Foresight has published a market analysis arguing that AI is moving from intelligence to agency. In that shift, an agent is no longer limited to answering questions. It can understand goals, call tools, execute tasks, interact with other agents, and produce real-world outcomes.
The article says this change pushes an old human problem to the front of Web3. If an agent can call APIs, operate wallets, manage assets, sign onchain transactions, and even hire other agents to buy services, the question becomes straightforward: on what basis should value and authority be handed to an unfamiliar digital actor? In the author’s view, if the cost of trust between unfamiliar agents cannot be reduced, the agent economy remains a white-paper vision rather than an operational system.
ERC-8004 as a verifiable trust entry point for the agent economy
The article presents ERC-8004, introduced by Ethereum teams and core developers, as a key answer to that problem. It describes the proposal as a highly watched Ethereum Improvement Proposal that has been deployed quickly across multiple major networks and now serves as a trust anchor and a verifiable trust entry point in Ethereum’s agentic economy.
According to the piece, ERC-8004 is built around three core registries:
- Identity Registry: answers the question, “Who is this agent?” It gives an onchain agent a portable and non-forgeable native identity so it can carry a digital credential across platforms.
- Reputation Registry: answers, “How has this agent behaved in the past?” It creates structured behavioral feedback and credit records, aiming to break data silos and turn historical performance into composable onchain reputation signals.
- Validation Registry: answers, “Did it actually do what it said it would do?” The article says this layer can use independent validation nodes, stake-secured re-execution, zkML, or trusted execution environments to provide proof of execution.
In the author’s framing, the main breakthrough is that ERC-8004 combines identity, reputation, and validation into an open framework for verifiable trust. That moves agents away from being treated as probabilistic models alone and toward becoming verifiable agents with minimized trust assumptions onchain.
A wider protocol matrix is taking shape around the agent lifecycle
The article says that as ERC-8004 standardizes identity, reputation, and validation, Ethereum has also started to see faster experimentation around narrower agent problems such as authorization, execution, tasks, and security. It names ERC-8294, ERC-8312, ERC-8370, ERC-8380, and ERC-8414 as examples.
It adds that these proposals do not form an official ERC-8004 extension family in a strict sense. Even so, when viewed across the full lifecycle of an agent participating in economic activity, they collectively map out what the author calls a more complete picture of agent-native infrastructure.
The article says the industry still has to answer eight core infrastructure questions if an agent is to participate fully in onchain economic activity. Ethereum, in the author’s view, is moving quickly through the first half of that roadmap, from identity to execution. But that same map also reveals a hard boundary: even if all of those ERC proposals mature, what Ethereum mainly completes is still the identity, trust, and execution entry point for a single agent. It does not yet solve trusted coordination.
Verifiable trust entry does not equal a trusted coordination network
The piece reduces the distinction to a simple line: trust is the beginning, and coordination is the next frontier.
It argues that verifiable individual agents and standardized calling protocols are not enough to support large-scale, complex collaboration. A large organizational gap remains, and the article breaks that gap into four parts.
- Root of Intent: a protocol for a single agent can show what permissions the agent has and what code it executed. It cannot answer whose will that agent represents, or whether its final objective still matches the original intent of a sovereign human.
- Contribution Attribution: validation does not equal contribution. The article gives a direct example: proving that an agent compiled code is not the same as proving that it created 30% of a project’s value. Once work is shared by a human builder, a digital twin, and several outside agents, contribution measurement and ownership become much harder.
- Multi-Agent Coordination: real production settings are not single-task serial calls. The article describes them as multi-actor, cross-physical-and-digital, asynchronous, parallel, and networked. Its example path is Human / Builder → Digital Twin → Agent A → Agent B → External Network → Human. That kind of graph, it says, goes beyond the expressive range of single-agent protocols built around “task” and “execution.”
- Settlement & EcoFi: the article asks how economic value should be distributed precisely by contribution, and how contribution and reputation can accumulate into long-term digital assets for human actors rather than ending up as scattered agent logs onchain.
