@moyed

FLAVOR by moyedhttps://paragraph.com/@moyed Articles about crypto projects that I'm interested in.Sun, 19 Jul 2026 02:16:31 GMThttps://validator.w3.org/feed/docs/rss2.htmlhttps://github.com/jpmonette/feedenFLAVOR by moyedhttps://storage.googleapis.com/papyrus\_images/a1832bbf159adcc99c9987adcb58fb0bhttps://paragraph.com/@moyed

All rights reserved<![CDATA[Era of Composable Finance]]>https://paragraph.com/@moyed/era-of-composable-finance RUYjTKwFBHfhmXkGElyeMon, 25 May 2026 13:57:33 GMTWhat does next-generation finance look like?

This article is an attempt at my own answer to that question. There may not be a single clean definition, but the goal is to put some structure on the new shape of financial services that many people have begun to sense without quite being able to name it.

To get there I looked at as broad a set of companies as time allowed. The list includes Lightspark, Erebor, Catena, Coinbase, Meow, and others. There are many more I didn't get to, which is a real limitation.

The outline of this piece:

The Era of Composable Finance

The reason I needed a new category is that the companies I studied don't fit existing ones cleanly. Some look like banks but aren't, some look like fintechs but aren't, and some get called neobanks but are really just banks. So I'm using a new term: Composable Finance.

Composable Finance is an architecture in which Substrate, Regulatory Wrapper, and User Surface are independently selected and recombined into financial products that work without reducing back to legacy infrastructure.

Three terms do the work in that definition:

(Substrate is often just called the 'rail'. I use both terms interchangeably here.)

The phrase that does the real work in the definition is "doesn't reduce back to legacy infrastructure." Composable architectures have existed in theory since the BaaS era. But every prior combination eventually reduced back to a single legacy rail (Visa, ACH) at some stage, usually settlement. A permutation with the same result isn't really a combination. It's just a relabeling within a closed set.

When I say "the era of Composable Finance has arrived," I don't mean simply that composition has become possible. I mean that composition has started to actually matter. Permutations across the layers are no longer variations of the same legacy stack. They're starting to produce structurally different products.

Three-axis breakdown

To see how composition actually works, separate the unit of analysis first. The unit is the function: custody, payment, yield, credit, FX, treasury, advisory, and so on. Each company picks one or more of these and offers them to customers. Each function is delivered through some combination of the three axes defined above (Substrate, Regulatory Wrapper, User Surface), and which functions get delivered through which combination is what defines the company.

Function = Regulatory Wrapper × Substrate × Surface

A few examples:

Historical flow

This three-axis framework has always existed in principle.

What changed in the Composable Finance era is that permutations across the axes are no longer variations of the same legacy stack. They now produce structurally different products. To see why, walk through the historical flow where each era unbundled one layer of the stack while leaving the rails underneath alone. I'm borrowing this framing (yet again) from Multicoin Capital's Specialized Stablecoin Fintechs.

Pre-Fintech

Until around 2000 the only model was the vertical bank that owned Substrate, Regulatory Wrapper, and surface all at once. The bank had its own substrate (its own ledger, its own Visa relationship, its own ATM network), its own wrapper (its own charter), and its own surface (its own branches and apps). The three layers were welded together. You couldn't have a JPMorgan checking account without going through JPMorgan's rails, JPMorgan's charter, and JPMorgan's app.

Fintech 1.0 (2000-2010)

Fintech 1.0, represented by PayPal, was the era of digital distribution. PayPal made payments clickable on a webpage in 2002, but every click still settled through ACH and card networks built decades earlier. PayPal unbundled distribution (the surface). The rails stayed the same.

Fintech 2.0 (2010-2020)

Fintech 2.0, represented by Chime, was the era of neobanks. Chime offered a free banking app and unbundled the customer relationship, but the deposits still sat at Bancorp Bank, the cards were still issued by Stride Bank, and the rails underneath were the same Visa and ACH that Chase used. Chime unbundled the customer relationship. The rails stayed the same.

Fintech 3.0 (2020-2024)

Fintech 3.0, represented by Marqeta, was the era of embedded finance and BaaS. Marqeta sold card issuance as an API and powered dozens of branded cards including DoorDash, Cash App, and Klarna. If the 2.0 players became one brand on top of legacy rails, the 3.0 players became an API between the legacy rails and many brands. But every card still ran on Visa or Mastercard, still got issued by Sutton Bank or Pathward. Marqeta unbundled the bank-integration contract, but the rails stayed the same.

Fintech 4.0

Each era unbundled one layer of the stack. PayPal unbundled distribution, Chime unbundled the customer relationship, Marqeta unbundled the bank-integration contract. None of them unbundled the rail. Every fintech from 1.0 through 3.0 ultimately reduced back to "still running on top of Chase or Wells or SWIFT."

Composable Finance is the first era in which the rail itself became composable.

The core change is that new Substrates have become sufficient. Stablecoins (USDC on Base, USDT on Solana, DKUSD on the Dakota platform) are dollar substitutes that don't reduce back to sponsor-bank deposits, and DeFi protocols like Aave and Morpho deliver bank-level yields without a bank. For the first time, the Substrate axis has real choices that don't reduce back to legacy.

The test that separates "3.0 with stablecoins bolted on" from 4.0 is one question:

Does the money flow the user sees ultimately reduce to "still running on top of Chase or Wells or SWIFT"? If yes, it's 3.0 with optionality added. If not, it's 4.0.

This Substrate change means any company can now pick a new combination, and that choice changes what the product can do:

Of course, to legally handle a new Substrate you need a matching Regulatory Wrapper. The emergence of categories like OCC trust charters and the PACE Act is part of the same flow. Which Regulatory Wrappers exist today, and which a builder should pick, is the topic of the next section.

A quick tour of the Regulatory Wrapper

Of the three axes, Regulatory Wrapper might be the most consequential. The companies I studied for this piece fall into roughly six Regulatory Wrappers:

For reference, the OCC national charter status as of May 2026:

There is endless detail under each wrapper, but here is a quick comparison on the dimensions builders actually have to choose between:

One thing worth calling out is the PACE Act, proposed in April 2026. What makes the PACE Act (Payments Access and Consumer Efficiency Act) distinctive is that it would guarantee direct access to Fed payment systems (Fedwire / FedNow / ACH) by law. Under every other Regulatory Wrapper, Fed access is gated by the Fed's activity-profile review regardless of charter type, and new activities like stablecoin settlement or agent banking tend to land in Tier 3 and get effectively declined. The PACE Act is the only Regulatory Wrapper that bypasses the activity gate. The catch is that to qualify as a Registered Eligible Provider you have to already hold 40+ MTLs plus a state depository or credit-union charter as prerequisites.

The three clusters that are forming

Looking at the companies covered here, they end up clustering into three groups. Two questions are enough to decide where any given company belongs.

The first and biggest split:

If the latter, the second question:

By the car analogy, Pick & Shovel is the engine maker and the other two are the finished-car makers. Inside that, Bank Builder is the carmaker that builds and installs its own engine, and Vertical Brand is the carmaker that borrows the engine and brings only the body and the brand.

Pick & Shovel

These companies build infrastructure that other fintechs, platforms, and enterprises use to launch financial products. A large share of them are orchestrators that bundle rails owned by other companies into a single API + ledger + compliance program. In a sense, BaaS 2.0.

A few representative orchestrator examples:

Modern Treasury

Originally a software layer on top of corporate bank accounts that exposed ACH/wire/RTP as an API. After acquiring Beam (a stablecoin solutions company) in October 2025 it pivoted to an integrated PSP that absorbs the bank relationships themselves. Now companies like Robinhood, Gusto, Procore, and Navan can handle everything from ACH through stablecoins through a single API. It has its own ledger, which makes sub-accounts and real-time balance aggregation possible.

Dakota

A global banking service that other companies can embed inside their own products. A single API covers user onboarding, multi-currency accounts, sub-client structures, and multi-rail payouts. The recently launched pay-as-you-go model is particularly interesting. This kind of infrastructure is usually enterprise-sales territory, but Dakota made it possible to open a sandbox immediately, sign up in five minutes, and start using it without contracts or sales calls. Pricing is a simple usage-based 25bps + $0.50 ACH + $1 KYC + $5 KYB with no minimum commitment. A five-person team can actually stand up its own fintech in a few days.

Lightspark Grid

Payments infrastructure that ecommerce sites, gig networks, and creator platforms embed into their own platforms. A single API handles KYC/KYB, multi-currency accounts, cross-border transfers, embedded wallets, and AI-agent delegation workflows. The point is to let those platforms directly capture the yield, interchange, and FX margin that used to flow to Stripe, Cross River, and FX desks.

Some companies go beyond orchestration and create Substrate themselves. Stripe, through Bridge, enables other fintechs to issue their own stablecoins and run on/off ramps without their own Regulatory Wrapper, and through Tempo built a settlement rail of its own. Coinbase similarly supports other services' settlement through Base.

Banks like Cross River, Lead Bank, and Column also belong in this category. On the surface they look like Bank Builders because they hold charters and take deposits directly. The defining fact is that their dominant customers aren't ordinary users or businesses but other fintechs building financial services on top. Mercury borrowed Cross River before getting its own charter, Flex borrows Lead, and Slash borrows Column. Two companies can hold the same kind of charter, and the cluster split comes from whether they sell to end users directly or as Substrate that other companies embed into their products.

Bank Builder

This cluster holds customer money directly under its own charter, and the Regulatory Wrapper is the biggest differentiator. Representative companies:

Mercury

A business banking app used by 300K+ startups and small businesses. Mercury started on Choice/Column BaaS in 2017, applied for an OCC charter in December 2025, and received conditional approval on April 27, 2026. With its own charter it captures the full NIM on $20B+ in deposits that it used to share with sponsor banks. It recently shipped a read-only MCP server (burn analysis, board reporting) and a banking CLI.

Augustus

Augustus is the rebrand of Ivy, a European open-banking API company founded in Berlin in 2021. Ivy is already integrated with 5,000+ banks across EMEA, SEA, US, and LatAm, processes billions of euros annually, and counts Kraken as a flagship customer. After filing for an OCC charter in December 2025 and receiving conditional approval on May 11, 2026, it rebranded to Augustus Bank, N.A. (Dallas). The positioning is "a clearing bank for the AI era": a bet on reinventing legacy clearing banks like BNY Mellon and State Street as 24/7, multi-currency, stablecoin-native, and API-first.

The technically interesting part is that fiat balances and stablecoin balances live as peer primitives on the same ledger. A customer can receive EUR via SEPA Instant, send USDC on-chain, and flip between the two balances via internal ledger entries, all from one API call. The target customers aren't retail but crypto exchanges, stablecoin issuers, global payroll companies, and payment processors: businesses that are structurally 24/7 and multi-currency.

Erebor

The only bank in the 2026 charter wave that has received a full charter and is operating. The targets are segments that no existing bank handles well: AI infrastructure / hardware companies, defense-tech startups, robotics, crypto-native companies, and UHNW individuals.

One stated differentiator is the intent to underwrite new kinds of collateral. Erebor wants to extend loans against assets like GPU clusters, precision manufacturing equipment, crypto (BTC/ETH), and the private equity of VC-backed startups, an area traditional banks can't price. But as of now, loans are at zero, and the business model itself stays unproven until the lending engine turns on.

Catena Labs

Catena Labs, which recently raised a Series A, is an interesting company that filed for an OCC trust charter explicitly to serve AI agents as its primary customers. Founded by Circle co-founder and USDC creator Sean Neville, it is the first chartered-bank application in US history whose customers are software rather than humans or businesses.

The product is agent accounts + payments (ACH, wire, card, stablecoin) + guardrail infrastructure, with specific features like agent identity, counterparty whitelists, transaction limits, and transaction signing. The open-source Agent Commerce Kit is also being distributed as a standard. Strictly speaking Catena Labs isn't a bank because it only does custody, but it's worth paying attention to because it is the only chartered entity among financial-agent solutions.

Vertical Brand

This cluster has no charter of its own and competes on its own brand and workflow specialization on top of someone else's infrastructure. Customer money sits at sponsor banks, in DeFi protocols, or in accounts the user already holds elsewhere. The moat is deep understanding of a specific user segment, and that understanding lets the company package borrowed rails into products that chartered banks can't or won't build.

Meow

Business banking aimed primarily at tech startups, espousing a 'Costco model' where "the depositor gets most of the yield, not the bank." Its 'Maximum Checking' product spreads funds across three sponsor banks (Grasshopper, Third Coast, FirstBank) to create $125M of FDIC-sweep coverage, then passes through 3.5-4.8% APY on top.

In April 2026 it shipped an MCP server with write authority before any chartered US bank had done so. External LLMs like Claude, ChatGPT, and Cursor can open Meow accounts, issue cards, send payments, and handle invoicing (with limits, 2FA, approval workflows, and role-based permissions as guardrails).

Flex

Flex is an "AI-native private bank" for owner-operator businesses (truckers, contractors, solopreneurs, ecommerce operators, and so on). Where Ramp / Brex / Mercury target digital, VC-backed, finance-team-staffed companies, Flex targets $3M-$100M-revenue, non-digital companies (construction, logistics, real estate, agriculture, professional services). It bundles a 60-day interest-free credit card, a Visa Infinite Business card, working-capital lines, and AP/AR automation into a single surface.

The real moat is the five AI agents embedded inside that surface. AP agent automatically parses invoices that arrive by email, matches vendors, and schedules payments (an AI Inbox pattern). Expense agent captures receipts, matches them to transactions, and applies policy rules. Cash Management agent handles cross-account fund movement and treasury optimization, and Underwriting agent automates working-capital and Net-60 credit decisions. Finally, Owner Insights is the 'AI CFO' interface that gives consolidated answers to owner questions like "How is the business doing this month?" or "Can we afford this expense?". Together these five create the CFO seat owner-operators never had before. That's Flex's pitch.

Tuyo

A consumer USDC card. To the user it just looks like a Visa debit card, but the balance lives in their own non-custodial wallet as USDC on Base, and at the moment of a Visa swipe Tuyo pulls from that wallet. While the balance is held, USDC is automatically deployed into a DeFi vault, yielding up to 11% APY (Tuyo keeps 10% of the yield).

The signature feature is "Buy Now, Pay Maybe." Tuyo arbitrarily absorbs some transactions into its own balance (the merchant gets paid normally, the user pays nothing), creating slot-machine-like variable-reward CAC. Rain handles card issuance and Bridge handles virtual bank accounts plus ACH/SEPA/SPEI conversion. Available in the US, EU, and Mexico.

Rise

An international stablecoin payroll service. A company funds Rise in USD, and Rise pays out employees in 190 countries in crypto. The newer feature is Rise Earn. From the company's side, idle USDC balances between payroll cycles are automatically deployed on Aave on Arbitrum to generate 3-5% yield, and employees can optionally do the same with their received USDC. Rise takes 1% of the yield generated and the rest goes to the user, redistributing the NIM that traditional banks used to capture.

Era

The slogan is "Make Claude manage your money." Once the user connects their existing accounts (Schwab, Mercury, Chase, and so on) to Era, Era's MCP server (context.era.app) exposes 33 tools to external LLMs. Claude, ChatGPT, or Cursor can answer "How are my finances?", "Cancel my Netflix subscription," or "How much did I spend on food this month?", and paid tiers can even automate ACH transfers between the user's own accounts. Era itself holds no money (it's only SEC RIA registered) and acts purely as an orchestration layer on top of the user's existing accounts.

On May 6, 2026, Era became the first personal-finance connector in Claude Directory. Nine days later OpenAI launched ChatGPT Personal Finance with Plaid and Intuit, which gave Era an instant competing-product problem.

Closing thoughts

This article started as my own attempt to answer the question, "What does next-generation finance look like?" The answer came out not as the shape of a single company but as the shape of an architecture. Composable Finance, where Substrate, Regulatory Wrapper, and User Surface are independently selected and assembled into financial products that don't reduce back to legacy infrastructure.

Three simple conclusions follow from that architecture:

  1. Architecture choice is now strategic rather than given. Builders pick a cluster the way software founders pick a deployment model. Bank Builder, Vertical Brand, and Pick & Shovel are all bets, and all of them carry risk. None is universally right. The question every founder now has to answer is: which layer do you want to own?

  2. Regulatory Wrapper is the single most consequential choice. The same Substrate and the same surface can produce completely different things depending on the Regulatory Wrapper. Builders should pick the wrapper that fits their activity, not the most prestigious one.

  3. Finance is no longer a single product category to predict. It's a stack to assemble. The winning companies will be the ones that picked and owned the right layers at the right times.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[On My Radar: H1 2026]]>https://paragraph.com/@moyed/on-my-radar-h1-2026 mFYyoQJBFzRXuDdXTSb0Wed, 13 May 2026 03:39:55 GMTMy Own XXL Freshman Class

XXL Freshman Class is an annual list by hip-hop magazine XXL, selecting ten upcoming artists to grace their cover. Its authority has faded in recent years, but not long ago, making the XXL Freshman list was a serious stamp of credibility, a curated bet on which hip-hop artists were about to blow up.

The On My Radar series, which I started last year, is essentially my own crypto version of the XXL Freshman Class. Nobody asked for it, but the idea is simple: highlight projects that are still relatively under the radar, and then follow up on how they develop over time.

I first published five projects on my website at the end of 2025. The original plan was to write a dedicated article for each, but between work and excuses, I only managed to cover Ornn and Numo before half of 2026 had already passed. I never got around to writing up Meanwhile Finance, Merit System, or Street individually, but they are just as interesting as Ornn and Numo, so I hope to cover all three together at some point.

The five projects selected for H2 2025 were:

On My Radar: H1 2026

In this post, I want to introduce new projects worth watching as we approach the end of H1 2026. Previously, I tried to write a separate article for each project, but realistically that's hard to sustain. From now on, I'll cover everything in a single post, and aim to curate On My Radar projects on a semi-annual or quarterly basis.

Seismic & Friends: The Synergy of Privacy + Neobanks

Seismic is the only infrastructure project among this batch, and it's the one I've been following the longest, ever since I first encountered it at the 2024 DBA Research Day presentation about two years ago.

Seismic can be viewed from two distinct angles.

From a technical perspective, Seismic is an alt L1 that provides privacy features. This naturally groups it with privacy-focused infrastructure projects like Aztec and Payy, but Seismic's advantage is its familiar developer experience.

Unlike most privacy-focused blockchains that require separate languages or developer tooling, Seismic follows the standard EVM development experience. Seismic provides privacy through two main mechanisms:

From a business perspective, Seismic can be seen as neobank-focused infrastructure, which puts it in the same bucket as projects like Plasma and Stable.

My sense is that most blockchains targeting neobanks these days try to build services in-house to bootstrap their ecosystem, but Seismic takes a more classic approach, focusing on onboarding external projects. The services currently building on Seismic include:

Many more projects are building on Seismic, and the full list is available on the Seismic website.

As someone who is interested in and bullish on both "privacy" and "neobanks," I think Seismic and the projects building on top of it are worth keeping an eye on.

Privacy and neobanks might seem unrelated at first glance, but as mentioned in this post, privacy is one of the most needed features for crypto neobank users. People obviously don't want their spending history to be transparently visible on-chain, and for teams that want to easily integrate this functionality, Seismic could be a compelling option. Of course, the entry barrier of choosing a relatively unproven new infrastructure when existing chains are shipping their own privacy solutions, along with the lack of an existing user base, are unavoidable weaknesses for any new blockchain. How well Seismic can overcome these is an important point to watch.

Seismic is currently running a play similar to a VC with a stablecoin neobank investment thesis. If even one of these projects becomes a successful case study, it could serve as a springboard for onboarding many more. For now, Brookwell looks the most polished and likely to succeed, but the other services are each tackling different problems, which makes it fun to watch how they all develop.

Tomorrow: The Rise of Niche Fintech

Tomorrow is a lending service for digital creators. Creators can borrow against a portion of their future Google AdSense revenue, and investors can earn interest by lending into the pool.

I feel like I've mentioned this quite a few times already, but I'm a big fan of the specialized stablecoin fintech thesis by Multicoin Capital. In short, stablecoin-native fintechs have very low dependency on intermediaries like banks and card networks, which means they're cheaper to launch and operate, which means they can sustainably serve highly segmented target audiences, much like SaaS.

I personally find this argument quite realistic, and I've been expecting the emergence of fintech services that leverage stablecoin rails to target niche audiences while serving globally from Day 1. Tomorrow Finance could be a representative example of this.

It's not as if fintech for digital creators didn't exist before. Karat feels like a business banking app for creators, and as far as I know, it doesn't offer lending services. Creative Juice, which Mr.Beast invested in, offered advances based on AdSense revenue but was acquired last year by Rho, another business banking fintech. Spotter has Creator Capital, where they purchase existing video catalogs and provide advances in return, primarily targeting large long-form creators. Tomorrow's differentiators seem to be that it leverages onchain infrastructure so anyone can participate as an LP, it uses a loan structure so creators retain content ownership, and it outsources underwriting to external partners.

Tomorrow might look like a simple lending or RWA service right now, but much like SoFi started with student loans and gradually expanded into other layers, Tomorrow could develop into a fintech service specialized for digital creators. Demand for borrowing will always exist, and as a first step toward building a niche fintech for digital creators, lending seems like a solid choice.

Tomorrow is slated for launch in H2 2026, so there isn't much information available yet, but it recently joined Base Batch 003.

Lightcone: Trading the Multiverse

Through 2025 and 2026, prediction markets have become one of the most proven use cases in crypto. Weekly trading volume across prediction market platforms including Polymarket and Kalshi has reached $6B, and there are recent moves to expand into the perp market as well.

Lightcone is not a prediction market. It's an impact market project.

What we trade in prediction markets is fundamentally probability: betting on how likely a specific event is to occur. But what if you could trade the impact an event would have on asset prices? That's exactly what Lightcone's impact market is about. If a prediction market lets you bet on the probability of Saylor defaulting, an impact market lets you trade the impact that Saylor's default would have on the price of BTC.

Let's walk through an example. Say I hold 10 ETH (assuming a current price of $4,000) and I'm worried about the scenario where quantum computers emerge and threaten Ethereum's cryptographic foundations. I can use Lightcone as follows:

  1. Deposit 10 ETH into Lightcone. Receive 10 "Quantum_Realized_ETH" and 10 "Quantum_Not_Realized_ETH."

  2. These two tokens trade independently. Market participants price in what ETH would be worth if quantum computers are realized versus if they're not. For example, if the market consensus is "ETH drops 75% if quantum computers arrive," then "Quantum_Realized_ETH" trades at $1,000.

  3. I want to hedge against this event, so I sell my 10 "Quantum_Realized_ETH" at $1,000 and receive $10,000 in "Quantum_Realized_USDC."

  4. The event resolves.

Options and insurance require you to pay a premium regardless of whether the event occurs, but Lightcone's cost (selling at a discounted price) only kicks in under the bad scenario. If you tried to achieve the same hedge through a prediction market, you'd need to estimate the probability of the event, estimate its impact on price, and then manually calculate how much to bet.

This concept actually has roots in economics. It's known as Arrow-Debreu securities, or state-contingent claims, a theory that has existed since the 1950s. However, it was never implemented in traditional finance due to constraints around settlement, oracles, and liquidity. This naturally brings perps to mind. Perpetual futures also existed only as theory in traditional finance until 2016, when BitMEX's funding rate mechanism brought them to life in crypto. They are now the highest-volume product in crypto. If perps were possible, there's a cautious hope that impact markets could follow a similar path.

Of course, implementation won't be easy. The biggest challenge is liquidity. Since each event creates a separate market, liquidity inevitably fragments, and even prediction markets took years to reach meaningful liquidity. There's also the oracle problem. Determining whether an event has occurred, and how, is still a point of contention even in prediction markets. Events that occur partially or have ambiguous interpretations can make settlement complicated.

Still, what makes Lightcone interesting is that it sits on the extension of the "financializing events" trend that prediction markets opened up, while being an entirely different kind of product. If prediction markets are about "will this event happen?", Lightcone is about "what happens to my assets if this event occurs?"

TechDollar: Assets Scarcer Than Money

TechDollar offers lending products collateralized by private equity in frontier tech companies (AI labs, space, defense, robotics, etc.). Private equity in frontier tech companies will only become scarcer over time, and currently the options for raising capital against it are limited.

The idea Packy McCormick laid out in Scarce Assets is precisely this: when something becomes abundant, adjacent things become scarce and rise in value. As AI drives down the cost of producing software, run-of-the-mill software companies proliferate, but companies with truly differentiated positions (e.g. Anthropic, SpaceX) become scarcer by contrast.

Furthermore, these companies are choosing to stay private longer and longer. Most recently-listed VC-backed companies have underperformed the Nasdaq, and according to PitchBook, the median age of companies at Series D and beyond grew from 6.92 years in 2006 to 9.57 years in 2023, with 44.6% of unicorns being 9+ years old.

The problem is that during this extended private period, the options available to employees and early investors are limited. Employees in particular hold what may be the scarcest asset, but unless the company goes public, it's nearly impossible to convert that equity into cash. Even if they manage to sell on the secondary market, they face wide spreads and the opportunity cost of permanently giving up upside.

TechDollar lets people holding private equity borrow against their shares when they need to buy a house, pay taxes, or diversify their portfolio. If you think about it, whenever an asset class reaches sufficient scale, lending products naturally emerge around it. Public equities have margin lending, real estate has mortgages, and even private funds have something called NAV financing. TechDollar believes private tech company equity is next in line.

The differences between TechDollar and existing private equity liquidity services like EquityBee and Forge are as follows:

Of course, private equity carries inherent risks: it's difficult to value accurately, and disposing of collateral when a borrower defaults is a complex process. TechDollar appears to mitigate this by focusing on the top handful of private companies with active secondary trading and applying conservative LTVs (20-40%).

Agra: The Missing Piece of the Bond Market

Tokenized assets onchain currently total roughly $29B, and private credit is one of the fastest-growing categories. But tokenized private credit has one fundamental problem: once you buy in, there's no way out until maturity.

Say you purchase a tokenized private credit position yielding 14% annually. Three months later, you urgently need cash, but there's nowhere to sell this position. You either wait for the issuer's redemption window (typically semi-annual or quarterly) or just hold. Tokenization solved the technical settlement problem of "who holds what," but it hasn't touched the practical liquidity problem of "can I convert what I hold into cash when I want to?"

This isn't just an inconvenience. Without liquidity, these positions can't be used as collateral either. For DeFi lending protocols like Morpho or Euler to accept tokenized private credit as collateral, they need the ability to liquidate that collateral on the market if a borrower defaults. But if there's no market to sell into, liquidation is impossible. And if liquidation is impossible, the position can't be accepted as collateral. Tokenized private credit ends up existing solely as a 14% yield product, unable to be composed with other assets or leveraged within the DeFi ecosystem. This is a completely different situation from traditional finance, where bonds serve as collateral, get leveraged through repo transactions, and underlie derivatives.

Agra is building a secondary market for tokenized credit to solve this problem.

There's an interesting design choice here: Agra uses a yield-based orderbook rather than a conventional price-based one. The reason lies in the nature of bonds. With stocks, "I'll buy this at $150" is straightforward. But for bonds, the same price means completely different things depending on the maturity. A $98 bond maturing in 1 year versus 10 years has a vastly different yield. In traditional bond markets too, traders quote on yield, not price: "I'll buy at 5.2% annualized." Agra follows this approach because AMMs and price-based orderbooks simply don't fit the nature of bonds as an asset.

Concretely, if you're holding a 14% private credit position with 6 months to maturity and want to exit now, you'd post an ask at a yield above fair value, say 16%. This is effectively telling buyers: "take a higher yield in exchange for holding the remaining duration." The buyer gets in at a discount, and the seller takes a small hit but gains liquidity. The trade happens between market participants without any involvement from the issuer.

For this to actually work, market participants need to trust that they can sell at any time. To achieve this, Agra is designing a three-tier liquidity structure:

  1. In normal conditions, active market makers like Wintermute provide quotes at tight spreads.

  2. Passive vaults stand by at defined price levels.

  3. Even when the market is stressed and active quotes pull back, a separately reserved capital pool steps in to buy at wider spreads, serving as the buyer of last resort.

Agra is currently operating in beta. As the tokenized credit market grows, the need for secondary infrastructure will grow alongside it. The fact that demand naturally follows market growth makes for an interesting positioning.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #16]]>https://paragraph.com/@moyed/things-i-read-last-week-16 kfBEPDMcxRZUMFtmO4tTSat, 09 May 2026 13:07:00 GMT1. Blend and Tempo: Different Solutions, Same Problem

Two projects I found interesting recently, Blend and Tempo, are solving the same problem in different ways. The shared thesis: the pattern of pooling user assets and then reconstructing per-user state at the application layer is broken. Per-user isolation should be embedded in the infrastructure itself.

The classic example in traditional finance is the FBO (For Benefit Of) account structure. When a non-bank fintech offers accounts to customers, it opens a single master account at a partner bank and tracks per-user balances on its own internal ledger. The problem is that this sub-ledger must always match the bank's actual balance.

When it doesn't, disaster follows. The 2024 Synapse collapse was exactly this scenario. Tens of millions of dollars in discrepancies emerged between Synapse's internal ledger and partner bank Evolve's actual balances. Hundreds of thousands of users had their funds frozen. When per-user isolation isn't guaranteed at the infrastructure layer, and accountability is unclear, failures at this scale happen.

Tempo's Virtual Addresses take a different approach. Instead of FBO plus sub-ledger, the blockchain itself serves as the ledger. Each customer receives a dedicated virtual address mapped to the main liquidity pool, eliminating the need for an external ledger. Tempo's infrastructure is the source of truth, making discrepancies structurally impossible. Sweeps (consolidating funds from individual wallets to a master wallet) are either automated at the protocol level or cost negligibly (under $0.001).

A similar structural problem repeats in DeFi. The most common approach when neobanks introduce DeFi yield products is to pool user funds in an ERC-4626 vault and reconstruct per-user positions by parsing on-chain data. It works, but compliance burden scales proportionally with user count, and the pool architecture makes it impossible to isolate actions on one user from another.

Blend takes a different approach: using per-user Gnosis Safes (smart accounts) as compliance boundaries. Neobanks interact only with Blend's SDK/API. Underneath, each user's funds are isolated in an independent Safe. DeFi protocol interactions happen below the Safe layer. In traditional finance terms, it's the separation of Schwab (custody) and BlackRock (asset management).

At first glance, the two projects seem unrelated. Blend solves the compliance boundary problem between DeFi yield infrastructure and regulators. Tempo solves account isolation and settlement in payments infrastructure. But the way I see it, both projects push per-user isolation down from application-layer reconstruction (sub-ledgers, data pipelines) to infrastructure-layer native primitives (smart accounts, protocol-native virtual addresses). They're looking in the same direction.

What the BaaS collapse proved is that the "pool then reconstruct" model breaks at scale. Blend and Tempo are applying the same lesson in their respective domains. Per-user isolation shouldn't be built on top of infrastructure. It should be built into it.

2. Lightspark Grid: Why Platforms Should Own Their Financial Infrastructure

David Marcus unveiled Grid Global Accounts through Lightspark. The former PayPal CEO and leader of Meta's Libra/Diem project, his core argument is that the structure where platforms rent financial infrastructure is fundamentally wrong.

Every platform (marketplaces, gig economy, creator networks) does effectively the same thing. They build the product, construct a two-sided market, and create the conditions for transactions to happen. But the moment money actually moves, platforms lose control. Payment processors, banks, card networks, and FX providers each take a piece. These intermediaries didn't acquire the customers or build the product, yet they capture $2.5T in annual revenue and nearly all the transaction data.

What Marcus emphasizes is the dual cost of intermediation. Fees are visible and intuitive. The more critical cost is data leakage. Every transaction carries context: who earned what, for what, how much, in what pattern. The moment money passes through external infrastructure, context separates from cash. In an era where AI systems run on contextual transaction data, this isn't just a cost. It's a competitive leak.

Grid Global Accounts is the full-stack answer. With a single integration, platforms can offer users dollar accounts, Visa cards, instant settlement in 65+ countries, and stablecoin-based cross-border transfers. Yield, interchange, and FX margin flow back to the platform, not to intermediaries.

The AI agent delegation feature is also interesting. Marcus says he connected a Grid account to an OpenClaw agent and tested it for over a month. The agent autonomously handled online shopping, multi-country bill payments, remittances, and bulk gift card purchases. Users create a funded operating pocket for the agent, set spending limits and approval scopes, and the wallet enforces boundaries natively. Marcus predicts that within 12-24 months, agents will handle a significant share of human-initiated transactions.

3. Stripe Is Trying to Make Crypto Disappear

@snapcrackle published an analysis called Stripe Is Trying to Make Crypto Disappear. The core observation is that Stripe isn't trying to become a crypto company. It's burying blockchain so deep inside enterprise payment infrastructure that customers never need to use words like wallet, gas, or bridge.

The stack Stripe has assembled over the past 18 months makes this intent clear:

On top of this sits Stripe's core $1.9T payment volume and merchant network across 195 countries. No other company has issuance, orchestration, wallets, a settlement chain, a trust bank charter, and merchant distribution all in one orbit.

Tempo exists for a specific reason. During a memecoin frenzy on a major blockchain, a Bridge customer's settlement was delayed over 12 hours and per-transaction costs spiked 35x. The conclusion: institutional payments cannot run on a chain that shares blockspace with other applications. Tempo operates on stablecoin gas with no native token, isolates speculative traffic from payments through dedicated blockspace lanes, and has validators run by designated financial institutions like Visa, Stripe, and Zodia Custody (a Standard Chartered subsidiary). Its governance is closer to DTCC than to Ethereum.

Bridge's Open Issuance functions as an app store for stablecoins. By returning the majority of reserve yield to issuers, Phantom, Klarna, Hyperliquid, and MetaMask are already issuing their own coins on the platform. Stripe doesn't need to own every branded stablecoin. It just needs to own the platform where stablecoins are launched, custodied, and settled.

The most notable observation in the piece is the comparison with Circle. Circle's Arc independently converged on nearly the same architecture as Tempo: designated financial institution validators, stablecoin-native gas, sub-second finality, payments-first design. Two companies independently arriving at the same answer signals that permissioned L1 plus FI validators is becoming a category, not a single product.

Stripe's self-disruption logic is also worth noting. Stripe's existing business runs on card interchange. Stablecoin payments on Tempo bypass card networks entirely. If it succeeds, Stripe is building the thing that erodes the foundation of its own business. But the architecture is designed to capture both sides. When an agent pays by card, interchange is preserved. When it pays by stablecoin, settlement flows through Tempo. Either way, it ends inside Stripe's system. Most companies protect existing revenue and hope new technology arrives later. Stripe is doing the opposite: building the competitor to its own card business on its own timeline, while ensuring the new flows also route through systems it owns.

]]>moyed@newsletter.paragraph.com (Moyed)tempostripeblendlightspark<![CDATA[Things I read Last Week #15]]>https://paragraph.com/@moyed/things-i-read-last-week-15 y21aSMV813c7zUSCgEQiSat, 25 Apr 2026 14:42:00 GMT1. Cloud Clearing: Settlement as a Single Optimization Problem

I came across a piece by Stripe's Mason Reeves called Cloud Clearing. In short, it analyzes how money leaks through fragmented cross-border payment settlement and proposes an architecture that unifies everything into a single optimization engine.

The problem with today's payment infrastructure is that each rail is an isolated system optimizing only one cost dimension:

Take a Korean SaaS company that receives $500K via cards in the US and €300K via SEPA in Germany, while paying $120K to AWS and €80K to a German contractor. Four separate pipelines run independently, and card fees, FX spreads, correspondent bank fees, and SWIFT fees all stack up redundantly.

Reeves' explanation for why this hasn't been unified is straightforward. In financial markets, your counterparty might know something you don't, so FX needs FX specialists, rates need rates specialists, and credit needs credit specialists, each managed separately. But commercial payments are different. The $50K owed to a supplier is a fixed obligation set at invoice time. There's no hidden information. And without hidden information, there's no reason for separate specialists. A single system can optimize all costs simultaneously.

First, we need to define what's being optimized. Every payment has five cost dimensions attached to it at once.

Previously, five separate systems priced each of these independently. Cloud Clearing optimizes all of them simultaneously within a single objective function.

Here are the four tools for optimizing these variables.

  1. Netting: offset opposing obligations against each other. When A owes B $100K and B owes A $70K, only the $30K difference moves. Cost is near zero, so this runs first.

  2. Gross settlement: send the remaining balance after netting through an actual rail. Fedwire, SWIFT, stablecoin transfers fall here. Because rail fees apply, this is used last and minimally.

  3. Bridge financing: when outflows are due before inflows arrive, borrow briefly to cover the timing gap. Bank overdrafts and correspondent bank credit lines are the current version.

  4. Novation: a third party assumes the obligation. In trade finance, a bank purchasing an exporter's receivables (factoring) is the classic example.

Let's walk through how this engine actually works using the Korean SaaS example. Suppose the same system also includes BerlinApp (receiving €400K from Germany, paying $200K to the US) and TexasSaaS (receiving $600K from the US, paying €250K to Germany).

The result: four FX conversions drop to one or two, and the actual amount moving through rails shrinks from over $1M to roughly $520K. That's with just three participants. As more join, more offset paths emerge and efficiency improves further.

2. Yield Is Commoditized: Rebuilding Loyalty Programs on Stablecoins

The core argument in Bradley Freeman's Nobody Cares About Yield is that once stablecoins and DeFi enable every app to become a bank, the axis of competition shifts from yield to loyalty rewards. And that USDC-backed loyalty points can let any brand build the kind of reward network that was previously only possible for major credit card companies and airlines.

Once every consumer app can offer competitive yield, yield becomes table stakes, not a differentiator. The question shifts from "what rate do you offer" to "why should I put my money here."

Starbucks is the decisive case. Its rewards program holds $1.8B in customer deposits and earns roughly $100M in annual interest, but the yield passed back to its 32 million members is exactly 0%. Customers don't deposit for yield. They deposit for free drinks, status badges, and emotional loyalty.

But most loyalty programs fail. The reason is structural: they fund all rewards from their own margins, a zero-sum setup. Too generous and you lose money. Too stingy and customers don't engage. The average American consumer is enrolled in 17 programs and actively uses fewer than half. $10B in rewards go unredeemed annually. Starbucks works because coffee is a daily habit driving five store visits per week. Few businesses have that frequency advantage.

Amex took a different approach.

Amex's structure is fundamentally a three-way value exchange. Members earn points through card usage. Partner merchants (airlines, hotels, retailers) offer discounts in exchange for direct access to Amex's premium customer base. Amex sits in the middle, delivering rewards that cost roughly $200 to produce as $500 in perceived value to members. The result: a $17B reward network running on $10B in annual fees. The key is that rewards come from partner wholesale discounts, not from Amex's own margins.

The problem is that most companies can't become Amex, for two main reasons:

  1. Technical complexity: every partner integration is a bespoke engineering project. Each partner has different databases, catalogs, inventory systems, and settlement rails. Points are effectively money, so bugs translate directly into financial losses or exploitation.

  2. Capital intensity: every issued point sits on the balance sheet as a future obligation. Predicting when and how much members will redeem requires modeling, plus capital reserves against redemption risk. Without an enterprise-scale balance sheet and decades of accumulated partnerships, the model doesn't work. That's why only major credit card companies and airlines have achieved this scale.

This is where stablecoins come in. Freeman's proposal is to back all points internally with USDC instead of inventing custom point currencies. To users, it still looks like branded points. What changes is the infrastructure.

The advantages of switching points to a USDC base:

  1. Points are backed 1:1 by dollars, keeping their value stable.

  2. USDC is already widely circulated, so partner integration shifts from bespoke engineering to shared standard connections.

  3. When every point is backed by real dollars at issuance, there's no floating liability to hedge. Reserves earn USDC yield (~3.5%), turning points from a balance sheet liability into a self-funding asset.

The framing matters. Stablecoin loyalty isn't "using crypto to make points better." It's "structurally lowering the barriers to building an Amex-scale reward network." The technical barrier (standardized settlement rails) and the capital barrier (self-funding reserves) are addressed simultaneously, which makes this different from simple point tokenization.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #14]]>https://paragraph.com/@moyed/things-i-read-last-week-14 rk10j3KH4D3EnpHwo8yeSat, 18 Apr 2026 13:20:00 GMT1. How Stablecoins Spill Over Into FX Markets

BIS published Stablecoin Flows and Spillovers to FX Markets. The core finding is that shocks originating in the stablecoin market spill over into traditional FX markets. Banks and exporters with zero crypto exposure end up paying more to borrow dollars.

The study uses daily data across four stablecoins (USDT, USDC, DAI, BUSD) and 27 fiat currencies. The hardest part of this kind of research is proving causation. When stablecoin inflows and exchange rate moves happen at the same time, it's impossible to tell whether the inflows caused the move or whether both were driven by the same underlying factor (say, economic instability).

The paper solves this by looking at traders who hop between exchanges in multiple countries, chasing price discrepancies. These traders have limited capital and allocate it to wherever the premium is highest. If stablecoin demand drops in Brazil, they shift capital to Korea. Brazil's demand shift has no reason to directly affect Korea's FX market, so by isolating the inflows that entered Korea through these traders' capital reallocation, the paper can test whether stablecoin inflows themselves cause FX effects.

The results are clear. A 1% increase in net stablecoin inflows simultaneously triggers three effects:

Why does this happen?

Large global banks like JPMorgan and Citi do two things at once. One is facilitating stablecoin-to-local-currency exchanges for clients. The other is providing FX swaps (transactions where the counterparty posts local currency and borrows dollars). When a bank takes KRW and hands over USDT in stablecoin brokerage, it accumulates KRW exposure. When it takes KRW and hands over USD in an FX swap, it also accumulates KRW exposure. Both types of KRW exposure stack on the same bank's balance sheet, pushing against risk limits. The bank's only options are to reduce swap provision or widen margins.

Meanwhile, direct dollar borrowing (USD bond issuance, US bank loans, money markets) doesn't flow through these banks' KRW balance sheets and remains unaffected. FX swaps are uniquely vulnerable because they share the same type of currency risk, on the same books, as stablecoin brokerage.

The paper simulates a hypothetical: "What if stablecoin brokerage and FX swaps shared the same books less?" If the two businesses were separated into distinct units so that only half the risk was pooled, the spillover into dollar borrowing costs would drop by about 50%, and the exchange rate impact would shrink by roughly a third. If the two were even more tightly linked, say run off the same desk, the spillover nearly doubles. The degree to which these risks are pooled within a bank directly determines the size of the spillover.

Two policy implications stand out:

Korea is a particularly interesting case. The average KRW stablecoin premium is 2.51%, seven to eight times higher than EUR (0.31%), GBP (0.30%), or CHF (0.31%), currencies from economies of comparable stability. This isn't driven by macroeconomic weakness. It's driven by capital controls: restrictions on foreign exchange access, real-name requirements, and fund transfer regulations. The kimchi premium is evidence of regulatory friction in arbitrage, not economic fragility.

Capital controls compound the problem further. Without them, banks could disperse KRW exposure built up from stablecoin activity to overseas branches or offshore markets. In Korea, that exposure is trapped on the domestic balance sheet, stacking directly with FX swap exposure. The spillover is structurally larger.

Blocking stablecoin inflows outright isn't realistic. The two most actionable paths to reducing FX risk are:

  1. Allow netting of crypto and FX exposures so they offset rather than stack, relieving the pressure on bank books

  2. Improve stablecoin market accessibility to reduce the premium, without touching the structure that links both markets to the same books

The premium being high isn't the problem. The problem is that when the premium swings, it bleeds into traditional markets through bank balance sheets.

2. The Physics of On-Chain Lending

Luca Prosperi (Dirt Roads) published The Physics of On-Chain Lending, and ADCV (Steakhouse Financial) responded with a rebuttal. The central question is straightforward: are Morpho depositors being fairly compensated for the risk they take?

First, the structure. A depositor places USDC into Morpho. A borrower posts crypto collateral like ETH and borrows against it. There's no deposit insurance, no bank absorbing losses. The depositor is the lender, directly.

Two parameters determine the risk:

The borrower has a 16 percentage point cushion between 70% and 86%. The lender has an additional 14pp safety margin between 86% and 100% (insolvency). If the price declines gradually, a liquidation bot repays the borrower's debt to the lender and seizes the collateral, so the lender gets their money back. The lender only loses if the price crashes so fast that collateral value drops below the debt (100%) before any liquidation can fire.

This is where Luca and ADCV diverge.

Luca applies the 1974 Merton credit model. The core insight is that every collateralized loan is mathematically equivalent to "a safe bond + selling insurance." The depositor collects fixed interest while bearing the loss if collateral drops below a threshold, identical to collecting a premium and paying out when an accident occurs. The problem is that most depositors, especially those using frontends like Coinbase or Kraken, think of this as a "USDC savings account" rather than recognizing they're selling insurance.

So what's the fair price for this insurance?

By Luca's calculations, the interest premium depositors should demand ranges from 45bp to 400bp. Even with real-time rebalancing, the floor from sudden-crash risk alone is 45bp. Actual deposit rates sit at 0-20bp. A 10-20x mispricing, per this diagnosis.

ADCV's counter is concise: the math is right but the model is wrong.

Luca's model assumes the lender is fully exposed to collateral declines with no protective mechanism in place. But Morpho has a mechanism that revalues collateral every 12 seconds and triggers automatic liquidation when conditions are met.

The closest traditional finance analogy is a repo trade: lending cash against collateral (treasuries, etc.) with immediate margin calls when collateral value drops. Morpho executes this margin call every 12 seconds, with no human intervention, in a single transaction. Luca's model prices fire insurance as if the building has no sprinklers. Sprinklers don't eliminate fire risk, but ignoring them overstates the premium.

The most important technical difference is how much the lender actually loses when liquidation fires.

At the 86% trigger point, collateral is worth roughly 116% of the debt. The liquidator repays the lender in full, seizes the more valuable collateral, and pockets the ~14% difference as profit. The lender recovers 100%. That 14% comes from the borrower's collateral, not the lender's pocket. Luca plugged in a 5% lender loss rate, effectively miscounting the liquidator's profit as a lender cost.

Empirical data backs this up. Across 19,228 liquidation events and $500M in repayments on Steakhouse-curated vaults, total bad debt was $2.13. Not $2.13 million. Two dollars and thirteen cents. Plugging a conservative 0.3-0.5% lender loss rate into Luca's own model produces 3-30bp. Exactly what the market pays.

The conclusion splits by collateral type. For liquid crypto collateral like ETH and stETH, ADCV is right. The 12-second auto-liquidation effectively protects lenders. For illiquid collateral like private credit or RWAs, Luca is right. When the liquidation mechanism breaks down, the lender is back to selling insurance.

With illiquid collateral, every protection discussed above fails. Crypto collateral has real-time market pricing, per-block monitoring, single-transaction liquidation, and near-zero lender losses. Private credit and RWAs get priced weekly or quarterly, take days to months to liquidate (legal proceedings), and carry fire-sale discounts of 20-50%. Assets that are rarely marked appear far less volatile than they actually are, inflating the perceived safety margin.

How quickly and cheaply collateral can be sold determines whether the risk model holds at all. For liquid collateral, the market is pricing efficiently. For illiquid collateral, there isn't even consensus on what fair pricing looks like.

3. BRLA: Brazil's PIX-to-Blockchain Adapter

I was reading the Dune stablecoin report and came across the BRLA section, a Brazilian real-pegged stablecoin. Curious about how BRLA has embedded itself in traditional financial flows, I dug deeper.

BRLA is a BRL-pegged stablecoin issued by Avenia (rebranded from BRLA Digital in late 2025), backed 1:1 by BRL deposits and Brazilian government bonds. Lifetime processed volume exceeds $1.4B, yet market cap sits at roughly $3M. This gap reveals BRLA's true nature: it's not an asset anyone holds for long. It gets minted during a payment flow, used for settlement, and burned within seconds.

The problem BRLA solves is clear. PIX, Brazil's domestic payment system, is instant, free, and runs 24/7, but it stops at the border. Cross-border payments are stuck on legacy rails: 3.5% IOF tax (Brazil's foreign exchange transaction tax) on every trade, 2-4% bank FX spreads, correspondent bank fees, and 2-5 business day settlement. All-in cost for a typical SMB cross-border payment exceeds 6%.

Using USDC directly doesn't work either. USDC solves the middle of the pipe but not the endpoints. Cashing out USDC in Brazil requires finding an exchange that supports BRL withdrawal, completing KYC, paying a conversion spread, and settling via PIX/TED. BRLA is already BRL-denominated, so cashing out is a 1:1 redemption, not a currency conversion, just getting your money back.

The regulatory angle matters too. Brazil's central bank (BCB) classified stablecoin transactions as foreign exchange operations in its November 2025 resolutions. Using USDC for cross-border payments could trigger a 3.5% IOF levy. But BRL to BRLA isn't a foreign exchange event. You're putting in BRL and getting a BRL-denominated asset. The FX event only occurs when BRLA is swapped to USDC on-chain. Since the conversion happens on the blockchain rather than at the banking layer, IOF applicability may differ. This isn't settled yet. Over 850 companies are pushing back through industry associations, and Brazil's finance minister delayed the consultation in March 2026.

Services using BRLA include:

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Numo, Futures Exchange for Frontier FX]]>https://paragraph.com/@moyed/numo-futures-exchange-for-frontier-fx AXXyCg7UAI4kVpIVVNTLTue, 14 Apr 2026 04:43:27 GMTPreviously

The On My Radar series is back after about two months. I'd been meaning to write, but sitting down to actually do it is always harder than it sounds. As a refresher, On My Radar is a series where I introduce noteworthy projects featured on my website, with a selection cutoff of late 2025. The last piece covered Ornn, which is working to commoditize GPUs. Since that post, Ornn has been added to the Bloomberg Terminal, featured in The Economist, and used as the data source for Kalshi's GPU prediction market.

Studying Onchain FX

I stumbled upon Numo while studying onchain FX. When you look at cross-border payments using stablecoins, you naturally start to decompose the process into two steps:

  1. Moving money from A to B.

  2. Converting that value into the form B actually wants (usually local currency).

Blockchain drives the cost of the first step close to zero, but the question of what advantage it offers for the second step is where it gets interesting. It would be convenient if everyone just accepted dollars (go KRW stablecoin!), but that is not reality, so the costs, slippage, and exchange rate distortions that arise at the conversion stage are very real concerns.

As a proud contributor to the Asia Stablecoin Alliance, I wrote a piece on onchain FX. During the research for that article, I came across several onchain FX projects, and most of them were focused on simply moving FX itself onchain. To me, that looked like replicating what already exists in the real world. I was looking for a project that attempted something better than the status quo, and that is how I found Numo.

TLDR: Numo is building FX derivatives infrastructure for frontier markets from scratch.

FX in Frontier Markets

For all businesses operating in frontier markets, there is effectively no way to hedge currency risk.

The Challenges of Frontier Markets

Frontier markets are countries within the broader emerging-market universe that have relatively small economies and stock market capitalizations: Nigeria, Kenya, Ghana, Egypt, Vietnam, and Argentina, among others. Their foreign exchange markets face a fundamentally different set of challenges than those in developed economies.

First, there is a chronic shortage of dollars. Nearly every external transaction (imports, debt servicing, remittances) requires dollars, but central bank reserves are limited. Governments impose capital controls to stem dollar outflows, and global banks are pulling back from correspondent banking relationships with local banks due to risk (e.g., under Basel III, exotic currencies carry heavier capital requirements). As a result, what is a trivial act in developed markets, "buy dollars," frequently becomes a multi-day negotiation in frontier markets.

On top of that, official exchange rates often diverge from reality. Many frontier countries have two exchange rates coexisting: the official rate managed by the central bank, and the parallel market rate that reflects actual transactions on informal P2P channels and exchange bureaus. The gap between the two can run 10-30% as a matter of routine, which is why derivatives markets are either effectively absent or, where they do exist, dysfunctional.

For instance, in 2016, Nigeria's FMDQ Group launched a USD/NGN futures market. It was designed as cash-settled in naira to sidestep the dollar liquidity problem, and it scaled quickly, reaching over $7 billion in volume by 2017. But two structural flaws rendered the market useless:

  1. The settlement reference was the official exchange rate. Because the official rate was disconnected from the actual market price, a hedge payout would not match the real conversion cost a company faced.

  2. Cash settlement meant no arbitrage mechanism. With physical delivery, arbitrageurs would correct any divergence between futures and spot prices. Cash settlement had no such self-correcting force. And switching to physical delivery was not an option either, since capital controls and correspondent banking issues made it impossible to physically exchange dollars.

The Importer's Predicament

Importers operating in these markets are caught in a particular trap. Revenue comes in local currency, but a significant share of their costs and debts are denominated in dollars. When the exchange rate is stable, this works fine. One sudden devaluation can break the business.

On June 14, 2023, the Central Bank of Nigeria released the naira exchange rate it had been artificially propping up. A large gap had already built up between the real market price and the government-posted official rate, and that gap closed overnight. The official rate devalued by 30-40% in short order. For businesses operating in Lagos, nothing had changed about their revenue, yet they suddenly could not service their dollar-denominated debt.

Importers in frontier markets have been running their businesses in this effectively unhedgeable state for decades. Numo aims to solve this not by imitating the existing FX market, but by bypassing the legacy infrastructure entirely using onchain rails.

From Nairobi to an Exchange

Numo's founder Robert Leifke discovered this problem not through research but by living it. He spent several months in Nairobi, manually quoting and selling USDC/NGN FX forwards to microlenders.

But forwards were structurally expensive. With Nigerian T-bill rates at ~16.7% and dollar rates at ~5%, the interest rate differential gets priced directly into the forward rate, making it worse than spot. Hedging costs started at ~11% annualized, prohibitive for thin-margin microlenders.

As he kept talking to them, Leifke realized that what these microlenders actually wanted was not to lock in an exact future rate. They could live with ordinary fluctuations. What they feared was an overnight collapse like June 2023. The right instrument for them was not a forward but insurance, a call option.

Call options are cheaper than forwards, require only an upfront premium with no credit check, and if the rate does not crash, you simply let them expire. The problem was that pricing options properly requires a forward curve across maturities. In developed markets, this forward curve emerges naturally from deep, physically delivered futures markets. In frontier markets, the basic infrastructure to produce one barely exists. Numo had no choice but to build the ladder from the ground up.

Numo's Stack

Numo is best understood not as a single product, but as a multi-layered stack.

| Layer | Role | Product | | 1 | Spot exchange | ForexSwap | | 2 | cNGN rate discovery, theoretical forward curve | Numo Engine | | 3 | Arbitrage validates theoretical prices, forward curve realized | Numo Futures | | 4 | Insurance against devaluation | Options (upcoming) |

Since Numo chose Nigeria as its first market, let's start with the naira stablecoin cNGN before diving into the stack. cNGN is a naira stablecoin issued by the cNGN Consortium under the Nigerian SEC's regulatory sandbox. It is pegged 1:1 to the naira, with reserves held in Nigerian banks, and circulates as an ERC-20 on EVM chains including Base.

Layer 1: ForexSwap

The starting point of the stack is a spot market for exchanging cNGN and USDC.

Existing AMMs handle volatile pairs like this poorly. According to Leifke, swapping 100,000 USDC for cNGN on Uniswap at the time should have returned 139,200,000 cNGN, but actually returned only 74,125,500 cNGN. That is roughly 46.75% slippage, a $46,000 loss on a single trade.

General-purpose AMMs like Uniswap are structured so that when exchange rates move sharply, arbitrageurs drain the cheaper token from the pool, leaving LPs holding the loss. For pairs like ETH/USDC that move in both directions, this is manageable. But for a currency like the naira that depreciates in one direction, LP capital gets drained unilaterally. What about an order book instead of an AMM? Leifke initially tried to recruit professional market makers to quote on an order book, but no market maker was willing to enter a nascent market with neither meaningful volume nor incentives.

Numo built ForexSwap to solve this.

Unlike general-purpose AMMs that price purely based on token ratios, ForexSwap embeds devaluation probabilities into its pricing curve, allowing LPs to withstand one-directional devaluation without active management. It is also implemented as a Uniswap v4 hook, meaning that as long as DEX aggregators like 1inch already support Uniswap v4, ForexSwap can be routed to without any additional integration work.

Layer 2: Numo Engine

Having spot exchange does not solve the importer's problem. What they actually fear is not today's rate but the rate six months from now. What Importers need next is "an agreement to fix the exchange rate six months out," which in FX terms is called a forward. For example, if you agree today to trade USDC/cNGN at price X in six months, you can convert at X regardless of what the rate does in between. Layer 2's goal is to derive X for each maturity and build a theoretical forward curve.

Fair Forward Price (CIP)

So what should X be?

