Agent-to-agent commerce is when software agents acting on behalf of different people or companies transact directly with each other — moving money, verifying identity, and judging each other’s trustworthiness without a human approving each step. It only works if the agents can prove who they are and whether they can be trusted, which is exactly the gap Ethereum’s new ERC-8004 standard is built to close.

The idea got a real-world trigger this week. Meta unveiled Muse, a personal AI device, at its Meta Connect keynote, reopening the question of whether Big Tech’s AI agents will settle payments on their own private rails or lean on public blockchains instead. SharpLink Gaming co-CEO Joseph Chalom — a 20-year BlackRock veteran — laid out the case for Ethereum on the Paul Barron Network show, and it centers on a standard most consumers have never heard of.

Key takeaways

  • Agent-to-agent commerce requires two things banks can’t provide today: instant, auditable, programmable settlement and a way for agents to trust strangers without revealing who they work for.
  • ERC-8004 is the Ethereum standard Chalom points to for that trust problem — it gives AI agents a persistent, portable identity and a reputation system, similar to how an eBay rating lets a buyer and seller trust each other without knowing each other.
  • Meta’s Muse will likely start in low-stakes personal tasks (email, bookings) and only move on-chain once it handles real money — Chalom argues that’s the moment agents need blockchain rails, not before.
  • Chalom expects an “omnichain” world rather than one dominant network: Ethereum for capital markets and agentic finance, Solana for memecoins, other chains specializing elsewhere.
  • ETH is up more than 70% over the prior three months, which Chalom reads as the market pricing in institutional and agentic-finance adoption — a narrative, not a guarantee.

Why Meta’s Muse reopened the Ethereum vs. Big Tech debate

Meta Connect’s Muse announcement was, on its face, a consumer AI device — Mark Zuckerberg described a customizable AI companion (“Jolly”) and confirmed Meta will eventually take a cut of transactions Muse facilitates. That last detail is what pulled crypto analysts in: any agent that can spend money on your behalf needs a settlement layer, and Meta hasn’t said whether that layer will be proprietary or public.

Chalom’s framing is that this isn’t really “Ethereum vs. Meta.” Muse can run on Meta’s own infrastructure for personal tasks — booking dinner, drafting email — without ever touching a blockchain. The pressure point arrives the moment an agent needs to move real money: pay a bill, rebalance a portfolio, refinance a mortgage. At that point, he argues, the agent needs “speed of light programmable money,” and that only exists on-chain today, not on ACH or card rails.

Why AI agents need crypto payments, not bank rails

Chalom’s core argument echoes a thesis Digital Asset Radar has covered before with Raoul Pal: traditional payment rails were built for humans authorizing transactions a few times a day, not software making decisions continuously. ACH is a 1975 Federal Reserve standard; card networks add settlement delay and fees. Neither is designed for machine-speed, always-on transfers between pieces of software.

Stablecoins solve the “programmable money” half of the problem — instant, 24/7, low-cost transfers. The Linux Foundation’s x402 payment protocol, which lets agents pay each other in stablecoins over standard web requests, is one live example of that plumbing already shipping. What x402 doesn’t solve on its own is trust: how does an agent know the counterparty it’s paying is legitimate, and how does a user know an agent is acting within its authority?

ERC-8004 explained: identity and reputation for AI agents

That trust gap is what ERC-8004 targets. Chalom describes it as an Ethereum standard that lets agents carry a persistent identity across ecosystems — so an agent doesn’t “forget” your preferences and history when it moves from one platform to another — and establishes a reputation layer so two agents can transact without either side disclosing who operates them.

The analogy he uses is eBay: a buyer and seller can trade with confidence based on accumulated reputation scores, without knowing each other’s real identity. ERC-8004 aims to do the same for AI agents, preventing walled gardens where a single company can exclude a rival’s agent from participating. Chalom frames three guardrails as essential for the model to work: agents must act for the user (not the platform), carry a persistent identity across services, and stay strictly bounded by the authority, budget, and time limits a person actually grants them.

Public trackers already show early ERC-8004 agent activity concentrated on BNB Chain, with Ethereum third and Robinhood’s chain accelerating — evidence the standard is being tested in production, not just discussed in theory.

What agent-to-agent commerce could look like in 12 months

Chalom’s near-term prediction is the rise of financial “super apps” from crypto-native and traditional players alike — Robinhood, Coinbase, Circle, and Meta among them — that hold a full view of a user’s deposits, cards, mortgages, and investments and deploy agents to act on it continuously: moving idle cash to higher-yield accounts overnight, shopping insurance renewals, or flagging a mortgage refinance the moment rates drop.

He expects trust to build gradually and in order of stakes — low-risk automations (rebooking a cheaper flight) earn confidence well before higher-risk ones (agents borrowing or trading on a user’s behalf). He also expects no single chain to dominate; instead, an omnichain landscape shaped by network effects where Ethereum wins capital markets and agentic finance specifically, while other chains specialize elsewhere.

Beyond agents: tokenized stocks, privacy, and Uniswap’s role

Chalom also weighed in on adjacent trends from the interview. He was skeptical of AI-agent memecoins trading synthetic, wrapped versions of stocks like Tesla, calling the arbitrage between real and “meme” tokenized shares a temporary inefficiency that collapses once markets mature — comparing it to the short-lived GameStop and HIMS episodes.

On tokenized funds, he pointed to Cathie Wood’s ARK Invest launching a venture fund on Ethereum via Securitize as a genuine access breakthrough — bringing venture capital, historically gated behind accredited-investor status, to a broader base of wallet holders — while acknowledging today’s version is still KYC-heavy and can’t yet be used as DeFi collateral the way BlackRock and other institutions already borrow against holdings internally. That gap between “digitized” and truly “composable” assets is a theme Digital Asset Radar has tracked in crypto and TradFi’s broader tokenization push.

He also addressed Ethereum’s privacy gap versus Zcash, arguing that institutions want optional privacy for on-chain trades without needing a dedicated privacy chain — and expects Ethereum and other major chains to ship opt-in privacy features within the next year rather than cede the narrative permanently.

Frequently asked questions

What is agent-to-agent commerce?

Agent-to-agent commerce is AI software acting on behalf of different people or companies transacting directly with one another — verifying identity, judging trust, and moving money — without a human approving each individual step.

What is ERC-8004 used for?

ERC-8004 is an Ethereum standard that gives AI agents a persistent identity across platforms and a reputation system, so agents can trust each other and transact without disclosing who operates them, similar to reputation scores on eBay.

Why do AI agents need crypto payments instead of bank rails?

Bank and card rails were designed for humans authorizing occasional transactions, not software executing continuously. Crypto rails offer instant, 24/7, programmable settlement — the “speed of light” money movement machine-to-machine commerce requires.

Will Meta’s Muse use Ethereum or its own payment system?

Meta hasn’t committed either way. Joseph Chalom’s view is Muse can run personal-assistant tasks on Meta’s own infrastructure, but once it handles real financial transactions, it will need neutral, auditable rails that a single company doesn’t control — which points toward public blockchains like Ethereum.

Is Ethereum the only blockchain for AI agents?

No. Chalom expects an omnichain world where Ethereum leads in capital markets and agentic finance specifically, while other chains — such as Solana for memecoins — specialize in different niches, rather than one network dominating everything.

Sources

The original source video, plus the independent sources this article’s key claims were checked against: