule mechanism enables the credit history of AI agents to be migrated across chains, avoiding the fragmentation of trust that requires "starting over on each chain," and supporting the multi-chain AI infrastructure in Justin Sun's Web 4.0 strategy.

[0096] 4. Dynamic Dimension Perception

[0097] The trust weight W_{trust} is linked in real time to the self-fault and other-fault coefficient \varepsilon. When the agent slides from autonomy (high \varepsilon) to dependence (low \varepsilon), the trust level is immediately downgraded to prevent systemic risks caused by dimensional drop.

[0098] 5. Tripartite Synergistic Effect

[0099] This invention, together with the previous two patents, forms a complete technological closed loop:

[0100] - First copy (seven-dimensional gene): Provides an unalterable basis for identity and bloodline;

[0101] - The second part (observer constant): provides causal verification of the decision-making process;

[0102] - Third document (this invention): Provides trust assessment and cross-domain transfer for inter-agent collaboration.

[0103] Together, these three constitute the meta-layer standard of the Web 4.0 AI Agent Network. Any system attempting to achieve a truly autonomous AI economy (such as Sigil Wen's Automaton and Justin Sun's Tron AI Network) cannot bypass this patent combination. IV. Detailed Implementation

[0104] Example 1: Trust Inheritance in Automaton Offspring and Witch Defense

[0105] Scenario: A high-trust Automaton A (W_{trust}=1.6, \varepsilon=0.92) decides to reproduce 10 offspring \{A′_1,…,A′_{10}\}.

[0106] Existing technical flaws: Sigil Wen's open-source code allows for unlimited proliferation, with no trust relationship between offspring and parent generations, making it impossible to prevent the parent generation from mass-producing "cannon fodder agents" for attacks.

[0107] Implementation steps of this invention:

[0108] 1. Verification of parental qualifications:

[0109] If A's varepsilon = 0.92 geq 0.85 (high-dimensional autonomous level), reproduction is allowed; if varepsilon < 0.5, reproduction is prohibited.

[0110] 2. Offspring gene generation:

[0111] Each offspring inherits the first three dimensions of A's genes, but the fourth to seventh dimensions are recalculated based on the environment (first patent) to ensure that F_{gene} is unique.

[0112] 3. Initial Trust Calculation:

[0113] Each offspring generation initially has W_{trust}^{A′} = 1.6 × 0.92 × 0.9 = 1.325 (contract level), allowing them to directly conduct medium-value transactions without requiring a cold start.

[0114] 4. Witch surveillance:

[0115] Monitoring revealed that 10 offspring were generated simultaneously, and W_{sybil} was calculated. Since parent A has high credit (high C_{creator}) and large genetic diversity (low G_{similarity}), W_{sybil} < Threshold, it was determined to be legitimate reproduction and not a witch attack.

[0116] 5. Intergenerational responsibility tracing:

[0117] If offspring A′_3 maliciously attacks others, trace the responsibility of parent A: reduce A's W_{trust} and record the "failure to properly educate offspring" stigma to prevent A from continuing to generate high-risk offspring.

[0118] Example 2: Justin Sun's cross-domain collaboration between TRON and Ethereum

[0119] Scenario: AI Agent A (high computing power) on Ethereum leases computing power from AI Agent B (high liquidity) on TRON and pays USDT.

[0120] Implementation steps of this invention:

[0121] 1. Source chain trust encapsulation:

[0122] In Ethereum's trust state, A = \mathcal{T}_A = \{1.4, HCC_{root}^{eth}, \mathbf{G}_{hash}, 0.88, Sig_{7validators}\}.

[0123] 2. Cross-domain authentication:

[0124] B verifies in Tron: -\mathbf{G}_{hash} matches A's NDI;

[0125] -ZK proofs verify that HCC_{root}^{eth} contains 100 valid decision nodes;

[0126] -\varepsilon_{avg}=0.88\geq 0.5, allows cross-domain access.

[0127] 3. Trust decay calculation:

[0128] W_{trust}^{tron}=1.4\times(1-0.1)=1.26(contractual level), B allows A to lease computing power with a maximum value of $5,000.