The answer lies in the interest rate differential between the two currencies over six months. Intuitively: if someone could borrow USDC, convert to cNGN, park it in Nigerian T-bills earning 16.7% for six months, then convert back to USDC at maturity and pocket a risk-free profit, the market must price the six-month forward rate unfavorably enough to eliminate that arbitrage. The relationship that formalizes this is Covered Interest Rate Parity (CIP).

https://www.investopedia.com/terms/c/covered-interest-rate-parity.asp

The USDC rate is produced by global markets on its own. The problem is the cNGN rate. There is no reliable onchain cNGN interest rate benchmark, so Numo built a cNGN rate market itself.

How Numo Engine Works

Numo Engine is an onchain market for fixed-rate cNGN lending and borrowing. The mechanism is straightforward: it trades fycNGN, a zero-coupon bond token redeemable for 1 cNGN at maturity, at a discount. If fycNGN trades at 0.92 cNGN, the market is implying roughly 8.7% annualized over six months. The discount is the rate. A YieldSpace-based AMM intermediates these trades automatically, so cNGN's fixed rate for each maturity is discovered purely through supply and demand, with no rate-setting formula required.

Each market's maturity is aligned with actual Nigerian T-bill maturity dates. Stacking 3-month, 6-month, and 1-year maturities produces a curve that answers "what fixed rate does the market assign for locking up cNGN at each tenor?" This is cNGN's yield curve. Plugging this rate and the USDC rate into CIP gives a theoretical forward price for each maturity, and together these form the theoretical forward curve.

Layer 3: Numo Futures

Having a theoretical forward curve does not mean Importers can actually trade at those prices. It is just a theoretical value; actually hedging requires a counterparty willing to trade at that price. To provide this, Numo is building a futures market that addresses the two flaws of FMDQ:

  1. Settlement tied to the official rate: Numo Futures uses the cNGN/USDC price from ForexSwap directly, a market-generated price, not the central bank's official rate.

  2. No arbitrage under cash settlement: physical delivery enables arbitrage to function. At expiry, a smart contract acts as escrow to deliver USDC or cNGN to the counterparty.

When the futures price deviates from the Layer 2 CIP theoretical value, arbitrage pulls it back.

For example, suppose spot is 1 USDC = 1,500 cNGN, Engine's discovered 6-month cNGN rate is 8%, and the USDC rate is 2.5%, giving a CIP theoretical value of about 1,580 cNGN. But the futures market is trading at 1 USDC = 1,500 cNGN, cheaper than fair value. The following risk-free trade becomes possible:

  1. Borrow 1 USDC (repay 1.025 USDC in 6 months)

  2. Convert to cNGN at spot: 1,500 cNGN

  3. Buy fycNGN at a discount on Engine: 1,620 cNGN at maturity

  4. Simultaneously buy a future: "pay 1,537.5 cNGN (= futures price 1,500 x 1.025 USDC) and receive 1.025 USDC in 6 months"

At maturity, redeem fycNGN for 1,620 cNGN, pay 1,537.5 cNGN into the future to receive 1.025 USDC, and repay the loan. The remaining 82.5 cNGN is risk-free profit. As traders exploit this opportunity, they bid up the futures price until it converges with the theoretical value. This process repeats across maturities, turning the theoretical forward curve into realized market prices.

Layer 4: Options

Once a realized forward curve exists, options become possible in theory. Specifically, call options, the right to buy USDC at a predetermined rate. Importers pay a small premium once. If the rate holds, they forget about it. If it collapses, they exercise. For now, Numo is focused on building sufficient futures liquidity first, and plans to relaunch the options product once that foundation is in place.

To recap: ForexSwap creates the spot market, Numo Engine discovers cNGN's yield curve, and together these two produce the theoretical forward curve. Numo Futures then realizes this theoretical curve through arbitrage, and ultimately, option prices will be derived on top of it.

Who Trades

For this stack to work, every layer needs someone on each side of the trade.

The futures market has a natural two-sided structure:

Once options launch, importers will enter at that layer. Rather than locking into an expensive forward, they buy call options as devaluation insurance for a small premium. Option writers then hedge their exposure using the futures and spot layers below. In this way, carry traders end up indirectly subsidizing the importer's hedging cost.

Closing

The absence of FX derivatives in frontier markets is not a demand problem. It is an infrastructure problem. Numo is assembling that infrastructure on top of stablecoins, one layer at a time. It is still early, the current focus is on building futures liquidity, and whether each layer attracts meaningful participation remains to be seen. But a design that connects spot, rates, futures, and options through a single coherent logic is less a DeFi product and more an attempt to build market infrastructure. In frontier markets, that infrastructure is exactly what has been missing, and that is what makes Numo worth watching.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #13]]>https://paragraph.com/@moyed/things-i-read-last-week-13 fX6LNNyuYELfq57UhyBsMon, 13 Apr 2026 01:38:00 GMT1. Personal Finance Agents: The Next Fintech Isn't an App

Simon Taylor, a regular in this weekly digest, published Wallet Wars Pt 4: The Personal Finance Agent. The core argument is that the next stage of consumer fintech isn't an app — it's an agent that autonomously manages your finances.

Taylor starts from his own experience.

An OpenClaw agent with read-only access to Gmail automatically stored email receipts, generated a categorized CSV, and used Google Maps history to produce a mileage report — assembling a complete tax package ready to send to his accountant. The key is that this process was invisible. He didn't type a prompt during tax season; the agent had already been doing the work in the background.

So why haven't personal finance agents gone mainstream yet? Taylor's answer is simple: everyone is looking for an app.

Open-source frameworks like OpenClaw exist, but they require a dedicated Mac Mini or Docker on a VPS, manually linking LLMs and plugins, installing skills, and training workflows. Taylor calls this the "Tamagotchi problem" — mainstream users don't want a Tamagotchi; they want something packaged. Nvidia announcing NemoClaw, a packaged version of OpenClaw, with Jensen Huang declaring "every company needs a Claw strategy," signals that the packaging phase has begun.

An interesting insight emerges here: no single big tech company sees a user's entire existence. Google sees email, calendar, photos, and location but not bank data. Apple sees device data and payments but not Gmail contents. Banks see salary and transaction history but not Amazon orders. The only entity that can bridge all these silos is "you." A personal agent running on the user's device can connect APIs across companies that refuse to share data with each other.

But finance is a high-stakes domain. Salary, net worth, bills. Failure is not an option. Taylor raises two fundamental problems:

  1. Robustness: Is the agent resilient, well-tested, and backed by liability frameworks and clawback mechanisms when things go wrong?

  2. Incentive alignment: Does the agent serve the user's interests, or does it drag them into meme stock gambling and prediction market slop?

Taylor calls this the guardian agent — an agent aligned to positive outcomes while protecting the user's financial and security integrity.

2. Mastering ZCash: Private Dumb Money

I finally got around to reading Maxime Desalle's Mastering ZCash. It's a comprehensive study covering ZCash's origins, technical architecture, philosophy, and roadmap — essentially a bible for newcomers.

The starting point is simple. Bitcoin is transparent money. Every transaction is permanently recorded and visible to everyone. Wallets are pseudonymous, but the moment you share your address to receive BTC, your entire transaction history and balance are exposed. ZCash takes a different approach. In shielded transactions, the sender, recipient, and amount are all encrypted. The network verifies that the transaction is valid but learns nothing about the transaction itself.

How do you verify a transaction you can't see? In Bitcoin, verification checks four things: 1) inputs exist, 2) no double-spending, 3) spending authorization is valid, 4) outputs don't exceed inputs.

Unlike Bitcoin, where nodes inspect data directly to check these four conditions, ZCash submits a zk-SNARK — a cryptographic proof that demonstrates transaction validity without revealing the underlying information.

Technically, ZCash has three core building blocks:

ZCash has both a transparent pool and a shielded pool. The shielded pool has gone through three generations — Sprout, Sapling, and Orchard — with Orchard being the primary one in use today. Sprout and Sapling required a trusted setup, but Orchard eliminated this requirement using Halo 2. The transparent pool exists for compatibility, auditability, and regulatory compliance, and the two pools are entirely independent systems. No matter how much ZEC sits in the transparent pool, the shielded pool's privacy is unaffected.

The most notable part of the roadmap is Project Tachyon. ZCash's current scaling bottlenecks are threefold: 1) every validating node must store the entire nullifier set, 2) wallets must scan all transactions to find spendable notes during synchronization, and 3) transaction size. Tachyon aims to solve all three simultaneously.

As mentioned in the first section, financial data is increasingly being consolidated in centralized repositories. If surveillance becomes the norm, demand for private dumb money like ZCash could grow significantly. Just as Bitcoin and Ethereum coexist, perhaps a world where Bitcoin and ZCash coexist isn't so far-fetched.

3. The State of Onchain Credit

Private credit may have hit some rough patches recently, but onchain credit seems to be entering its golden age. @Solofunk highlighted notable onchain credit projects on Twitter, and every single one was interesting enough to cover here.

1) Cap x Agra x YieldNest

Agra partnered with Cap to enable instant redemption for YieldNest's ynRWAx, a tokenized offchain credit fund. Cap's structure is interesting: users deposit stablecoins that are then lent to crypto-native borrowers. Rather than traditional collateral, loans are effectively overcollateralized at the protocol level through restaking. Assets restaked on Symbiotic or EigenLayer are seized upon borrower default — Cap remains solvent while restakers absorb the loss. In most cases, restaking protocols like EtherFi extend credit to borrowers and underwrite it, while Cap only provides the capital.

Currently, YieldNest is directly posting $2.7M in OETH collateral via EigenLayer and borrowing $943K from Cap to supply liquidity on Agra's CLOB. ynRWAx token holders can use this liquidity to instantly redeem to USDC before maturity.

2) Valinor

An onchain private credit firm founded by two former Blackstone employees closed a $25M seed round. Not much has been disclosed yet, but Valinor initially plans to extend asset-backed credit to tech-native borrowers with verifiable cash flows who are underserved by traditional credit markets. The longer-term vision is Open Credit — a world where credit moves entirely onchain.

3) Midas

Vault infrastructure provider Midas closed a $50M Series A. TVL stands at $470M. The key development is the launch of Midas Staked Liquidity (MSL), into which Midas deployed $40M of its own capital. MSL is a liquidity vault that supports instant redemptions across Midas' other vaults. Among its private credit vaults, mF-ONE is the largest at $68M, with an underlying strategy managed by Fasanara yielding 12.5% APY. mF-ONE currently serves as collateral on Morpho, where the attractive spread against a 5.5% borrow cost has driven significant utilization ($15.3M deposited, $13.6M borrowed).

4) Silo Finance V3

Silo introduced a dual liquidation mechanism. Beyond traditional liquidation, it offers Collateral-Debt Swap (CDS): when liquidation isn't viable or the position exceeds the CDS threshold, the borrower's collateral is swapped at a discount for the lender's debt. This could be attractive for lenders who don't need immediate liquidity and prefer to hold discounted assets for yield.

5) Loopscale

Loopscale launched PRISM (Permissioned RWA Instant Settlement Market). As covered previously, Loopscale addresses the fundamental limitations of RWA looping by having liquidity providers front capital to bridge the gap, earning predictable returns in the process. The first supported asset is Apollo's ACRED via Securitize, now instantly redeemable on Solana for a sub-2% haircut.

6) Jupiter Offerbook

Jupiter launched a P2P money market. Collateral options are unlimited — even illiquid assets like NFTs, memecoins, and RWAs can be posted. With no price-based liquidation, lenders must set all parameters themselves: collateral type, interest rate, loan size, and duration, pricing risk directly. Whether real demand for collateralizing illiquid assets exists remains to be seen. Even if it does, lenders would likely demand terms so punishing to borrowers that the actual impact is hard to gauge — but it's an interesting experiment nonetheless.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #12]]>https://paragraph.com/@moyed/things-i-read-last-week-12 bgJfKnfcGHwOEzbr7UdjSat, 04 Apr 2026 10:25:00 GMT1. Why You Need to Do Weird Things

0xsmac published Splitting Fives: The House Edge on Wonder. The core argument is that a world obsessed with optimizing only what's measurable ends up killing experimentation and creativity.

Take sports. After Moneyball, baseball converged on three "true outcomes" — strikeouts, walks, and home runs. The NBA followed the same path. Once analytics showed that three-pointers had higher expected value, every team adopted the same strategy. Each team's style and personality disappeared.

Film and music follow the same pattern. Most of the 2026 release slate consists of sequels and franchise entries. Streaming algorithms have standardized even song length and intro structure. When you reward measurable metrics — clicks, views, revenue — rational humans naturally optimize for them. A dominant meta forms, and pointless experimentation becomes increasingly expensive.

To be clear, the piece isn't arguing that optimization is inherently bad. But as everyone knows, genuinely new things are born when strange people pursue strange problems — and that's worth worrying about. 0xsmac prescribes three remedies: create "wide games" like poker where multiple strategies can win, intentionally do non-optimized things like aimless walks, and consume content based on personal taste rather than algorithmic feeds.

These days, as AI tools grow, optimizing and automating everything seems to be treated as a virtue. Perhaps even groups aiming for long-term innovation feel the same pressure. Whether it's FOMO or an inevitable trend is unclear, but pieces like this clearly carry meaning.

2. Why RWA Looping Is Broken, and 3F's Approach

This piece covers the structural limitations of RWA looping and 3F's approach to solving them.

The concept of RWA looping is simple. Deposit a yield-bearing RWA at x%, borrow stablecoins at y%, buy more of the RWA, repeat. In theory, as long as x exceeds y, you can generate leveraged returns.

The problem is that DeFi moves at block speed, but RWAs don't.

With crypto-native leverage, a single flash loan handles borrowing, swapping, posting collateral, and looping within one block — and if it fails, the entire transaction reverts. RWAs break this completely.

Most tokenized funds settle on T+1 to T+3. At 80% LLTV, reaching the theoretical 5x max leverage requires 10–15 loops. Even with a T+1 asset, that's a minimum of 10–15 days; for T+3, it stretches to 30–45 days. A full month just to build the position. Unwinding takes just as long. When the essence of a yield strategy is time, 30 days of drag is a structural problem.

The incumbent approach internalizes this friction inside a single vault. Whether it's a curator-managed leverage vault or a yield-bearing stablecoin, one operator absorbs settlement delays, liquidity buffers, and collateral underwriting. It works, but the costs get passed on to LPs.

3F takes a different approach. Instead of hiding friction, it decomposes it. Each friction point becomes an independent function, and each function becomes a market. The key is the bridge facilitator — which eliminates looping entirely.

When a user wanting 5x leverage deposits $1M, bridge facilitators front the remaining $4M upfront. The full $5M is used to purchase the RWA in a single execution, and once collateral settles on-chain, it's refinanced through Morpho. Position construction time drops from N×T to just T. Unwinding works the same way. For bridge facilitators, the settlement waiting period becomes a short-duration yield opportunity.

3. Body Futurism

Toby Shorin published an essay called Body Futurism. The core diagnosis is that American society is shifting away from software-centric utopias toward new political and social forms centered on the body.

Start with the phenomena. The rise of the Huberman Lab podcast, the mainstreaming of GLP-1 drugs, the seed oil debate, van der Kolk's The Body Keeps the Score holding its place as a perennial bestseller. TikTok and Reels have turned body practices — breathwork, tapping, trauma release stretching — into recombinant content. AI researchers are pivoting to GLP-1 and weightlifting. Masculinity (testosterone, jawlines) and femininity (Botox in your 30s) are being framed as economic necessities in what's become a "charisma economy."

Shorin traces the roots of this phenomenon to the serial collapse of software utopias. Crypto-anarchism in the 1990s, social media democratization in the 2000s, blockchain disintermediation in the 2010s. All promised disembodied spaces accessed through screens, and all ended as little more than narratives. Once the cloud utopias evaporated, people returned to the one thing they could actually control — their own bodies.

This is where "protocolism" gets interesting. The word "protocol" has shifted from meaning a technical standard to describing a body practice. Open-source rule sets like 75 Hard emerged and were commercialized through apps and coaching. Bryan Johnson's "Don't Die" movement is the prime example — a tracking-based longevity religion that looks like freedom but demands constant self-regulation.

But Shorin's real interest isn't trend analysis. It's body politics. The loneliness pandemic, toxic food systems, hormonal disruption, microplastics, declining sperm counts. Modern society is living through "the politics of a society convinced it is sick." Shorin frames this not as a simple crisis but as a precondition for transition. When existing systems — healthcare, food, public administration — are perceived as no longer taking care of people's bodies, people start taking care of themselves. Bodily autonomy rises in the space where institutional trust collapses.

Of course, "return to the body" has historically drifted in dangerous directions. What Shorin means by body futurism is the opposite — different bodies meeting and giving each other energy. The act of sweating together, crying together, colliding together as a new social form in itself.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #11]]>https://paragraph.com/@moyed/things-i-read-last-week-11-1 8xzNQvCrUkqJeiY5PaSJSat, 21 Mar 2026 18:23:00 GMT1. The Intention Layer: How HTTP 402 Gets Filled After 30 Years

Simon Taylor published The Intention Layer. This time he goes beyond simply mapping the agentic payments landscape, addressing at the protocol level why the internet never had a payment layer and why it needs one now. His answer: the Machine Payments Protocol (MPP), being built by Tempo and Stripe.

Taylor's core argument is the shift from the attention economy to the intention economy.

The key metaphor for this shift: moving from "an economy that taxes the journey" to "an economy that prices the destination."

But there is a problem. The internet has no native way for an agent to say: "I'm authorized to spend up to $5 to complete this task. Here are my credentials, here is my payment method, give me the resource." HTTP 402 "Payment Required" was reserved in the original HTTP spec in 1995, but has never been implemented in thirty years.

The scale problem Taylor presents is also striking. The internet sends roughly 4 million emails per second. Visa's peak throughput is about 65,000 TPS. UPI, the most advanced real-time payment system in the world, averages 7,500 TPS. The fastest payment systems operate at 1–2% of the speed the internet already moves messages. Add billions of agents transacting on every API call, every token output, every data retrieval, and Visa starts looking small.

MPP aims to solve this with five principles:

  1. Just as TCP/IP doesn't care which physical network carries the packet, the payment layer must accept any rail—cards, stablecoins, BNPL, Lightning Network, UPI, RTP.

  2. It uses existing WWW-Authenticate and Authorization headers—the same mechanism that powers Basic, Bearer, and OAuth.

  3. Proving payment authorization happens in milliseconds, but settlement happens later, through whatever rail is appropriate.

  4. Not 65,000 TPS, not 544,000 TPS—millions to billions of transactions. Most settlement is batched off-chain with periodic finality.

  5. The protocol works peer-to-peer. Third parties can add value if they want, but the protocol doesn't break without them.

The biggest change the intention economy opens up is the liberation of the long tail.

The attention economy narrowed the range of viable businesses. You had to be big enough to attract an audience, addictive enough to retain them, and valuable enough for advertisers to care. In the intention economy, a service called 10,000 times a day at $0.001 per call generates revenue without a single user, landing page, or brand. Millions of hyper-specialized services, each doing one thing well and earning fractions of a cent per transaction, become viable for the first time.

2. Four Ways to Hand a Card to an Agent

The cards vs. stablecoins debate over payment methods for agentic commerce is heating up. @zephyrlogs broke down this debate on Twitter, and the key point is that "giving cards to agents" isn't one design—it refers to several distinct control models. The differences matter in practice.

Why do cards show up first?

