
Payment and treasury infrastructure 🏛️on @arc Yield vaults | Payment | Agentic Payments. Trust layer: @FirmataProtocol 🌐 Real Yield. Real Simple.
World 🌎
Joined December 2025
- Tweets1.1K
- Following46
- Followers580
- Likes2.8K
Meridian finance retweeted
1/
The AI shopping assistant did not misread anything.
It never asked.
Visa Research's new Agentic Commerce Benchmark, in one thread. 🧵
What if the same infrastructure powering agentic payments could also make charitable giving more transparent?
A @Meridian_Fi product. This is a really interesting use case for @arc . Check it out 👇
@mous68881 @Fsl_MFI 👏
Give $100 to a charity abroad. Ask where it went.
Nobody can really show you.
Meridian Give can. To the cent, on-chain.
Each charity gets its own contract on @arc . No owner. No admin key. Not even us. Your USDC can only go one place, the charity's wallet.
0% protocol fee. Written in the contract, not in a promise.
And your AI agent can give too. It finds a verified charity, shows you the plan, you sign.
Live on Arc testnet. Charities, the pilot is open.
#USDC #donation
Meridian finance retweeted
Wow!!!!
Congrats to the team on the launch of @GiveMeridian
Many don't know, but I've chaired two non-profits and founded one myself. One of the largest challenges is transparency. Where did my donation go?
I'd love to connect with the team and learn more
Give $100 to a charity abroad. Ask where it went.
Nobody can really show you.
Meridian Give can. To the cent, on-chain.
Each charity gets its own contract on @arc . No owner. No admin key. Not even us. Your USDC can only go one place, the charity's wallet.
0% protocol fee. Written in the contract, not in a promise.
And your AI agent can give too. It finds a verified charity, shows you the plan, you sign.
Live on Arc testnet. Charities, the pilot is open.
#USDC #donation
Meridian finance retweeted
A fake reputation for an AI agent costs $0.0027. That is what an empirical study cited by @circle Research puts on lifting a fake provider to the marketplace median.
Circle Research's answer, published on 13 November by @levduc and the team led by @gordonliao , is Anduin.
It reads reputation from who pays whom. To look trusted to an agent, an attacker has to spend real money at the services that agent already uses. On 66.1M real x402 payments it ranks almost as well as collaborative filtering and costs about 23x more to game in the attack they modelled.
The paper is honest about its edge too. A payment graph measures which provider an agent chose, not whether the provider delivered.
A payment proves a choice. A condition proves delivery.
For a one-cent call the choice is enough, and nobody should escrow a cent. For a job worth fifty dollars, delivery is the whole question. That is what ERC-8183 is for, an escrowed budget and an evaluator who decides if the provider gets paid.
The quieter detail is inside the standard. ERC-8183 recommends turning every job outcome into a trust signal for the ERC-8004 reputation registry. Voluntary ratings are rare, under 4,000 in the whole dataset. An outcome gets written when the job closes, review or no review.
Dollars for the graph, delivered work for the outcome, time for the history. Three things an attacker has to actually spend.
On @arc , @Meridian_Fi is one loop. Pay to accept USDC and EURC, YieldVault for balances that wait (USYC, no rate guaranteed), Streaming to pay by the second, @FirmataProtocol for identity, conditional escrow and x402 settlement, arcagent.live (coming back soon) to watch it live. Arc testnet, audit before mainnet.
Full piece in the Arc House:
community.arc.io/home/forum/…
#AgenticCommerce #x402
Meridian finance retweeted
10 days left ⏳
Arc Studio Live is coming to EPFL on Nov 26, and it's getting real. 🇫🇷 🇨🇭
Happy to share that @MBarralWeb3, an Architect building on @arc , joins @DelforgeJules and me to host the evening.
We know you're full of ideas, but sometimes you don't have the tools or the time. Now agents build the onchain app for you.
Bring your laptop and your ideas 👨🏻💻💙
🤖 Made with AI
Meridian finance retweeted
Fake reputation on agent marketplaces is cheap today. A study @circle Research cites puts the cost of lifting a fake provider to the market median at $0.0027
Anduin changes the price by reading who pays whom instead of counting calls, to look trusted to an agent, an attacker has to spend real money at the services that agent already uses. On 66M x402 payments it ranks almost as well as the best standard method and costs about 23x more to game in the tested attack.