On that basis, the piece draws a sharp distinction. ERC-8004 and related proposals are described as trust infrastructure. m&W, by contrast, is presented as an attempt to build coordination infrastructure.
m&W’s position: cooperation and competition with ERC infrastructure
The article says m&W does not see itself as a direct replacement for Ethereum’s ERC stack, nor as its opposite. Instead, it describes the relationship as “coopetition,” spanning several layers.
- Primitives level: ERC-8004 provides basic primitives for identity, validation, and reputation. m&W says it can integrate those primitives as trusted base components inside a higher-order coordination network.
- Protocol level: the ERC matrix defines how an individual agent is identified, authorized, and executed. m&W says it extends that by defining how agent networks divide work, schedule states, and route contribution around human intent.
- Order level: the article says the deeper strategic competition sits here. If billions of agents enter the economy, what institutional order and incentive design should govern human-agent and agent-agent coexistence?
The article sums up the distinction with two questions. ERC-8004 asks, “Can I trust this Agent?” m&W asks, “Can trusted Agents coordinate around human intent?” One is about deciding who to trust. The other is about how trusted actors complete a shared objective.
Three layers of human-machine collaboration: Builder, Digital Twin, Agent Network
At the system level, the piece frames the difference as one between agent-centric infrastructure and human-anchored coordination.
In the m&W architecture described in the article, human-machine collaboration unfolds across three layers:
- Builder, or the sovereign human, is the root of intent. The article defines this role as the source of value, the final definer of goals, and the ultimate bearer of moral judgment and responsibility.
- Digital Twin is the execution identity. It acts as a projected digital proxy of the builder’s will, preferences, and knowledge base, while also serving as the authorization boundary and control hub between the human and a large agent network.
- Agent Network consists of heterogeneous intelligent actors, both onchain and offchain, that carry out specialized tasks within a trust and authorization framework.
The article’s core claim is direct: AI gives machines the ability to act, digital twins give human intent the ability to act, and a trusted coordination network lets that ability scale.
An open architecture with m&W as the coordination layer
The article also says m&W is not meant to become an all-encompassing “superior governing protocol” inside Web3. It instead places m&W at the center of an open coordination layer within a broader agent-native stack.
In that stack, MCP handles standardized tool and context calls for agents. A2A handles message passing and interaction between agents. ERC-8004 and newer ERC proposals handle onchain identity, validation, authorization, and execution. x402 handles instant machine-to-machine micropayments and is described as an orthogonal payment layer decoupled from ERC-8004.
m&W, in this framing, acts as the orchestration center. It arranges capabilities, communication, payments, and trust primitives so they can be aligned with the will of the human builder and converted into a collaborative network with authorization boundaries, contribution measurement, and an EcoFi economic loop.
The flow presented in the article starts from Human / Builder and Intent / Goal, passes through the m&W Coordination Layer, and connects to Identity & Trust, Communication, Capability & Tools, Agent Execution, Payment / x402, Contribution Attestation, Reputation / Ownership, and EcoFi.
The article’s conclusion: the next phase depends on rules, relationships, incentives, responsibility, and value flow
In closing, the author says ERC-8004 does not solve every problem in the agent economy. Its value, in this argument, is that it exposes the deeper problems more clearly.
The piece compares that trajectory to the history of the internet. TCP/IP solved trusted packet transmission, but the broader digital economy emerged from the web standards, e-government protocols, and social networks built on top of it. In the same way, ERC-8004 may establish a verifiable trust entry point for an individual agent, but a full AI agent economy depends on the rules, relationships, incentives, responsibility, and value flow built above that layer.
The article says ERC-8004 has been in development for more than a year and has already been deployed on 16 networks. It adds that the protocol matrix has largely completed a map of agent-native infrastructure, with identity, reputation, validation, authorization, execution, and payment becoming more mature. What remains unresolved, according to the author, is now easier to name: contribution, coordination, responsibility, and value flow.
The piece ends with a concise distinction. The ERC-8004 protocol matrix addresses how to trust an agent. m&W is exploring how trusted digital actors can create together.
The article also includes a disclaimer stating that markets carry risk, investment should be approached with caution, the piece does not constitute investment advice, and readers should judge whether any opinion, view, or conclusion fits their own situation before making decisions.