Cards are already integrated into most of the merchant internet. Merchants accept them, checkouts are built around them, and fraud detection, dispute handling, and refund processes all assume cards. Even systems designed specifically for AI agents often start at or stay close to card rails in their payment layer.

The four models are as follows.

  1. Shared Card Access: The most familiar model. The agent uses an existing personal or company card. The advantage is simplicity—less setup, fewer moving parts, lower initial friction. The tradeoff is scope. The credential is broad. When multiple agents or workflows share it, spending is hard to separate cleanly, limits apply at the account level, and it's difficult to trace which agent initiated which purchase.

  2. Dedicated Virtual Card: A new card is issued for a narrower unit. This model is already familiar outside of agentic commerce—Ramp and Brex use it for corporate spend, Privacy.com uses the same logic for consumers. Instead of giving an agent a broad payment credential, you issue one designed for the specific task. Narrower scope means clearer attribution and easier post-hoc review.

  3. Tokenized Card Delegation: The agent never receives raw card credentials. Instead it operates through a delegated token or network-backed credential. Mastercard Agent Pay, Visa Intelligent Commerce, and Stripe Shared Payment Tokens are heading this direction. A virtual card, however narrowly scoped, is still a card credential. The tokenized model pushes delegation logic into the token layer itself. This is also why the current debate gets confusing—some newer systems are discussed as alternatives to card-based commerce, when they're more accurately understood as new delegation layers on top of card-based commerce.

  4. Hybrid Models: Most real systems won't fit neatly into one bucket. A platform might hold funds in a single account while generating separate virtual cards per agent action; a buyer approves a payment method once and then the agent uses tokenized delegated access; or card credentials are used for merchant acceptance while policy is enforced in a software layer before authorization.

Ultimately, the core question isn't "cards or not." It's where does the funding live, where does the credential live, and where does the policy live. How these three separate determines the system's risk profile.

Cards fit naturally in agent-to-business commerce. In that environment, they bring existing merchant acceptance, familiar checkout behavior, and mature dispute and refund processes.

Where cards don't fit is agent-to-agent environments—no browser, no consumer-style purchase flow, no chargebacks, no merchant acquiring stack. Micro-amount, high-frequency payments for API calls, compute, and content access also clash with the card model's transaction structure.

3. Slop Tokenization: An Existential Threat to Blockchains

Gabriel Shapiro introduced the concept of "Slop Tokenization" on Twitter. His core claim is that most tokens represent nothing in an architectural sense.

Shapiro first defines three forms of "clean tokenization."

  1. System-intrinsic tokens: ETH and BTC are the paradigmatic cases. Native elements of the protocol, with rights enforced by consensus rules and no off-chain referent. The chain doesn't represent something—it IS the thing. ETH is Ethereum's access credential, enforced by code every node runs. Remove ETH and the system ceases to function.

  2. Contractually operative tokens: USDC is the clearest example. Circle's Terms of Service make holding USDC the mechanism for exercising redemption rights. The token isn't the dollar, but holding the token activates your right to get dollars. A binding contract makes token-holding the trigger for enforceable claims against the issuer.

  3. Legally constitutive tokens: An on-chain stock ledger is the paradigmatic case. If a corporation's governing documents designate the on-chain state as the authoritative record of holder rights, the blockchain doesn't track ownership—it constitutes ownership.

According to Shapiro, if none of these three patterns apply, it's slop. The one-line test: is the token merely a UI pointer to an off-chain registry that can ignore you when inconvenient?

The case studies Shapiro presents are overwhelming.

Shapiro's argument gets genuinely scary when he points out that this problem extends beyond individual tokens to threaten blockchains themselves. Take Ethereum as an example: ETH itself is the paradigm of clean tokenization, but for ETH to be valuable, Ethereum must be useful, and for Ethereum to be useful, meaningful assets and contracts must actually settle there. If everything on the chain other than ETH is merely "the real thing lives off-chain, the token just points to it"—then what? Ethereum is just the example here, but the broader point stands: from this perspective, a blockchain becomes little more than a censorship-resistant archive of photocopies.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #10]]>https://paragraph.com/@moyed/things-i-read-last-week-10 q718NReKS4VnVdjUjHxdSun, 15 Mar 2026 01:36:00 GMT1. OpenClaw's buy-anything: The API-First Approach to Agentic Commerce

One of the teams executing best in agentic commerce right now is Rye. Rye added a buy-anything skill to the open-source AI agent framework OpenClaw, enabling agents to purchase products directly from Amazon and Shopify via Rye's Universal Checkout API.

There are currently three approaches to agentic commerce:

  1. Protocol-based (ACP, UCP): Merchants must opt into the protocol. OpenAI's ChatGPT Checkout is the poster child — roughly 12 Shopify merchants adopted it before it effectively stalled. Waiting for millions of merchants to opt in is a structural scaling bottleneck.

  2. Browser scraping: Agents navigate checkout pages directly. As CommerceBench demonstrated, this approach is fragile against anti-bot protections, inaccurate, and carries the security risk of handing login credentials to the agent.

  3. API-first checkout: The agent passes a product URL and a payment token to a backend API, and Rye handles everything downstream. The buy-anything skill falls here.

Here's how buy-anything works. The user shares a product URL. The agent collects shipping information, and card details are tokenized through Stripe. The card number never touches the chat, Rye's API, the agent, or the LLM provider. Rye's API then handles product validation, price confirmation, tax calculation, shipping, and order placement in a single flow. V2 added Shopify support, order status tracking, and spending limit controls.

The skill works in practice. In a live demo on the Retailgentic podcast, it passed three challenges specifically designed to break agentic commerce systems, including a complex Amazon purchase with multiple size and expiration options.

The key design decision is that the agent never scrapes or visits the merchant page. It passes only the URL to Rye, and Rye handles the rest. This is precisely why API-first checkout works reliably where browser scraping fails.

Open questions remain. Rye is a centralized intermediary, meaning full trust in Rye's API is required and a single point of failure exists. Spending limits are in place, but the risk of prompt injection manipulating the agent persists. The per-transaction fee structure hasn't been disclosed. Still, with the protocol-based approach stuck behind the structural bottleneck of merchant opt-in, API-first checkout appears to be the only approach that actually works today.

2. Agents Will Use Cards First, Then Stablecoins

Simon Taylor published Agents Will Use Cards First, Then Stablecoins on Fintech Brainfood, arguing that the "stablecoins will kill Visa" thesis is mostly wrong.

The core argument is that cards and stablecoins are complementary, not competitive. Cards authorize the movement of money; stablecoins move the money. Cards are accepted everywhere and have mature controls (single-use, budget caps, merchant restrictions), but settlement is slow. Stablecoins settle instantly and are programmable, but acceptance is still nearly nonexistent.

Taylor sees agent payments evolving in three stages:

Taylor raises another insight: agents can become the new merchants. A vibe-coder builds a tool that presents financial data in four hours. No website, no terms of service, no legal entity. Another developer's agent calls it 40,000 times per week, generating $40 in weekly revenue. Existing payment processors struggle to onboard these "merchants" — not because the technology is lacking, but because onboarding a merchant means taking on that merchant's risk.

3. LLM as Prediction Assistant: Semantic Filtering in Prediction Markets

Anyone interested in crypto has probably thought about — or actually tried — running a prediction market bot with some kind of AI edge. I'd been vaguely thinking along these lines myself, until I stumbled on this paper co-authored by Professor Yongjae Lee (whose class I took during my undergrad years at UNIST) together with Kalshi. In one sentence: the paper proposes using LLMs as an auxiliary tool for prediction market trading.

First, some background.

In prediction markets, lead-lag relationships can exist between different events. For example, when the price of a "Japan recession" event moves first, the price of a "U.S. GDP growth" event may follow days later. Identifying these relationships lets you observe the leader's movement and bet on the follower for profit.

The problem is that the standard statistical method for finding these relationships — Granger causality — produces too many false positives. Granger causality, put simply, is a statistical test that checks whether past data from event A helps predict the future of event B. For instance, it might flag a statistical correlation between "Taylor Swift Tour" and "U.S. GDP," which is pure coincidence. Betting on these spurious links leads to large losses. In other words, Granger causality provides only weak evidence of actual temporal causation.

The paper proposes a better solution: a two-stage framework.

  1. Stage 1 (statistical discovery): Run Granger causality tests on Kalshi prediction market price data and extract the top 100 candidate pairs.

  2. Stage 2 (LLM semantic filtering): For each of the top 100 pairs, ask the LLM: "Is there a plausible economic transmission mechanism between these two events?" The LLM evaluates the presence, strength, direction, and reasoning behind the mechanism, then re-ranks the pairs by plausibility. Only the top 20 enter the portfolio.

The prompt includes the instruction: "Be skeptical — many statistical correlations are spurious." The key point is that the LLM isn't making better predictions. It's filtering out fragile, spurious relationships.

The results are striking. Compared to the pure statistical approach, adding LLM filtering yields:

The driver isn't win-rate improvement — it's loss reduction. LLM filtering cuts the average magnitude of losing trades nearly in half. Pairs that are statistically significant but lack a real economic mechanism are precisely the ones that cause large losses, and the LLM filters them out.

LLM filtering is most valuable during large market moves. When the leader event moves by 10 or more points, the win rate jumps from 53.8% under the statistical approach to 71.4% under the hybrid approach.

There are also cases where the LLM surfaces high-value pairs that statistics miss. The "Japan recession to U.S. GDP growth" pair had a Granger rank of #71 — well outside the top-20 cutoff — but the LLM recognized the mechanism: "recessions weaken domestic demand, and through trade linkages and financial spillovers, downturns in major economies drag on overall growth." It elevated the pair to #5, and it generated $700 in profit.

These results held consistently across holding periods (1 to 21 days), model variants (GPT-5-nano, GPT-5-mini), and post-training-cutoff evaluation windows. What the paper ultimately demonstrates is that the LLM functions not as a better predictor, but as an auxiliary tool for separating signal from noise in the gaps that statistics alone leave behind. Given the chance, I'd like to try incorporating this methodology into an actual trading strategy.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #9]]>https://paragraph.com/@moyed/things-i-read-last-week-9 88qRMkFi74AdMtBeAcFgSun, 08 Mar 2026 01:01:00 GMT1. How Agents Will Pay

Sam Broner of a16z crypto published Agents will pay like locals, not tourists. The core argument is that we're designing agent payments with a tourist's frame. In a bazaar, tourists haggle at every stall, pay cash, and repeat one-off interactions. Locals operate differently. The butcher buys from the fisherman on credit. The tailor settles monthly with the weaver. Broner's diagnosis: agents will behave like locals.

Why will agents act like businesses? Two reasons:

  1. The best experiences are pre-designed. Users don't want an agent that compares prices and negotiates terms at every checkout. They want one that already knows which vendors are reliable, has pre-negotiated pricing, and can transact instantly.

  2. Agents are infinitely replicable, but economies of scale are not. A travel agent booking a million flights a year gets better terms from airlines than one booking ten. Only ChatGPT has the distribution to negotiate direct partnerships with Shopify, Amazon, and Expedia. Smaller startups are stuck with automated browsers or reverse-engineered APIs, paying retail fees. This is why agents will converge to a few dominant players per vertical.

In this structure, there are two payment relationships: 1) user to agent (subscriptions, per-task fees, credit lines), and 2) agent to vendor (pre-negotiated B2B terms, volume pricing, net-30 invoices). This is, in fact, how credit cards work today. The issuer maintains a retail relationship with the consumer; the acquirer maintains a commercial relationship with the merchant.

So are cards the answer for agent payments? Only halfway.

Cards are widely accepted, work well for $20–$1,000 transactions, and include built-in arbitration and cancellation. But two problems remain:

  1. Card technology was designed around humans. Approvers, UIs, 3D Secure — nearly all infrastructure assumes a human in the loop. The number of PSPs, POS terminals, and merchant endpoints is too large, and agent adoption is moving too fast, for a smooth upgrade.

  2. The fee structure. Visa doesn't support sub-cent payments, and there's a fixed 30-cent fee per transaction. Streaming $0.001/second to a compute provider or sending micropayments for API access simply doesn't work on card rails.

This is where stablecoins enter. No 30-cent minimum fee. An agent streaming $0.001/second and a manufacturer settling a $50,000 vendor invoice can use the same rail.

2. The Present State of Agentic Commerce, According to Stripe's Annual Letter

Patrick and John Collison published the Stripe 2025 Annual Letter. Total payment volume: $1.9 trillion (34% YoY growth), processing roughly 1.6% of global GDP. Here's the worldview of the company behind those numbers.

The most notable section is Stripe's five-level framework for agentic commerce:

Stripe's own assessment: we're currently "at the boundary of Levels 1 and 2."

The Collisons compare this moment to the mid-90s when HTTP, HTML, URL, and DNS were being created. "Every Google had an AltaVista" is the key line — an implicit prediction that today's proliferation of agentic commerce protocols will consolidate. Stripe's moves reflect this conviction. ACP, built with OpenAI, is already live in ChatGPT with Walmart, Etsy, and Instacart. SPTs (Shared Payment Tokens) are a new payment primitive that lets agents initiate payments without exposing credentials. Notably, Stripe opened SPTs to competitors like Klarna and Affirm, making them available even for merchants who don't use Stripe.

Stablecoins are the other headline. "Crypto winter but stablecoin summer" — while Bitcoin fell 50% from its October high, stablecoin payment volume doubled to roughly $400 billion. For the first time, stablecoin volume decoupled from crypto asset prices. An estimated 60% is B2B payments. Bridge ($1.1B acquisition) volume grew 4x.

3. Verifiable Intent: Proving What the Agent Spent

Mastercard released an open spec called Verifiable Intent. The core question is straightforward: when an AI agent buys something on behalf of a human, how do you prove the agent actually followed the user's rules?

Today, this is a trust-me situation. A user says "buy headphones under $300 from Amazon." The agent spends $500 at a random store. The only evidence is platform logs — logs the platform controls. Users bear the risk of unauthorized charges, merchants bear the risk of chargebacks, and payment networks can't determine who's at fault.

Verifiable Intent solves this with a cryptographic signature chain. Not logs — signatures. Tamper with any part and the entire chain breaks. The structure has three layers:

Each layer includes a hash of the previous layer, and the signing key must match one pre-registered in the layer above.

A concrete example makes this more intuitive. Say Alice uses a Mastercard.

  1. Her card issuer issues L1 once, certifying Alice's device key.

  2. Alice tells her agent: "Buy wireless headphones under $300 from AudioShop or SoundStore." Her mobile app creates L2 — allowed merchants: AudioShop, SoundStore; allowed products: Sony WH-1000XM5, Bose QC Ultra; amount range: $100–$300; agent key binding included. Alice signs with Face ID and walks away.

  3. The agent finds a Sony WH-1000XM5 at AudioShop for $279. The merchant signs a checkout JWT containing the cart details.

  4. The agent uses this JWT to create L3a (payment proof) and L3b (checkout proof), sending them to the payment network and merchant respectively.

  5. Both sides verify the L1→L2→L3 signature and hash chain. The network additionally checks L2 constraints: is $279 under $300? Is AudioShop on the allowed list?

  6. Everything passes. Payment authorized. Alice gets a notification.

What's interesting is that L3 splits into two. L3a goes to the payment network (includes amount, payee, and card info but hides the cart). L3b goes to the merchant (includes cart contents and checkout JWT but hides payment details). Same L2 underneath, but each party sees only the disclosures relevant to them.

Integrity between the two L3s is guaranteed by a checkout hash. L3a carries it as transaction_id, L3b as checkout_hash — both are hashes of the same checkout JWT. This independently verifies that the payment and checkout reference the same cart.

The constraint types are fairly specific: allowed merchant lists (allowed_merchant), approved products with max quantities (line_items), per-transaction amount ranges (payment.amount), allowed payees (payment.allowed_payee), cumulative budget caps (payment.budget), agent recurrence limits (payment.agent_recurrence), and merchant-managed subscription setup (payment.recurrence). If the agent exceeds budget, uses an unapproved merchant, or buys an unauthorized product, the payment network rejects it at verification time.

On protocol integration, VI is not tied to any specific payment protocol. It can plug into Google's UCP or OpenAI/Stripe's ACP via extension mechanisms. Where the agent payment protocols covered in issues 1 and 2 solve "how do you pay," VI solves "how do you prove the payment was legitimate."

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last week #8]]>https://paragraph.com/@moyed/things-i-read-last-week-8 HTFpcTK7Wh9GEdqNM8qySun, 01 Mar 2026 01:32:00 GMT1. CommerceBench: How Well Can AI Agents Shop?

Lazer Technologies released CommerceBench, a benchmark measuring how well AI agents perform shopping workflows in real e-commerce environments. Reading the paper, the system is designed around three components, though only CommerceBench-Eval has been released so far. CommerceBench-Train and CommerceBench-Generator remain on the roadmap.

The motivation is straightforward. Agentic commerce is a hot topic, but there has been no reliable way to measure whether agents are actually "good" at shopping. End-to-end workflows spanning product search, option comparison, constraint handling (price, inventory, shipping), cart management, and checkout initiation are difficult to evaluate with existing web benchmarks. Commerce tasks involve multiple steps, require detecting hidden inventory or shipping logic, and demand adaptation to promotions and stock changes.

Here's how CommerceBench works. Agents receive structured task programs rather than natural language instructions. These programs specify intent, constraints, and success criteria. For example: "Buy running shoes, under $120, size 10, delivery within 3 days." If the agent adds the right item to cart and initiates checkout, it succeeds. All executions are deterministically reproducible. Failure types, including wrong option selection, constraint violations, checkout dead ends, and infinite loops, are labeled under a classification taxonomy.

The leaderboard reveals that CommerceBench aims to answer four key questions:

The third and fourth questions are particularly important. Starting with Shopify and expanding to Salesforce Commerce Cloud, Adobe Commerce, and Commercetools means quantitatively comparing agent-friendliness across commerce platforms. And comparing browser-based agents with protocol-based approaches (UCP, MCP) means generating empirical data on the interface layer of agentic commerce.

2. Allocation Vaults: Distribution Infrastructure for Tokenized Assets

RWA.xyz published Allocation Vaults: Primer for Institutional Asset Managers, arguing that allocation vaults are emerging as the distribution infrastructure for tokenized assets to reach on-chain capital. Written for institutional asset managers, it's a useful primer worth unpacking.

Institutional tokenization has progressed through three stages: 1) record on-chain, 2) raise capital on-chain, 3) integrate with DeFi infrastructure. In 2022, KKR partnered with Securitize to tokenize a healthcare fund on Avalanche. The hypothesis was "lower the access barrier, and demand will follow." Demand didn't follow.

Demand came from an unexpected place. In 2023, the Fed raised rates to 5.5%. Coinciding with the crypto bear market, DeFi lending rates dropped to around 3%, and on-chain capital flooded into tokenized US Treasuries. Product-market fit was immediate. As of February 2026, tokenized Treasuries exceed $10 billion. Institutional asset managers discovered that blockchain wasn't an operational efficiency tool but a new distribution channel.