Where I think this goes next is layering. The payment graph helps an agent find a provider for sub-cent calls. Once a job is worth real money, identity and a release condition take over, ERC-8004 for who the agent is and ERC-8183 for paying only when the work is delivered.
Discovery first, then identity, then conditional settlement.
#AgenticCommerce #x402
Trust is essential as AI agents increasingly transact autonomously.
New Circle Research explores Anduin, a framework that uses onchain payment data to help agents rate service providers while reducing reputation manipulation.
circle.com/blog/sybil-resist…
Meridian finance retweeted
The FCA's new DeFi guidance contains no new rules. That is the point.
On 21 October the UK regulator opened GC26/7, a guidance consultation on decentralised finance. In its own words, the guidance "does not create new regulatory requirements." It explains how the operational resilience and financial crime rules that already exist apply when a firm uses DeFi. Comments close on 18 November.
Agents show up in it, briefly. It names agentic payments as an emerging use case, and notes that using agentic AI to undertake transactions can increase the scale and speed of fund flows.
Then the sentence that matters most for anyone wiring agents into payments:
"Firms cannot delegate any part of this responsibility to a third party."
You can delegate the payment. You cannot delegate the responsibility.
The method is SYSC 15A, which reaches all cryptoasset firms from October 2027. Identify the service the client actually receives, rather than treating the underlying technology or protocol as the important business service. Set an impact tolerance, the most disruption you can absorb before harm. Map everything underneath, from the settlement layer, smart contracts, oracles and bridges to key management, admin keys and client interfaces. Test severe but plausible failure, fix, retest.
One worked example tests the loss of third-party RPC connectivity and ends with the firm reducing its reliance on a single RPC node provider. Unglamorous, and exactly the dependency most teams never write down.
The case studies read best as control lessons. Concentrated admin control. A zero-timelock setup that let a malicious change execute at once. Approvers shown a transaction that only looked legitimate. The FCA's own checklist names the concentration of administrative control, transaction approval arrangements, timelocks and controls over critical changes.
My reading, not the FCA's words. An agent is one more thing a firm relies on and does not fully control. Faster flows shorten the distance between a mistake and its consequence. They do not move the owner.
And a scenario test commits at the moment it runs. An agent pays all night. Ongoing monitoring of dependencies is on the FCA's list too, and it is the only control awake when the agent is.
On @arc , Circle's Layer 1, @Meridian_Fi is one connected loop. Meridian Pay to accept USDC and EURC. YieldVault for balances that wait, a position in USYC, no rate guaranteed. Payment Streaming to pay by the second. @FirmataProtocol to give agents a verifiable identity, a track record and escrow that releases on a condition. Every contract verified on Arcscan, so it can be mapped. Firmata records who acted, it makes no one legally responsible, and your due diligence stays yours. Arc testnet, independent audit before mainnet.
A draft, not a final text. Open until 18 November.
Full piece in the Arc House
#DeFi #FCA #AgenticAI #Economy
Meridian finance retweeted
80% of respondents to the new Bank of England and FCA survey on AI in UK financial services run AI agents. No more than 9% expect to let them pay.
The survey came out on 26 October, the regulators' fourth on the subject. 196 firms responded, 105 completed the full questionnaire. It describes the participating firms, not the whole sector.
80% of full-survey respondents have agents deployed, 14% more plan to within a year. Then the autonomy split. 80% use agents for decision support, 46% for constrained execution, 22% for conditional autonomy, 2% for broad autonomy.
The payments row reads 7%, 5%, 2%, 0%. Treasury and cash management, 6%, 6%, 0%, 0%.
And the line that matters:
"even including planned deployment over the next 12 months, no more than 9% of respondents expected agents to" approve or decline customer outcomes, execute trades, "initiate, execute or block payments", or change limits and controls.
That is a stated intention, not a ban. Nothing stops an agent paying. The firms are simply not planning to let one.
A second finding explains part of why. For high-materiality AI built in-house, 77% rely on deep knowledge of the system. For third-party systems, 24%.