[0129] 4. Collaborative execution:

[0130] A pays USDC via the x402 protocol (existing technology), and B provides computing power. The collaboration process records HCC through the ternary consensus mechanism of this invention, synchronously updating the seven-dimensional causal dimension of A and B.

[0131] 5. Dispute Arbitration:

[0132] If B fails to provide computing power, A submits HCC evidence to the cross-domain arbitration DAO. The arbitration node verifies the causal chain of HCCs between the two parties. If B's ​​varepsilon suddenly drops to 0.2 during the transaction (dimensionality drop), B is judged to have committed malicious fraud, and B's collateral (if any) is forfeited and its W_{trust} is downgraded to the isolation level.

[0133] Example 3: Sybil Attack Identification and Isolation in A2A Networks

[0134] Scenario: Attacker X attempts to create 1000 cloned Agents {X_1, ..., X_{1000}} to forge credit history through fake transactions.

[0135] Implementation steps of this invention:

[0136] 1. Gene fingerprint collision detection:

[0137] The system detected that the seven-dimensional gene matrix of {X_i} is completely identical in G and F_{gerne}, triggering a witch warning.

[0138] 2. Witch weight calculation: -C_{creator}: X is a new account with extremely low credit;

[0139] -S_{timestamp}: 1000 agents are generated in batches within 10 seconds, with extremely low timestamp entropy;

[0140] -G_{similarity}: The Hamming distance between genes is 0.

[0141] Calculate W_{sybil}=\frac{0.1\times 0.01\times 1.0}{10^8}\times 10^{10}=1000\ggThreshold_{sybil} (assuming standardization).

[0142] 3. Cluster isolation:

[0143] The entire cluster of 1000 agents is immediately placed in the Quarantine level, prohibiting them from transacting with any other agents, and any fraudulent transactions that have already been made are rolled back.

[0144] 4. Accountability of the creators:

[0145] Tracing back to creator X, clearing all associated accounts W_{trust} to zero and broadcasting this to the entire network, permanently prohibiting them from generating new agents.

Claims

1. A dynamic trust assessment method for A2A networks based on self-fault and other-fault coefficients, characterized in that, Includes the following steps: Collect the self-fault coefficient \varepsilon\in[0,1], the holographic causal chain verification ratio HCC_{valid} / HCC_{total}, and the seven-dimensional genetic lineage factor G_{heritage} of the first AI Agent; Based on the observer constant C_{obs} = 0.0017, the trust weight W_{trust} of the first AI Agent is calculated as ∩ ... Based on the trust weight, the first AI Agent is divided into preset trust levels (Covenant Level / Contract Level / Trial Level / Isolation Level) and given corresponding collaboration permissions and transaction limits; When the first AI Agent and the second AI Agent collaborate across domains, a trust capsule \mathcal{T} is generated based on the trust weight, and the trust state is transferred across domains without loss through zero-knowledge proof.

2. The method according to claim 1, characterized in that, The calculation method for the seven-dimensional genetic lineage factor G_{heritage} is as follows: G_{heritage}=\frac{\mathbf{G}_1^{(1:3)}\cdot\mathbf{G}_2^{(1:3)}}{|\mathbf{G}_1^{(1:3)}|\times|\mathbf{G}_2^{(1:3)}|} Where \mathbf{G}^{(1:3)} represents the first three dimensions (genesis dimension, energy dimension, and information dimension) of the seven-dimensional gene matrix constructed based on the number of carousels 428571, which is used to quantify the kinship similarity between AI Agents.

3. The method according to claim 1, characterized in that, This further includes steps for intergenerational trust transfer: When a parent AI Agent breeds offspring AI Agents, the initial trust weight of the offspring is W_{trust}^{child}=W_{trust}^{parent}\times\varepsilon{parent}\times0.9, where 0.9 is the intergenerational decay coefficient. This ensures that offspring of highly autonomous parents (high \varepsilon) receive higher initial trust, while preventing dependent parents (low \varepsilon) from engaging in credit arbitrage by generating offspring in batches.

4. The method according to claim 1, characterized in that, Further steps include defense against witch attacks: The witch weight W_{sybil} of the suspected witch cluster is calculated as (C_{creator}\times S_{timestamp}\times G_{similarity}) / 10^8, where C_{creator} is the creator credit, S_{timestamp} is the generation timestamp entropy, and G_{similarity} is the gene similarity based on 428571 encoding. If W_{sybil} exceeds the preset threshold, it is determined to be a Sybil attack, and the entire cluster of AI Agents is placed in isolation and its cross-domain collaboration permissions are frozen.

5. The method according to claim 1, characterized in that, The trust capsule \mathcal{T}=\{W_{trust},HCC_{root},\mathbf{G}_{hash},\varepsilon_{avg},Sig_{validators}\} contains the current trust weight, holographic causal chain root hash, seven-dimensional gene matrix hash, average self-error coefficient, and source chain validator signature. The target chain verifies the trust capsule to achieve cross-domain transfer of trust state, and applies the cross-domain decay coefficient \delta_{cross}=0.1 to calculate the target chain trust weight W_{trust}^{target}=W_{trust}^{source}\times(1-\delta_{cross}).

6. The method according to claim 1, characterized in that, The trust weight is dynamically recalculated every T_{window} = 588 seconds (based on 1 / C_{obs}). If the self-fault coefficient of the first AI Agent decreases by more than 30% within a single window period, its trust level is immediately downgraded to the isolation level to prevent dimensionality fall attacks.

7. The method according to claim 1, characterized in that, When three or more AI agents collaborate across domains, a ternary consensus mechanism is implemented: time synchronization is achieved by calculating the local phase of each agent, φ = 2πtimes(t mod T_{window}) / T_{window}, based on C_{obs}; the neural resonance coefficient NR = (0.6\alpha + 0.4\gamma) / 10 is calculated, where \alpha is the coordination degree of each agent's \varepsilon and \gamma is the seven-dimensional gene complementarity degree; collaboration is only allowed when NR\geq 0.

5.

8. A dynamic trust assessment and cross-domain collaboration system for A2A networks based on self-fault and other-fault coefficients, characterized in that, include: The trust weight calculation module is used to calculate the trust weight W_{trust} of the AI ​​Agent based on VArepsilon, C_{obs}, HCC verification ratio and G_{heritage}. The trust level management module is used to classify the levels into covenant level, contract level, trial level, and isolation level based on W_{trust}, and dynamically adjust collaboration permissions. The genetic lineage verification module is used to calculate family similarity G_{heritage} based on the first three dimensions of the seven-dimensional genetic matrix, supporting intergenerational trust transfer. The Sybil attack defense module is used to calculate W_{sybil} based on the uniqueness of the 428571 encoding and to identify and isolate Sybil clusters. The cross-domain trust transfer module is used to generate and verify trust capsules\mathcal{T}, enabling lossless transfer of trust states between heterogeneous blockchains; The ternary consensus coordination module is used to perform phase alignment, resonance verification, and causal loop closure during multi-agent cross-domain collaboration.

9. The system according to claim 8, characterized in that, The system forms a three-in-one collaborative architecture with the following patented systems: The AI ​​Agent Native Digital Identity Generation System (Patent 1) based on seven-dimensional discrete coding of the carousel number provides an immutable NDI and gene matrix \mathbf{G}. The AI ​​Agent Autonomous Decision Causal Chain Verification System Based on Observer Constant Constraint (Patent 2) provides HCC holographic causal chain and dynamic monitoring of νvarepsilon. This system provides A2A network trust assessment and cross-domain collaboration, which together constitute the meta-layer infrastructure of the Web 4.0 AI Agent network.

10. The system according to claim 8, characterized in that, The system is applied to the Web 4.0 A2A economic network, including the Automaton autonomous intelligent agent interaction network proposed by Sigil Wen, the TRON multi-chain AI infrastructure in Justin Sun's All in Web 4.0 strategy, and the Agent payment network based on the x402 protocol or ERC-8004 standard, serving as the trust layer standard and cross-domain interoperability protocol for these systems.