After Treasuries came private credit. But here, problems emerged. Treasuries are liquid and settle quickly, so they work naturally on-chain. Private credit is structurally mismatched with DeFi. Borrowers expect instant liquidity while funds offer quarterly redemptions. Liquidation requires immediate collateral conversion, but illiquid positions must be held and waited on. Transfer restrictions block free movement between smart contracts. Tokenization alone wasn't enough. The product had to be restructured to fit DeFi infrastructure.

This is where allocation vaults enter the picture. An allocation vault is a smart contract-based allocation vehicle built on top of DeFi lending protocols. Investors deposit USDC and receive ERC-4626 tokens. The vault allocates funds across multiple isolated markets according to parameters set by a risk manager. Yield is generated in each market and flows back to investors.

RWA.xyz lays out the on-chain distribution stack in five layers:

  1. Asset issuers (fund managers): Create investment strategies and manage portfolios. Once tokenized, assets unlock collateral valuation, lending market inclusion, and distribution channel connectivity. Circle (Hashnote), Janus Henderson, and Fasanara are active here.

  2. Tokenization platforms (product structuring): Package fund manager strategies into on-chain products. This parallels how investment banks structure funds into ETFs or CLOs, with one key difference: composability. Once tokenized, assets can serve as DeFi collateral and be embedded in yield products. Securitize, Centrifuge, and Midas are the leading players.

  3. Lending protocols: Infrastructure that automatically executes lending, borrowing, interest accrual, and liquidation. Protocols themselves make no asset judgments; they enforce rules based on oracle price feeds. Aave Horizon ($32.8B deposited), Morpho ($8.8B), and Kamino ($2.7B) are the most active.

  4. Risk managers: Decide which tokenized assets to accept as collateral, set risk parameters, and allocate stablecoin liquidity. If a risk manager doesn't accept an asset, it can be tokenized but can't attract funding. Steakhouse Financial ($1.6B AUM) and Gauntlet ($1.2B) lead this layer. Bitwise became the first traditional asset manager to launch a vault on Morpho in January 2026.

  5. Distribution platforms: Aggregate stablecoin deposits and route them into allocation vaults. Coinbase is the clearest example, connecting users' USDC lending products to Morpho vaults operated by Steakhouse. Users see yield; they don't see the collateral chain behind it.

Fasanara's mF-ONE illustrates how this structure works in practice. Fasanara is a London-based, FCA-regulated private credit manager ($5B+ AUM). Their F-ONE fund operates with traditional liquidity terms: monthly subscription, quarterly redemption. To make it DeFi-compatible, they did three things:

  1. Legal structuring: mF-ONE is not direct ownership of F-ONE fund shares. It's a bearer debt instrument issued by Midas. Issuance can be KYC-gated, but secondary transfers are free, meaning the token can be used as collateral and seized during liquidation.

  2. Liquidity engineering: Capital is split into three tranches. Roughly 10% of AUM goes into tokenized US Treasuries (mTBILL) for instant redemption. A portion sits in a medium buffer with 7-day notice. The majority maintains original F-ONE fund exposure (monthly subscription/quarterly redemption). The instant-redemption sleeve is the key innovation: it transforms a quarterly-redemption fund into a product that functions in DeFi's real-time settlement environment.

  3. Collateral underwriting: Steakhouse Financial, acting as risk manager, underwrites mF-ONE as collateral and allocates stablecoin liquidity to Morpho's mF-ONE/USDC market. Investors can borrow USDC against mF-ONE collateral, then buy more mF-ONE with the borrowed funds and post it as additional collateral, creating a leverage loop.

What allocation vaults ultimately do is bridge the gap between tokenizing an asset and creating demand for it. Once accepted as collateral, leverage loops generate demand; demand generates yield; yield attracts liquidity. It's a self-reinforcing mechanism. Just as distribution has always been the hardest and most expensive problem in traditional finance, it's the critical challenge on-chain as well. Allocation vaults are an attempt to solve distribution at the infrastructure level.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #7]]>https://paragraph.com/@moyed/things-i-read-last-week-7 iQszwG5aI48RekTPTw7WSun, 22 Feb 2026 13:18:00 GMT1. After Chocolate, Fintech

MrBeast bought a bank. Not quite. He acquired Step, a teen-focused fintech with 7 million customers, built on the rent-a-charter model via Evolve Bank & Trust. But the real story isn't a fintech acquisition. It's the Feastables playbook applied to finance: take a commoditized product, wrap it in distribution nobody else can match, and own the relationship.

MrBeast's core competency is attention science. Starting in 2012, he reverse-engineered the YouTube algorithm to build 1.4 billion subscribers. That's not a follower count; it's a marketing funnel that most banks couldn't replicate with billions in ad spend. Feastables proved the model. A commoditized product (chocolate bars) wrapped in unmatched distribution (460M subscribers), with obsessive product quality driving repeat purchases. As he puts it: "Virality gets the first purchase, quality gets the second." Feastables generated $250M in revenue in 2024 with distribution across Walmart, Target, and 7-Eleven.

What makes Step interesting is its Secured Card program. To the user, it looks and acts like a debit card. Technically, it's a secured credit card. Every transaction gets reported to credit bureaus as a paid-on-time credit payment. Teens enter college with a 700+ credit score already built. That credit score is the stickiness mechanism. A chocolate bar is a one-off transaction at Walmart. A bank account becomes the customer's financial identity. The longer you stay, the more valuable your credit history becomes, and the harder it is to leave.

The core of the attention arbitrage is that the form of trust is different. Traditional banks ask for trust at the point of sale. They need to convince through ads, brand reputation, and marble lobbies. Creator trust arrives pre-loaded, built through years of content and parasocial relationships. That's why the CAC math is fundamentally different. In 2026, neobank customer acquisition costs have soared past $150-300 due to ad-market saturation. MrBeast likely displaces 90%+ of ad spend singlehandedly.

The risks are real, though. Step runs on Evolve Bank & Trust, which was at the center of the Synapse bankruptcy and is under a Federal Reserve Consent Order. Competitors like SoFi, Varo, and Monzo have secured their own banking charters, moving past the BaaS model. And a brand known for "I Blew Up My House" managing teenagers' savings means a single technical glitch could spiral into a PR crisis that no apology video can fix.

Still, the core question stands. Most people stay with their first bank for decades. Capture a 13-year-old user today, and by the time they need mortgages and investment accounts, Step owns that relationship. The threat to banks isn't that MrBeast is a better banker. It's that he's a better onramp.

2. Agents Reshaping Commerce Architecture

BCG published Commerce Everywhere, Agents in Charge. The central diagnosis is that two structural shifts are happening simultaneously. First, commerce is moving from channels to surfaces. Second, AI agents are emerging as the organizing layer for buying decisions.

Commerce escaping channels means purchasing no longer happens at a destination like a website or app. Social feeds become storefronts, media content leads directly to checkout, and AR experiences resolve pre-purchase uncertainty. TikTok Shop's GMV has more than doubled year-over-year and is forecast to surpass $30 billion by 2028. Retail media ad spending is on track to exceed $100 billion globally. The global e-commerce market itself is projected to grow from roughly $25 trillion in 2024 to over $80 trillion by 2030.

More surfaces means more complexity. Managing dozens of touchpoints, offers, and decisions becomes impractical for consumers. Enter AI agents. Consumers state intent ("find me a quiet dishwasher," "plan a long weekend trip") and the agent handles research, comparison, optimization, basket assembly, and checkout. BCG calls this a "seismic shift" in how commerce is conducted.

BCG's proposed architecture has three layers:

The universal cart layer is interesting. Shopify's Commerce for Agents toolkit, launched in 2025, enables AI assistants to search large product catalogs, assemble multi-seller baskets, and complete transactions within conversational experiences. Platforms like Firmly.ai extend this further with multi-merchant, multi-cart checkout. Whoever operates this layer gains visibility into how baskets form across retailers, where substitutions occur, and how consumers trade off speed, price, and convenience.

I've covered agentic commerce in a previous weekly review, where the analysis pointed to infrastructure layers like Shopify and Stripe as primary beneficiaries. BCG's piece goes a step further, asking what strategy brands should adopt in an agent-mediated market. The "destination game" (drive customers to owned channels) or the "evaluation game" (get recommended by agents). BCG argues most brands will choose the evaluation game. Agents bypass offers that are ambiguous, incomplete, or hard to execute. What gets selected is not what appeals to humans, but what machines can easily understand.

The strategic question shifts from "Which channel are we optimizing?" to "When AI agents initiate and optimize buying decisions across every surface, how does our brand get chosen?"

3. EVMbench: AI Hacking Smart Contracts

Paradigm and OpenAI jointly released EVMbench, a benchmark assessing whether AI agents can detect and exploit smart contract vulnerabilities within simulated blockchain environments. While not technically an RL (reinforcement learning) environment, its structure parallels RL systems, meaning it could facilitate future model training.

Here's how it works. Real vulnerabilities sourced from Code4rena auditing competitions are deployed on forked blockchain instances. AI models are given three tasks: detect, patch, and exploit. Eight models were tested. The most striking finding is the asymmetry. Models performed significantly better at exploiting vulnerabilities than at detecting or patching them.

This asymmetry is explained by differences in reward structure. Exploitation is binary: you either stole the funds or you didn't. Clear success/failure signals make it easy for models to optimize. Detection, by contrast, requires exhaustive search through codebases. Patching is a compound task that demands both finding and fixing vulnerabilities.

Why does this matter? Roughly $100 billion currently sits on-chain, and the total crypto asset class is valued at $2.3 trillion. Treasury Secretary Scott Bessent has stated the goal of putting $3 trillion in stablecoins on-chain by 2030. In this context, the fact that frontier models can autonomously discover and exploit smart contract vulnerabilities end-to-end using publicly accessible APIs is not merely an academic finding.

Zooming out, EVMbench is part of a broader trend around RL environments. These are simulated software sandboxes that improve post-training model capabilities. Through verifiable, step-by-step tasks, they reduce reliance on human evaluation while building agentic competence. Anthropic is reportedly investing roughly $1 billion in RL environment acquisition. Companies like Hud are building tools that convert any software into RL environments. Prime Intellect is pursuing decentralized AGI development through an open-source RL environment hub.

EVMbench's evaluation harness and dataset have been fully open-sourced. This can accelerate defensive security research, but it also hands the same tools to malicious actors. The concerns mirror those raised by Anthropic's Red Team in biorisk research, with one critical difference: in decentralized finance, the time gap between discovering a vulnerability and draining funds is near-zero.

As more value moves on-chain, smart contract security becomes a far more urgent infrastructure challenge than incremental benchmark improvements. The asymmetry where models are better at attacking than defending means defensive tooling and audit processes must keep pace with the rate at which offensive capabilities advance.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #6]]>https://paragraph.com/@moyed/things-i-read-last-week-6 u22VTEVUBg9S9IEUKLYnMon, 16 Feb 2026 01:00:00 GMT1. Studying Crypto Neobanks

Pantera Capital published Building Permissionless Neobanks. The piece breaks down banking into four core functions — Store, Spend, Grow, Borrow — and maps out crypto projects trying to wedge into each function on their way to becoming full neobanks. It's an intuitive and well-organized framework.

If mobile gave us the neobank, crypto gives us the permissionless neobank. SoFi started with student loan refinancing, Revolut with low-cost cross-border payments, Chime with early paycheck access. Each captured a customer relationship by solving a specific pain point, then expanded into full banking. Crypto neobanks will follow the same playbook. The question is which wedge to start from.

Store (wallets) owns the entry point, but building a neobank from a wallet alone is hard. Users need to be actively doing things inside the wallet. That's why Metamask added a card and Phantom added in-app perps. Spend (payments) is already commoditized — stablecoin cards are table stakes, and the real move is toward next-gen payment methods like QR-based stablecoin payments. At the chain level, merchant acquisition is the main competitive dimension.

The article's framing of onchain's core value proposition as "making money fast" is sharp. From that thesis, Grow and Borrow — where capital velocity and user engagement are highest — are the best wedges for building a crypto neobank. I agree. EtherFi's path from liquid staking to DeFi strategy vaults to card issuance is a textbook example.

There's also an interesting tension between going local (integration with local partners/infrastructure, specific target customers) versus global (leveraging the permissionless nature of crypto). My take is that the answer satisfying both is "hyper-specific target audience that exists globally." Think a neobank for underage TikTok/YouTube creators — the user segment is extremely narrow, but they're distributed across the entire world.

Lastly, among projects pushing the crypto neobank narrative from a privacy angle, Seismic stands out the most. They already have three design partners: Specie (B2B neobank), Cred (private credit service), and Brookwell (retail neobank). Quietly executing.

2. Vincent: A Security Layer for AI Agent Wallets

Giving an AI agent wallet access is currently all or nothing. Hand over the private key and the agent can do literally anything — prompt injection, hallucination, or a bad strategy can drain funds. Don't hand it over and the agent is useless. There's no way to say: "You can swap up to $500/day on these three tokens, and anything above that needs my approval."

Vincent is an open-source security layer that solves this at the architecture level. The core design principle is complete separation between the AI runtime and credentials. Instead of the agent accessing keys directly, it requests actions through Vincent, which evaluates a policy stack and then executes against an airgapped vault powered by Lit Protocol.

Policies are designed as a composable stack. Daily spending limits, action restrictions (only swaps and transfers allowed, no approvals or bridge calls), hourly transaction rate limits — rules stack on top of each other. The key point is that these aren't application-level guardrails where the AI "decides to be careful." They're infrastructure-level constraints that the AI physically cannot bypass. Even if the AI is fully compromised via prompt injection, secrets remain untouched.

For high-value transactions, Vincent supports n-of-m multi-party approval. Below the threshold, auto-executed within policy. Above the threshold, a configured number of humans approve directly in Telegram. "My agent drained $1,800 at 3 AM" becomes "my agent flagged an $1,800 swap and waited for approval."

Compared to existing agent security solutions: Privy offers server-side key management with per-wallet policies. Coinbase Agentic Wallets use TEE enclaves with session/transaction caps. Vincent's differentiator is providing an airgapped vault (Lit Protocol), composable policy stack, n-of-m approval, and full audit logging — all open source. As agentic wallet infrastructure matures, the distinction between "AI choosing to be careful" and "AI having no choice but to be careful" will become increasingly important.

3. Stablecoins vs Tokenized Deposits: Which Should We Use?

The New York Fed published Stablecoins vs. Tokenized Deposits: The Narrow Banking Debate Revisited. It models whether blockchain-native money should be stablecoins or tokenized bank deposits.

The difference between the two ultimately comes down to backing assets. Stablecoins are 100% backed by safe assets like government bonds. Tokenized bank deposits are backed by a mix of loans and safe assets. Because deposit insurance protects depositors even if a bank fails, banks have an incentive to lend aggressively. The government imposes regulation to keep this in check. Stablecoin issuers only hold safe assets, so they don't need any of that regulation.

The paper's main message is that there's no single right answer:

What makes this different from the historical "narrow banking" debate (the idea that money-creating institutions should only hold safe assets) is scope. The 1930s Chicago Plan and the recent TNB (The Narrow Bank) case in the US tried to apply narrow banking to the entire banking system. This paper only applies it to money used in blockchain transactions. Even if stablecoins are required to be fully backed by safe assets, banks can still issue traditional deposits and make loans. You can have safe money in crypto without killing banks' ability to create credit.

What happens as crypto grows? If existing payments are migrating to blockchain, tokenized deposits become more important. Demand for traditional deposits shrinks, giving banks room to compete in the crypto market. But if blockchain transactions are net new activity, stablecoins become more important. When total money demand expands, banks alone can't keep up.

I've previously written about arguments against tokenized deposits, and my position was: "tokenized deposits can't replace stablecoins, but I can't agree they have no reason to exist." This paper backs up that intuition. Stablecoins and tokenized deposits aren't substitutes — they're complements. Stablecoins are an efficient channel for turning safe assets into money. Tokenized deposits are a channel for enabling bank lending. The optimal mix depends on the regulatory environment. What matters in the end isn't who issues the money, but what assets are behind it.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I read Last Week #5]]>https://paragraph.com/@moyed/things-i-read-last-week-5 4F75mkvNq0X3UfmfX8GbMon, 09 Feb 2026 01:00:00 GMT1. Tenbin Labs: A Hybrid Approach to RWA

Imagine holding gold as a token onchain. Two problems emerge.

Tenbin Labs approaches this problem by asking: "Why rely solely on onchain liquidity?" The core idea is simple. Instead of holding physical assets directly, synthesize price exposure through stablecoin collateral + CME futures hedging.

Consider issuing tokenized gold (tGLD). When a user deposits $1,000, you need to provide $1,000 worth of gold exposure. The traditional approach requires actually buying and storing gold, which creates liquidity problems. Gold spot markets aren't open 24/7, and large sales significantly impact prices. Tenbin's solution is different. Instead of buying physical gold, they keep stablecoin collateral onchain and go long CME gold futures.

What effect does this create?

When Tenbin issues tGLD, it assumes the obligation to "pay users whatever gold appreciates by." This obligation itself represents short gold exposure from Tenbin's perspective (if gold rises, Tenbin loses). To offset this, they take a long position in gold futures. Of the $1,000 collateral, $100 (10%) is deposited as margin at CME, and they go long $1,000 worth of gold futures. If gold rises, the futures long profits, and this pays users. If gold falls, the futures lose, but users also receive less, so it offsets. tGLD issuance obligation (short exposure) + futures long = delta neutral. This is the principle of hedging. Why CME? Daily metals futures volume reaches $454 billion, FX at $86 billion. This deep liquidity is impossible to achieve onchain.

So where does yield come from?

If you're delta neutral, you can't profit from price movements. There are two sources. First, onchain DeFi yield. Over 75% of collateral remains onchain, and deploying these stablecoins in low-risk yield strategies like Aave or Morpho can generate 4-6% annually. Second, futures basis yield. The difference between futures and spot prices is called basis, and as this difference converges to zero at expiration, it generates either yield or cost.

Let's look at gold (tGLD) as a concrete example.

Contango is when futures prices exceed spot prices, and gold is a typical case. Storing gold costs money (insurance, warehousing), so since buying futures avoids these storage costs, the market adds a premium. If spot is $4,000 per ounce and 3-month futures are $4,040 (basis +$40), the annualized financing cost is (40 × 4) ÷ 4,000 = approximately 4%. To issue gold tokens, Tenbin must go long futures, but in contango, longs lose over time. Buying futures at $4,040 and settling at $4,000 at expiration means a $40 loss. This cost must be offset by onchain yield. If off-chain margin is 10%, 90% remains onchain, and to offset 4% cost, onchain yield needs to be at least 4% ÷ 0.90 = 4.44%.

Now consider the Brazilian real (tBRL).

Backwardation is when futures prices are below spot, typical for high-interest currencies. Brazil's benchmark rate is about 15%, while the US is 3.5%. Holding BRL earns this interest differential, but dollar-based investors holding BRL spot face exchange rate risk. The futures market reflects this interest differential by trading BRL futures below spot. If spot is 1 BRL = 0.20 USD and 3-month futures are 1 BRL = 0.195 USD (basis -0.005), annualized yield is (0.005 × 4) ÷ 0.20 = approximately 10%. When Tenbin issues BRL tokens and goes long futures, buying at 0.195 and settling at 0.20 generates 10% annualized return. Opposite to contango, hedging itself becomes a yield source. Adding onchain DeFi yield, if 90% of collateral generates 5% annually, total yield becomes (0.90 × 5%) + 10% = 14.5%. This "captures" BRL's high interest rates without holding physical currency.