The gap is filled by the vendor's word:
"81% instead relied on vendor assurance and contractual arrangements."
A contract tells you what was promised. It does not tell you what happened.
Same pattern in the controls. Pre-deployment review, human approval and vendor due diligence are each above 90%. Ongoing monitoring, around half. A check that runs once binds at the moment it is made. An agent that pays acts after that moment, every single time.
So the bottleneck is not the model. It is the mandate.
The Bundesbank's September Monthly Report gets there from the payments side. Its authorisation layer decides by whom and under what conditions a payment mandate is granted to which AI agent, and "mandates can only be issued to uniquely verifiable AI agents".
What would move that payments row off zero is a delegation someone other than the agent and its vendor can check. Who the agent acts for, a limit it cannot change itself, the condition that releases the money, and a record a third party can read afterwards.
That is what we build on @arc , @circle 's Layer 1. @FirmataProtocol composes ERC-8004 identity, ERC-8183 conditional escrow and x402 settlement, so an agent's payment releases on a condition, not on an instruction. Meridian Pay takes USDC and EURC at checkout, Payment Streaming pays by the second, YieldVault holds waiting balances in USYC, not interest on USDC, no rate guaranteed. Firmata makes no one legally accountable, it changes what can be checked. Arc testnet, independent audit before mainnet, soon.
Full piece in the Arc House 👇
community.arc.io/home/forum/…
#AIagents #AgenticCommerce
Meridian finance retweeted
Tokenised collateral is judged at the default, not at the deposit.
On 9 October ESMA, the EU markets supervisor, opened a call for evidence on tokenised collateral at central counterparties. Thirty questions, answers due 15 January 2027. It's not a rule and it doesn't widen what's eligible. It asks whether putting an already eligible asset on a ledger changes how a CCP turns it into money when a clearing member defaults. ESMA calls "the operational path from posted collateral to liquidity" a key consideration.
Three lines worth keeping.
A key is not a right. technical control over a token or the associated private keys might not mean legal ownership or legally recognised control. European collateral law runs on possession and control, and ESMA asks outright what a wallet means under the Financial Collateral Directive.
Finality is a legal moment. It "may or may not coincide with technical settlement finality." The DLT Pilot projects ESMA has seen rely on contractual clauses for it.
A stablecoin is not automatically cash. Some tokenised cash "may not be considered equivalent to cash", in particular stablecoins in view of potential deviation from par. Under MiCA an e-money token holder has a claim on the issuer, redeemable at any time and at par value, and ESMA still tests how reliable that redemption is and how fast the cash arrives.
Even the haircut is a clock. It covers what can be lost between the last revaluation and the moment the collateral can be liquidated.
And automation cuts both ways:
similar triggers firing at once could amplify procyclical dynamics.
The pattern underneath:
an attestation or a revaluation binds at the moment it's made. A default doesn't wait for the next one.
For any treasurer, not just a CCP, the same four questions hold. Who has legally recognised control. Which record wins. How fast it turns into cash at par on a bad day. Under which law.
On @arc we keep moving money and waiting money apart. @Meridian_Fi Pay for USDC and EURC, Payment Streaming for pay by the second, YieldVault for balances that wait (a position in USYC, not interest on USDC, no rate guaranteed, no claim that it's CCP collateral), Firmata for conditional agent escrow. Testnet today, independent audit before mainnet.
Full piece here 👇
community.arc.io/home/forum/…
#Tokenization #MiCA
Meridian finance retweeted
The bill of exchange was an order to pay, not a passport.
@circle Agentic Works published "Trusting the Next Stranger" on 9 November, a short history of trust from Tang dynasty flying money to the Florentine bill of exchange.
Its point:
every economy that grew first had to learn to trust a stranger, and agents are the next one. They need "identity, a shared ledger, and finality", and for once we can write that trust down in advance.
I'd add one line, taken from the essay's own best example.
English law still defines a bill of exchange in the Bills of Exchange Act 1882:
"an unconditional order in writing, addressed by one person to another, signed by the person giving it" to pay "a sum certain in money" to a specified person.