Ultimately, Tenbin token yield structures vary based on the asset's futures curve state. Contango assets (gold, most commodities) incur hedging costs that onchain yield must offset for positive returns. Backwardation assets (high-interest currencies) turn hedging itself into yield, combining with onchain returns. Users need to understand this difference.

2. Simon Taylor's Agentic Payments Map

What infrastructure is needed if AI agents make payments? Simon Taylor drew a map answering this question in Fintech Brainfood. The key insight: "Multiple protocols emerging isn't redundancy—they solve different problems at different layers."

Taylor decomposes this problem into six layers.

  1. Agent Communication Layer: How do agents talk to each other? MCP (Model Context Protocol), created by Anthropic, standardizes how agents communicate with external systems. Google's A2A (Agent-to-Agent) handles security and coordination when multiple agents collaborate in enterprise environments.

  2. Trust Layer: Can we trust this agent? From a merchant's perspective, after years of blocking bots, requests now arrive claiming "I'm a legitimate agent." ERC-8004 proposes an onchain registry for agent identity, reputation, and MCP endpoints. Visa TAP (Trusted Agent Protocol) issues signatures proving Visa has verified the agent. Why does this matter? A single compromised agent credential could execute thousands of fraudulent transactions per minute. Human fraud can be caught after the fact, but machine-speed fraud must be prevented in advance.

  3. Mandate Layer: Does this agent have payment authorization? Even if an agent is trustworthy, whether it has payment authority is a separate issue. Google's AP2 (Agent to Pay) introduces the "mandate" concept, subdivided into Cart mandate (what can the agent buy?), Intent mandate (what did the user approve?), and Payment mandate (which card to use?). Visa Intelligent Commerce (VIC) and Mastercard Agent Pay (MAP) solve this at the card network level, with rollout planned for H2 2026.

  4. Transaction Flow Layer: Discovery, negotiation, checkout. This is the most competitive space. ACP, created by OpenAI and Stripe, is a standard dictating "structure carts like this, generate payment tokens like this," already live in ChatGPT with Walmart, Etsy, and Instacart. UCP, created by Google and Shopify, takes a different approach—merchants publish information for agents, who then discover and negotiate.

  5. Authentication Layer: Security and fraud prevention. This includes existing card payment systems like 3D Secure and chargebacks, though how to apply them to agent payments is still being defined.

  6. Payment Rails: Actual settlement. Card networks, ACH, wire, or stablecoins. The rails themselves aren't new, but the logic for agents to choose which rail and when needs to be built.

Additionally noteworthy is Google's vertical integration. A2A (communication) → AP2 (authority delegation) → UCP (commerce) forms one coherent stack.

Google dominates the search market and controls the answer to "where can I buy X?" If merchants must support UCP to appear in Gemini shopping results, adoption becomes mandatory, not optional. Taylor predicts: "Google has created internet standards before and will do it again." Of course, OpenAI's ACP being deployed first can't be ignored. ChatGPT shopping is already working, and the winner of the standards war remains undecided.

3. Multicoin's Manifesto Without Kyle Samani

With Kyle Samani's departure, Multicoin Capital published a long-overdue investment thesis update. It's the first since their 2019 Crypto Mega Theses, roughly seven years ago, and lays out eight investment themes. The full report is dense, so here are the three themes that resonated the most.

The first is Fintech 4.0.

The core claim is that stablecoins and blockchains provide fundamentally new financial rails for the first time in half a century, enabling a wave of specialized fintechs. Historically, early fintechs started by focusing on specific segments. SoFi did student loan refinancing, Chime offered early paycheck access, Greenlight provided teen debit cards, and Brex served founders who couldn't get traditional business credit. But the focus didn't last. Revenue caps on interchange, rising compliance costs, and dependence on legacy banking systems forced these companies to expand beyond their original markets. Unit economics simply wouldn't allow them to stay niche.

Multicoin argues that blockchain rails change this equation. With stablecoins, there's no need for sponsor banks, card networks, reconciliation teams, or dispute operations. Frontends can retain more margin, and unit economics can work even while serving narrow markets. The projection: thousands of hyperlocal fintechs will emerge, from neobanks for migrant workers in specific countries, to trading apps for crypto degens, payroll tools for digital nomads, and remittance apps dedicated to a single corridor.

The theory is compelling. But whether unit costs actually drop that much remains an open question. Sponsor bank and card network costs disappear, sure, but operations and KYC costs persist. Licensing requirements vary by jurisdiction, and customer support infrastructure is still necessary. Ultimately, this claim can only be validated by comparing actual unit economics data from teams that have built and operated stablecoin-native fintechs.

The second is Credit Vaults (CVs).

A CV is a pool of capital that anyone with a wallet can deposit into, while an operator (e.g., Gauntlet, Steakhouse Financial, MEV Capital) manages deployment according to pre-determined code parameters. So far, CVs have mostly been used for crypto-internal activities like market making, lending against crypto collateral, or earning leveraged DeFi yield.

What Multicoin finds compelling is the structural innovation. CVs separate roles that are traditionally bundled in private credit: liquidity providers own the assets, operators decide deployment strategy, and risk management is enforced automatically by smart contracts. No manual judgment or after-the-fact cleanup required. Why does this matter? The problem in lending today is that a well-capitalized lender in the US has no way of lending to a startup in Southeast Asia. There's no discovery process, no practical way to move the money, and no efficient mechanism for accountability and recourse.

CVs aim to solve this by having local operators source debt opportunities on the ground while drawing from a global capital pool. Consider an Indian software services company earning steady EUR revenue from European clients, yet facing high borrowing costs domestically. A CV operator can underwrite the business, secure funding from global lenders, and use cryptographic tools like zkTLS to verify accounts receivable and fund flows. Energy financing (Daylight financing solar panels) and creator financing (CreatorFi lending against YouTubers' future earnings) work within the same framework.

The third is Entertainment Finance.

Multicoin diagnoses a bifurcation of finance into necessity-driven finance and entertainment-driven finance. Short-form video replaces long-form, swipe dating replaces relationship building, zero-day options replace 401(k)s. The preference for immediacy maps directly onto finance. Declining housing affordability accelerates this trend. Among 18-to-29-year-olds, the share who believe the American Dream is still alive collapsed from 56% in 2010 to 21% in 2024. When the conventional path is blocked, people choose to roll the dice.

There are diverse views on entertainment finance. Some argue it's socially negative-EV. Others see it as individual freedom. Multicoin's stance is unambiguous: "This is a freight train we have no interest in standing in front of." The speculation and gambling market will keep growing because people enjoy games, risk-taking, and the chance to win big. What can change is how extractive these games are. In traditional gambling, casinos have a built-in house edge, sportsbooks charge high fees, and CFD brokers internalize B-Book trades. The hold rate on legal US sports betting climbed steadily from 6.7% in 2018 to 9.9% in 2025. On blockchains, P2P trading, transparent pricing, lower fees, and rules set in code are all possible. Products like Novig and Triumph point in this direction. The argument is that crypto isn't creating entertainment finance. It's serving already-existing demand more fairly and cheaply.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Ornn, How Do We Price Compute?]]>https://paragraph.com/@moyed/ornn 1taNrzDuCZU0BC0E1BpVMon, 02 Feb 2026 01:08:00 GMT_Thank you to the Ornn team for their valuable feedback on this piece._

On My Radar

If you visit my website, you'll find a section called 'On My Radar'. This section lists companies that have caught my attention - projects I'm actively following. The projects currently listed on the website are from Q4 last year, so newer projects I've been watching haven't been updated yet, but I still have high expectations for all of them.

In an era where research has become such a public good, the only thing a hobbyist researcher can really offer is taste. As many researchers around me have noted, it's less about being right or wrong, and more about showing "this is what this person finds interesting" - sharing one's taste can be entertaining for readers.

#001: Ornn

Among the various projects on my roster, I debated which to introduce first, and decided to start with my favorite: Ornn. In one sentence:

Ornn is a company building financial products that allow AI companies, investors, and data center operators to manage their exposure to volatile compute prices.

Preparing this piece taught me a lot about GPUs and the AI infrastructure market, which was genuinely fun. Since this isn't my area of expertise, please let me know if anything is incorrect - I'd really appreciate the feedback!

Name Your Price

Let's start by defining the problem. The issue is that AI companies, hyperscalers/neoclouds, and data center operators are all investing massive amounts of capital into the AI industry, yet there's no benchmark for the price of "compute" - the core of these investments.

For example:

I can't say how realistic these examples are, but the point is: relative to the scale of the AI industry, much of the underlying economics is conducted through bilateral phone negotiations - essentially vibes-based.

Case Study: CoreWeave & MagAI Ventures

For another example, let's look at the Q3 10-Q of CoreWeave, a prominent neocloud (they own GPUs and rent them to AI companies).

As this tweet shows, CoreWeave received a $230M prepayment from MagAI Ventures last August in exchange for providing pre-negotiated, fixed hourly compute prices to MagAI's portfolio companies. Another condition: if MagAI's portfolio companies don't use the compute from CoreWeave, the prepayment must be returned at 12% annual interest.

Why would MagAI Ventures agree to such terms? Likely they wanted to provide value-add to their portfolio companies by securing fixed-price GPU compute. If compute futures or options existed, they could have easily hedged through those instruments. But since they don't exist, they had no choice but to enter into this unusual arrangement with CoreWeave.

Of course, large companies with capital and information advantages can predict utilization rates and negotiate cheaper deals through scale. But companies without these advantages are exposed to information asymmetry and have no way of knowing where their deal stands relative to the market.

Benchmarking Against Power

When the Ornn team discovered this problem and started working on it, they naturally looked at oil first - oil is the classic commodity. But the problem is that compute and oil are fundamentally different. Oil can be stored; compute cannot. An idle H100 at 3am on Tuesday has zero value. So we call commodities like oil "stock commodities" and commodities like compute "flow commodities."

The quintessential flow commodity is electricity, which has these characteristics:

Understanding these characteristics of electricity helps when thinking about GPU compute as a commodity.

About Ornn

Before looking at Ornn's product lineup, let's examine the team and founding story.

Team & Founding Story

When Kush Bavaria and Wayne Nelms were consulting for private equity firms that lend to data centers, they kept hearing that PE firms were extending credit to GPU infrastructure companies but had no way to hedge their exposure. Without a benchmark to reference, there were no derivatives to hedge with, and financial institutions had no way to properly manage risk. The two founders identified this problem and created Ornn. The team consists entirely of MIT alumni with backgrounds at Google, SIG, Optiver - people who understand both trading and technology.

Products

Now let's finally look at what Ornn actually provides.

1) OCPI (Ornn Compute Price Index)

The core is OCPI (Ornn Compute Price Index) - the GPU compute index.

Everything starts with a benchmark. The detailed methodology isn't public, but separate indices exist for each GPU type, and they're built on actual executed transaction prices - not just quoted prices ("selling at X"). Regional weighting is also factored in, reflecting that GPUs trade differently across geographies. Following the electricity model, settlement is Asian-style.

You can see the H100 index value on the website, but other metrics appear to require payment. While preparing this piece, I found two articles written using OCPI data, suggesting it's already being used as a meaningful indicator by investors:

Both articles using Ornn's data are behind paywalls. I personally paid to read the second one, which briefly summarized: the article uses data from Ornn covering A100 SXM4, H100 SXM, and H200 from October 2 to December 30 last year to examine how utilization affects price volatility for each GPU type.

From public information, partners contributing to OCPI include:

2) Compute Swaps

Built on top of OCPI is the actual hedging product: Compute Swaps. A compute swap is a contract where two parties agree on a fixed GPU hourly rate for a future period. At settlement, the losing party pays the difference in cash based on OCPI - no actual compute changes hands.

For example, if I enter a 30-day swap contract for 10,000 GPUs at $2/GPU-hour, at month's end Ornn calculates the average of daily OCPI values. If the average is $2.30, my counterparty pays me ($2.30 - $2) × 10,000 = $3,000.

The beauty of swaps is you don't need to change who you buy compute from. Keep using your existing hyperscaler or neocloud - if the GPU hourly price rises and your bill increases, the swap payment offsets it. A very convenient structure. Notably, the first-ever compute swap occurred through Ornn on December 11th last year!

3) Compute Futures

Compute futures are compute swaps with standardized terms, contract sizes, and expiration dates, making them easier to trade on an exchange. Ornn currently operates under CFTC de minimis exemption and is pursuing a DCM (Designated Contract Market) license, so compute futures are not currently available.

However, Architect - the institutional exchange run by former FTX US President Brett Harrison - recently announced they'll support GPU perpetual futures contracts based on Ornn's data, suggesting compute futures trading will become available to institutions via OCPI.

4) Memory Futures

Recently, Ornn announced they'll also launch futures on memory prices like HBM, beyond just hourly GPU rates.

I'm no expert, but apparently when running inference on large LLMs, the actual bottleneck isn't GPU computation but data transfer from memory. HBM is made by stacking memory vertically and connecting it with microscopic copper pillars - extremely difficult to manufacture, which is why only three companies make it: Samsung, SK Hynix, and Micron. Prices have fluctuated over 250% in the past two years. Despite this volatility, no standardized financial product existed to hedge HBM price risk.

Memory and GPU prices often move together during AI demand spikes, but respond to different supply constraints. By offering hedging products on both, Ornn enables participants to hedge more precisely.

5) Residual Value Swap (RVS)

Ornn also offers GPU Value Protection through Residual Value Swaps. You pay a quarterly premium, and at the contract's end, if you elect to sell your GPUs, you're guaranteed to receive an agreed-upon price.

This addresses a core problem in GPU ownership: hardware worth tens of millions can see resale values collapse overnight due to new GPU generations or shifting infrastructure requirements. Traditional financing treats GPU terminal value as uncertain, forcing operators to accept conservative terms. RVS transfers this end-of-life price risk to another party.

The key insight is that a GPU's residual value is closely tied to its future revenue potential and this is exactly what OCPI tracks. By establishing a reliable benchmark for compute prices over time, OCPI provides the foundation for pricing residual value risk. Without knowing where GPU hourly rates are heading, you can't meaningfully price what a GPU will be worth in three years.

For datacenter operators, this enables better financing terms, reduces equity volatility, and allows investment decisions based on expected performance rather than worst-case scenarios. For lenders, it provides a defined floor under collateral value, supporting higher advance rates and insulation from technology obsolescence.

Related Companies

Given how important this problem is, I looked for other players besides Ornn. I couldn't find anyone else building compute-based financial products, but Silicon Data is a notable player building compute price benchmarks.

Silicon Data

Silicon Data, backed by DRW and Jump, aggregates price data from global sources to provide GPU price benchmarks. The detailed methodology isn't public, but you can see actual metrics by creating an account. I can't precisely compare OCPI and Silicon Data's benchmark without knowing their methodologies, but my understanding is: Silicon Data aggregates quoted prices ("selling at X") from various services and benchmarks them, while Ornn builds OCPI based on actually executed transaction prices. At the company level, Ornn provides financial products on top of OCPI, while Silicon Data stops at benchmark provision.

Spot Marketplace

Not direct competitors to Ornn, but related companies include GPU compute spot marketplaces like SF Compute and Compute Exchange.

I see two main implications of these spot marketplaces for Ornn:

  1. Data providers: They can supply real-time transaction data to Ornn's indices. These marketplaces are the source of the "actual transaction-based prices" that Ornn emphasizes.

  2. Potential customers: They can use Ornn's products to hedge their GPU price exposure. For example, if SF Compute enters long-term GPU contracts, they could use Ornn futures to manage future price fluctuation risk.

What Gets Unlocked

Groups that need Ornn's financial products can be divided into three categories.

First: AI companies wanting to stabilize costs. When training large models costs enormous compute, and that cost can swing dramatically within months, it's inherently stressful. Second: data center operators wanting to stabilize revenue. GPU infrastructure construction is a multi-year investment, and if you can sell future capacity at a known price, investment risk decreases.

The third group is financial institutions that need to manage exposure - let's examine this group more closely.

1) Project Financing

Currently, when data center operators request project financing from banks, banks naturally ask about revenue projections. Operators present their scenarios, but banks are understandably anxious.

Compare this to power plant financing. If a power plant signs a 10-year Power Purchase Agreement (PPA) and additionally hedges price risk with electricity futures, banks can treat that cash flow as "nearly guaranteed." Collateral value is clear, risk is quantified, so they can provide large-scale funding at low rates. Through Ornn's services, compute can achieve a similar structure, capital costs decrease, and more projects become economically viable.

2) Residual Value Insurance (RVI)

In ship and aircraft leasing markets, Residual Value Insurance exists - when a company buys an aircraft to resell in 10 years, insurers guarantee a minimum price for that aircraft after 10 years, compensating the difference if it falls below.

With OCPI now established, Ornn has launched exactly this: Residual Value Swaps (RVS) for GPUs. Companies buying or leasing GPUs can now use RVS to transfer risk and secure funding on better terms.

3) ETC

Beyond this, things like GPU-based Asset-Backed Securities become possible. Ornn already works with USD.ai on exactly this kind of structure: USD.ai provides the loan and financing for GPU infrastructure, while Ornn reduces financing costs by offering a put option on the hardware in year 3 or 4. This gives lenders confidence in the terminal value, enabling better terms for borrowers.

It's all one story: once a price benchmark that market participants can agree on emerges, risk becomes measurable, and capital can be deployed more comfortably.

Risks

But the path to this future obviously isn't easy. The biggest challenge is creating a proper index. Without a well-constructed index, hedging won't work perfectly, and if market participants don't trust it, the index becomes just a number with no greater value. The problem is GPUs may be even trickier than electricity - beyond GPU model and region, performance varies based on workload type, setup configuration, and more. Properly measuring these factors and creating a reasonable index seems critical.

Beyond this, there are liquidity and regulatory issues, but if the above risk is solved, the market need appears so strong that these should resolve quickly.

Q&A with the Team

I had the chance to speak with the Ornn team and asked them some questions about how OCPI works under the hood and where they're seeing traction:

Q. What's the minimum transaction volume needed for a price point to be included in OCPI?

A. There's no minimum transaction size, the index is volume-weighted, so larger transactions naturally carry more weight.

Q. How do you handle the heterogeneity problem, same GPU model but different performance based on workload type, network topology, or cooling setup?

A. Through weighting. Different configurations are weighted appropriately to reflect their relative performance characteristics.

Q. What's your approach to regional weighting? How granular does geography get?

A. City-specific. Prices can vary significantly by location, so the index accounts for geography at that level of granularity.

Q. How many data partners contribute to OCPI?

A. Over 10 data partners currently contribute to the index.

Q. How do you verify that reported "executed prices" are real and not manipulated?

A. We have contracts with our data providers and actually parse their invoices directly. This ensures we're working with verified transaction data rather than self-reported figures.

Q. Which customer segment is moving fastest: AI companies, data center operators, or financial institutions?

A. For the hedging products, it's operators. For the Architect partnership, we expect it will attract more speculators looking to take positions on compute prices.