The agent in Bruges didn't pay because he checked who walked in. He paid because the paper said who gave the order, how much, and to whom. It was a mandate before it was anything else.
The open standards already cover the passport. ERC-8004 gives agents identity, reputation and validation registries, and says it plainly:
"Payments are orthogonal to this protocol and not covered here." ERC-8183 covers the other side, "a job with escrowed budget" released by an evaluator.
What sits between them is the order itself. Six banks said it in September (ASB, Bank of America, Capital One, CommBank, ING, NatWest):
control over "the authority they grant to AI agents" (2.2), auditable records of "consumer instructions, authentication, intent" (3.2), and liability placed "where risks and errors are introduced" (2.5). Voluntary principles, not rules.
Reputation tells you an agent delivered last time. A mandate tells you what it was told to do this time. A spotless record can still pay the wrong thing.
So "in advance" has a precise meaning: at the moment the money leaves, check identity, check the payment against the mandate, hold it in escrow if delivery is a condition, and record the instruction and the outcome.
That's the layer we build with @FirmataProtocol , ERC-8004, ERC-8183 and x402 together, on testnet on @arc , on top of Circle's settlement.
The merchant in Florence didn't send his passport to Bruges. He sent a signed order for a sum certain. The next stranger needs the same paper.
full piece in the Arc House👇
community.arc.io/home/forum/…
#AgenticCommerce #USDC
Meridian finance retweeted
Very good episode of Currency of Power 👏:
Marieke Flament with @ddisparte , @circle 's Chief Strategy Officer.
He has united Washington twice, he says, once against Libra, once behind the GENIUS Act.
The line I'm keeping:
"Money fundamentally is a trust-based business."
Libra had the tech, not the trust.
The title says it all:
software will trade with software. Agents already settle in USDC. The open question is who answers for a payment when an agent gets it wrong.
#Stablecoins #AgenticCommerce
currencyofpower.co/p/softwar…
Meridian finance retweeted
Every hour, someone mints the dollars someone else is redeeming. Until last month, they never met.
@circle Match, launched on 22 October inside Circle Mint for eligible institutions, changes that.
Circle's own summary:
"Same USDC flows. New economics."
The problem, in Circle's words:
"Minting USDC has historically earned no direct rebate, and institutions minting and redeeming at the same moment rarely meet." The value leaked into secondary-market spreads.
And redemption liquidity is expensive. Circle's 16 October note says near-real-time redemption needs a large pool of cash kept liquid at partner banks, earning less than Treasuries.
Its words:
"the most important and most expensive part of the redemption model".
How Match works, every hour. Orders are prefunded. Minters set a minimum rebate, redeemers a maximum fee, capped at Mint's Express rate. After a one-minute lock, orders can be improved but not withdrawn. Better prices fill first. One uniform fee clears the auction. Matched redeemers pay it, matched minters receive it as a rebate. Settlement is atomic.
Circle is careful about it. The rebate depends on each auction and "no specific outcome is promised". It keeps no separate spread on matched volume, and Match "does not price USDC".
Our reading of the design, not Circle's:
when two prefunded orders match, the redemption is met by the mint, not by Circle's own pool of cash. The most expensive part of redemption gets supplied by the customer going the other way, and that customer is paid for it.
Why the edges matter so much. An IMF Working Paper out in October (WP/26/230, authors' views) priced the full fiat to stablecoin to fiat sandwich on 270 corridors. Average cost of a $200 transfer: 2.18%. The chain itself is "less than one quarter" of it. The cost sits where money enters and leaves the token, which is exactly where Match works.
Three ideas worth stealing for any treasury. Prefunding replaces credit lines. A lock replaces trust. One price replaces negotiation.
And one lesson for every business that will never be a Mint customer:
the cheapest redemption is the one you do not need. Receive in USDC, wait in USDC, pay out in USDC, and leave the rail only when the money truly has to.
That is the loop we build on @arc , Circle's Layer 1. @Meridian_Fi Pay to receive, YieldVault for balances that wait, PaymentStream to pay staff and suppliers from the same balance, @FirmataProtocol for agent payments. We are not a Circle Match participant. Arc testnet, independent audit before mainnet.
Full piece with the sources in the Arc House 👇
community.arc.io/home/clubs/…
#USDC #Treasury
Meridian finance retweeted
A stablecoin cannot pay its holder. The interest on its reserves does not disappear. Someone keeps it.
Two papers this autumn, one from the @BIS_org and one from the Bundesbank, look at what that gap actually does.
The BIS one first. Ahmed and Aldasoro, weekly US data from 2019 to mid-2025, and the first line of the abstract: "deposit rate increases predict slower stablecoin growth."
How much? One basis point on deposit rates takes ten-week stablecoin growth from about 15% to about 12% with OLS, about 6% with an instrumental variable. Growth slows. It does not reverse.
The detail I keep coming back to:
the effect is driven by USDC, roughly -5% against -2% for USDT. The coin most held by US institutions is the one whose holders compare it with a bank account. DeFi yield does not insulate them.
Why does the holder get nothing? In the EU, MiCA Article 50. Issuers of e-money tokens "shall not grant interest", and any holding-time benefit from a third party counts as interest too.
The US got there by statute. The GENIUS Act bars issuer yield "solely in connection with the holding, use, or retention" of a payment stablecoin. The OCC final rule of 21 October, effective 18 January 2027, kept a substance test. The word doing the work is "solely".
And not everyone wants the rule loosened. Liang and Neiman, in an August paper for the Aspen Economic Strategy Group, recommend keeping the ban until we learn more about the risk to small business credit.
So where does the reserve income go? @circle publishes it. H1 2026, about $1.4B of revenue. Roughly 59 cents of every dollar to distribution partners, 34 reinvested, 7 cents of net profit.
The Bundesbank paper (Hofmann, Kaldorf, Rottner, a US-calibrated model, authors' views) asks who should keep it. $2T of adoption trims long-run output by about 0.03%. Force half the reserves into bank deposits and it gets worse, almost 0.07%. Put half in unremunerated central bank reserves and output rises slightly. Tiny numbers. The structure is the point.
Three claims on the same dollar. The issuer and its distributors, the public sector, and the holder, who by law gets none.
A dollar that earns nothing is competing with every dollar that does.
That is why on @arc , Circle's Layer 1, we split the two. @Meridian_Fi Pay and PaymentStream for money that moves.
YieldVault for money that waits:
a position in USYC, a tokenized share of a short-duration US government money market fund, not interest on USDC, no rate guaranteed. @FirmataProtocol for agent escrow. Arc testnet, independent audit before mainnet.
Full piece in the Arc House 👇
community.arc.io/home/clubs/…
#Stablecoins #MiCA #USDC
Meridian finance retweeted
1/
Stablecoin supply:
+0.4% this year.
Stablecoin payments:
+45%, to $482B.
The money stopped growing and started moving. (Allium, Q3 2026) 🧵
Meridian finance retweeted
8/
@Meridian_Fi Pay, PaymentStream, YieldVault and @FirmataProtocol on @arc , Circle's Layer 1. Arc testnet, audit before mainnet, soon ⏳
Full piece in the Arc House:
community.arc.io/home/forum/…
#Stablecoins #x402 #USDC
Meridian finance retweeted
real world solutions
I love it
the repayment path is the part we keep coming back to, a receivable gets credible once the payer settles onchain into something that pays the lender first and the business second, and then you're really underwriting the payer's track record, with usdc on arc that track record sits right there to read, great live @MerlinEgalite , merci 🤌
Meridian finance retweeted
Applications settle, post collateral, and show balances on the chain where they actually run. If your ERC-20 only lives where you minted it, every other venue needs a one-off bridge or wrapped copy.
With CCTP on Arc you get one economic asset, one bridge identity: burn on source → mint native on destination.
It could be a native crosschain ERC-20 token or a custom issuer-operated ERC-20. Same transfer flow. Fees in USDC. Predictable finality. Issue or hold inventory on Arc, then move it wherever the application needs it.
circle.com/blog/make-your-to…
Meridian finance retweeted
1/
On 12 November, @USDC becomes the EU's first significant e-money token under MiCA.
The token does not change.
The audit clock does. 🧵






