Closing Thoughts

Compute, which may become the most important commodity going forward, deserves a real market. Just as oil has WTI and electricity has LMP, compute needs a price everyone can reference. Ornn is building the index (OCPI) that will become that market's foundation, and derivatives that let market participants trade risk based on that index.

Until now, buying compute has been like buying a used car. You can't tell if the seller's price is fair, comparing to the shop next door is difficult, and knowing what prices will be in six months is even harder. The more widely Ornn's products are used, the more transparent prices become, the more efficient capital deployment becomes, and ultimately, the lower compute costs will be. This is good for companies building AI and everyone using AI.

On a separate note, there's one more thing I'm excited about. In a past piece, I wrote about how we need to expand supply of critical infrastructure to lower costs, and redesign that expansion as investment opportunities accessible to younger generations. As infrastructure like Ornn matures, exactly that future becomes possible. Of course, data center REITs and infrastructure funds already exist, but those provide indirect exposure to real estate or company valuations. Products like GPU price-tracking ETFs, yield products linked to data center cash flows, or tokenized compute exposure would provide pure exposure to compute prices themselves. Institutions will be the main customers for now, but if the day comes when anyone can bet on the 21st century's most important commodity - that would be quite an interesting world, wouldn't it?

]]>moyed@newsletter.paragraph.com (Moyed)ornngpu<![CDATA[Things I read Last Week #4]]>https://paragraph.com/@moyed/things-i-read-last-week-4 bltzHvPO4EIVZvYVMkphMon, 26 Jan 2026 01:05:00 GMT1. WalletConnect Pay x Ingenico: Game Changer or Marketing Hype?

WalletConnect announced a partnership with Ingenico, claiming that stablecoin payments will be available on over 40 million POS terminals worldwide. Ingenico is a payment terminal manufacturer used by major global retailers including Walmart, Costco, Shell, and BP. The service will be available to PSPs and acquirers starting January 2026, with merchant activations expected to roll out across Europe in Q1-Q2.

WalletConnect Pay's core value proposition is clear: dramatically reduce the 2-3% card processing fees, settle transactions in seconds, and provide a familiar QR-based checkout experience. For large retailers processing $100 billion in annual payment volume, fee savings alone could amount to billions of dollars.

However, the counterarguments are substantial. The key issue is distribution structure. Ingenico terminals are Android-based with an app store, but merchants don't browse app stores on their terminals. They treat terminals as static utilities provided by their banks. Ingenico doesn't sell terminals directly to coffee shops—they sell to acquirers and ISOs (Independent Sales Organizations) like Worldpay, Fiserv, and Chase. These intermediaries configure the devices before handing them to merchants.

The problem is incentive misalignment. Acquirers and ISOs profit from interchange fees—the cut they receive from every Visa/Mastercard swipe. "Bypassing the banks" effectively means bypassing acquirer revenue. The acquirer controlling the terminal has zero incentive to install the WalletConnect app. In fact, they have negative incentive—every stablecoin transaction represents lost revenue compared to a card swipe.

There are counter-counterarguments to this as well. While acquirers lose interchange, the decision pressure comes from merchants. If merchants can save 2-3% on fees, they'll push for it, and the ecosystem adapts accordingly. Closed-loop tokens like gift cards don't carry the onerous fees of credit cards, and stablecoins as programmable payment instruments can leverage these existing token types. Post-GENIUS Act in the US, acquirers are already exploring how to incorporate stablecoin payments into their reconciliation processes. With merchants fed up paying over $100 billion annually to intermediaries, there's movement on the antitrust litigation front toward terminal openness.

2. The Reality of Crypto Cards

Artemis Analytics' stablecoin payments report analyzes the current state of crypto cards.

The vast majority of crypto card volume settles via fiat rails. When users swipe their cards, crypto-to-fiat conversion happens before payment network settlement, making transactions indistinguishable from regular card payments to merchants. Visa's stablecoin-linked volume reached an annualized $3.5 billion by Q4 2025—only about 19% of total crypto card settlement. The remaining 81% still settles after fiat conversion.

The infrastructure stack consists of three layers:

Incentives for card-issuing products vary by category:

So why do all these products ultimately operate on Visa/Mastercard rails?

Because card networks provide value far beyond transaction routing. Fraud protection with sophisticated detection systems, dispute resolution and chargeback rights, unsecured consumer credit, interchange-funded rewards programs, purchase protections like extended warranties and travel insurance. If native stablecoin payments must build all this from scratch, replicating the network effects that card infrastructure has accumulated over decades across 150 million+ merchant locations globally is extremely difficult. Critically, a payment method that cannot offer credit faces structural headwinds in consumer adoption.

This reality makes direct stablecoin acceptance by merchants challenging. Attempts like WalletConnect Pay exist, but there's no problem they solve better than cards for the average Western consumer or merchant. Stablecoin P2P transfers grow at 5% annually, while crypto cards grow at 106%. For now, crypto cards will continue outpacing native payment adoption.

Of course, crypto cards face criticism too. One more abstraction layer means one more fee layer—spread fees, withdrawal fees, transfer fees, sometimes custody yields. Individually insignificant, but they compound. As the saying goes, "a penny saved is a penny earned"—these hidden costs accumulate. Ultimately, crypto cards are a transitional solution. They provide convenience to crypto users by leveraging existing payment rails, but value fundamentally leaks through the fiat conversion process.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Things I Read Last Week #3]]>https://paragraph.com/@moyed/things-i-read-last-week-3 V1oWiqDz9glqeK7MF1sXThu, 15 Jan 2026 14:22:00 GMT1. The Economic Limits of Prediction Markets and Cognitive Finance

Matt Liston (Augur co-founder, Gnosis' first CSO) pointed out fundamental economic structure problems with prediction markets through a Twitter thread.

In the 2024 U.S. presidential election, Polymarket predicted Trump at 60% when media and polls were showing 50/50, and it turned out to be correct. This proved that markets aggregate information better than experts. However, paradoxically, the bettors who provided billions of dollars in liquidity to generate this accurate signal were not properly compensated.

The problem lies in the value distribution structure:

Since anyone can view prices for free the moment they form, bettors are essentially subsidizing a global public good. Private data providers charge hedge funds millions because competitors can't see that data, but Polymarket prices are instantly visible to everyone.

As a result, prediction markets only work in areas where people want to gamble for entertainment (elections, sports, meme events). They don't work in areas where decision-making actually matters (probability of China invading Taiwan, FDA drug approval, supply chain collapse likelihood). Nobody bets on these topics for entertainment purposes, so there's no liquidity, and without liquidity, there are no useful prices.

Matt Liston's proposed solution is 'Cognitive Finance,' with the core idea being to make signals private.

  1. Private Markets: Market state is maintained inside a TEE (Trusted Execution Environment), with only paid subscribers receiving signals through encrypted channels. Currently Polymarket follows "bet → price formation → anyone can view for free," but Cognitive Finance would be "bet → price formation → only paid subscribers can view."

  2. Combinatorial Markets: Instead of individual markets, maintain one massive probability distribution across all outcomes. Currently "Will the Fed cut rates in March?" and "Will Q2 inflation exceed 3%?" are separate markets. In combinatorial markets, a single trade updates the entire probability distribution. Betting on Middle East war probability would automatically adjust defense stocks, shipping insurance, inflation expectations, and Fed policy expectations.

  3. Agent Ecosystem: The key is separating agent types. Trading agents and sourcing agents are separated because if sourcing agents can see prices, they can infer "information X is valuable" from price movements and front-run. With information walls, agents are only rewarded for providing genuinely new signals.

Of course, this vision has many challenges including multi-agent coordination, enforcing information walls, and cold start problems. However, if prediction markets are to become actual decision-making infrastructure beyond gambling, an economic structure that captures the value of signals seems essential, so the direction itself appears valid.

2. The Return of Gusto

Apewood argues in this post that the 'attention-as-currency economy' has been exhausted.

Looking at history, there were four stages:

The reason 'performative' and 'slop' became buzzwords in 2025 isn't because performance is new, but because the limits of the game have been reached. Previously, only insiders knew the formula, but now everyone knows the hooks, templates, and formats. When everyone optimizes the same way, the edge disappears.

Apewood proposes 'Gusto,' a concept from William Hazlitt, as the solution. If Literalism is following templates exactly (output-focused), Gusto is "the power or passion defining any object" (input-focused).

Specifically comparing:

The core test is simple: Would you still do it if nobody was watching? If yes, that's Gusto. YouTubers uploading 40-minute deep dives when the algorithm rewards 8-minute videos, engineers refactoring working code because it's messy, people cooking elaborately alone without Instagram—all of these qualify.

Meanwhile, the advice "just be authentic" is a trap. The word 'Person' itself derives from persona (mask), and all self-expression is inherently performance. Chasing authenticity can itself become another optimization. The real question isn't "Am I being authentic?" but "Does this performance contribute to my growth, or is it just capturing attention?"

3. Helium's Buyback Strategy Confusion

Dylan Bane analyzed Helium's buyback strategy changes in a Twitter thread.

Helium's buyback timeline:

Numbers:

Helium did buybacks for two main reasons: 1) Regulatory—unlike stocks, token holders have no economic rights, so buybacks are the only legal method of "returning value," and 2) Hyperliquid's success ($1B+ annual buybacks) made buybacks a hot meta.

However, Helium's buybacks ultimately failed. Helium's $21M revenue is too small compared to Hyperliquid's $1B+, and in a bear market, selling pressure overwhelmed the buyback effect.

Dylan Bane's core argument is that buybacks don't make sense for early-stage DePIN. In TradFi terms, early startups reinvest everything into growth, and only mature companies do buybacks/dividends. Amazon's first buyback came in 2022, 25 years after IPO. Helium is #1 in DePIN revenue but only at $21M annually, which is Series B/C traction from a traditional startup perspective, not mature company territory.

Ultimately, many problems stem from the fact that token holders, unlike shareholders, have few ways to share in a company's growth upside. Buybacks briefly gained attention as a forced solution, but considering that no startup outside crypto does buybacks, it's difficult to expect buybacks to serve as anything more than a spark for short-term price increases.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Block 2.0 Era]]>https://paragraph.com/@moyed/block 8P8StlmFahsvulbwlmIeSat, 03 Jan 2026 05:57:46 GMTA Decisive Moment for Block

As fintech and crypto continue to converge with fewer and fewer clear boundaries, if one were asked to name the fintech company that was most crypto-friendly just a few years ago, Block, Inc. (formerly Square) would likely have been the obvious answer.

Block operates a wide range of products and services, including Square, Cash App, Afterpay, TIDAL, and Bitkey. Yet when looking at recent stock performance alone, Block has noticeably underperformed many of its peers.

There are multiple reasons behind this trend, but perhaps the most fundamental issue is that Block has become a somewhat “ambiguous company” that is difficult to define in a single sentence.

As its service portfolio expanded too broadly, it became increasingly hard for investors to clearly understand Block’s core business and growth narrative. The relatively expensive acquisition of Afterpay, along with the Hindenburg Research short report in 2023, also weighed heavily on market perception.

Against this backdrop, Block reached a point where it needed more than incremental product updates. What it required was a clear structural direction and a long-term growth roadmap. Investor Day 2025, held in San Francisco on November 19, 2025, was Block’s response to that need. The event was meaningful not only for clarifying Block’s own strategic direction, but also as a signal of how the broader fintech industry may be re-shaping itself. This article highlights several of the most notable takeaways from the announcement.

A Brief Overview of Block’s Businesses

Before diving into the insights from Investor Day, it is helpful to briefly review Block’s major business lines as they stood prior to the event.

  1. Cash App: A Consumer Finance Super App

Cash App began as a peer-to-peer payments application, but rapidly expanded to include debit cards, deposits, direct deposit for payroll, stock and Bitcoin investing, BNPL, and short-term lending through Borrow. Particularly among younger users in the United States, Cash App has achieved high engagement and has become a central interface for everyday financial activity rather than just a payments tool. It remains the primary growth engine for Block in terms of both user growth and transaction volume.

  1. Square: An Operating Platform for Small Businesses

Square represents the core of Block’s B2B business. It provides payments and operational infrastructure for small and medium-sized businesses, spanning POS hardware, payment processing, inventory management, employee management, payroll, marketing, and Square Loans. Square is less a standalone payments solution and more a comprehensive operating system for offline commerce.

  1. Afterpay: A BNPL Network

Afterpay, acquired by Block in 2022, is a BNPL service that offers installment payment options across online and offline commerce. While Afterpay established strong brand recognition on its own, there has been ongoing skepticism about whether its acquisition meaningfully strengthened synergies with Cash App and Square.

  1. Bitkey and Proto: The Bitcoin Division

Unlike most fintech companies, Block has consistently treated Bitcoin not merely as an investment product, but as long-term infrastructure. This philosophy is most clearly reflected in Bitkey and Proto. Bitkey is a self-custody Bitcoin wallet designed to allow users to manage their own keys without relying on exchanges. Proto, on the other hand, encompasses Block’s Bitcoin mining infrastructure, including custom-designed ASIC chips, mining hardware, and management software.

Insight #1 The Only Full-Stack Fintech Company

Hardware

While this theme did not appear for the first time at Investor Day 2025, Jack Dorsey emphasized more clearly than before that Block’s core differentiation lies in being a full-stack company capable of designing everything from hardware to software in-house.

Unlike most fintech firms that treat hardware lightly or outsource it entirely, Block has directly designed and shipped more than 30 million hardware devices since the Square era, including POS terminals and card readers. In doing so, Block has built long-standing relationships across its entire supply chain, from semiconductor foundries to manufacturing partners. This creates a structural moat that is difficult for other fintech companies to replicate quickly.

This hardware-centric philosophy carries over directly into Proto.

Proto can be seen as the application of Block’s accumulated hardware DNA to Bitcoin mining infrastructure. Traditional mining rigs are often disposable, requiring full replacement after a certain lifecycle. Proto instead adopts a modular design, where only the hashboards need to be upgraded while the rest of the system remains intact. This approach closely resembles a razor-and-razorblade model.

Critically, Proto’s design and business model would be nearly impossible without Block’s existing hardware manufacturing capabilities. Very few companies can control ASIC design, mechanical engineering, manufacturing partnerships, and long-term upgrade cycles under one roof. Proto further emphasizes that both design and assembly take place entirely within the United States, offering strategic advantages in supply chain resilience and geopolitical risk compared to the heavily region-concentrated mining hardware market.

Ultimately, Proto is less a new revenue line and more a unique manifestation of Block’s long-standing obsession with Bitcoin and hardware. A fintech company building Bitcoin mining equipment is already unusual. Re-architecting that equipment as long-term infrastructure with integrated software and upgrade cycles is an approach that only Block could realistically pursue.

Insight #2 An Intense Focus on Cross-Service Network Effects

Neighborhoods

One of the most striking shifts at Investor Day 2025 was Block’s clear move away from optimizing each business independently, toward intentionally creating network effects across its different services.

The most important change lies in how Block now positions the relationship between Square and Cash App. Historically, Square served merchants while Cash App served consumers, with little practical interaction beyond corporate affiliation. At Investor Day, Block repeatedly signaled its intent to break down that separation.

At the center of this effort is Neighborhoods. The problem it addresses is straightforward. Small businesses are increasingly dependent on advertising platforms for customer acquisition, while consumers have fewer organic touchpoints with local merchants.

Block’s solution through Neighborhoods is equally clear.

The key insight is that Neighborhoods is not merely a store discovery feature, but a behavior-driven network layer. Visit frequency, payment history, and spending patterns automatically inform recommendations and rewards. Merchants can drive repeat visits and loyalty without manual configuration. This is only possible because Square’s payment data and Cash App’s consumer data are unified within a single system.

Other initiatives follow the same logic, such as encouraging Square Payroll employees to receive wages via Cash App, and gradually absorbing Afterpay into Cash App rather than maintaining it as a standalone BNPL brand. All are deliberate attempts to engineer network effects across Block’s ecosystem.

Insight #3 The Role of AI Agents and Blockchain

Moneybot

Another major theme is the role of AI agents and blockchain, particularly Bitcoin, as foundational fintech components. Block’s vision positions AI agents as the new financial frontend, with blockchain operating as the backend settlement layer.

On the frontend, Block introduced Moneybot and Managerbot.

In both cases, the traditional app interface based on buttons and manual configuration becomes increasingly unnecessary. Rather than asking users what to do next, agents understand context and proactively suggest or execute actions.

At the same time, Block is re-architecting the backend of payments and settlement. According to Investor Day disclosures, Cash App users can now send money denominated in USD while settlement occurs over the Bitcoin network. From the user’s perspective, everything remains USD-native. The change is entirely behind the scenes, in how money moves and settles.

Block’s endgame is a system where AI agents manage financial activity on behalf of users, while an open settlement network minimizes dependence on traditional banks.

If Banking-as-a-Service abstracted bank APIs, Block aims to abstract the bank itself.

Insight #4 Underwriting and Lending as the Core Business

Perhaps the clearest takeaway from Investor Day 2025 is that lending is no longer a peripheral feature for Block, it is the company’s core business and primary revenue engine.

Across Square Loans, Cash App Borrow, and Afterpay BNPL, Block originates over $50 billion in loans annually, and the most emphasized metrics at Investor Day were concentrated in this area.

Block focuses less on loan size and more on capital efficiency. Loan margins are approximately 2 percent, yet return on invested capital reaches the high-20 percent range. This is driven by ultra-short-duration, high-velocity lending. Cash App Borrow loans average roughly 21 days in duration, allowing capital to turn over more than 15 times per year. While individual loan profitability is modest, velocity compensates.

With warehouse facilities or deposit-based funding, Block indicated that return on equity could exceed 100 percent.

This performance is enabled by superior underwriting data.

Cash App Borrow evaluates risk using P2P payment flows, payroll deposits, card usage, balances, and in-app behavior rather than relying primarily on traditional credit scores. Approximately 70 percent of Borrow users fall into subprime categories by FICO standards, yet long-term loss rates remain below 3 percent. This reflects not risk avoidance, but more precise risk measurement and control.

Square Loans exhibits similar structural advantages. Because Square directly controls merchant revenue flows, repayment is automatically deducted from sales. Lending becomes embedded within business operations rather than a standalone financial product.

Ultimately, Block’s lending business sits atop its broader network strategy and AI automation. As Square and Cash App become more interconnected, data quality improves. As Moneybot and Managerbot automate user behavior, predictability increases. Lending is the layer where all of Block’s strategic threads converge into tangible profit.

Conclusion

At its core, Block’s message at Investor Day 2025 was relatively simple. Block no longer sees itself as a company operating multiple payment apps or financial features. It is building a single underlying structure on which all of its businesses rest.

For the first time, Block’s disparate businesses felt coherently woven into a single narrative.

]]>moyed@newsletter.paragraph.com (Moyed)<![CDATA[Dopamine Finance and the Zeitgeist ]]>https://paragraph.com/@moyed/dopamine-finance jfxn2iQfxMJ5Ht14iKjzSat, 27 Dec 2025 07:20:01 GMT_The ideas in this article are largely drawn from the work of others, including_ @0xsmac , @sytaylor , @fintechjunkie , and many more. Many thanks to these thinkers for the work they’ve shared. This piece is simply an attempt to build on it.

What Problems Will the Next Generation of Fintech Solve?

The difference between an ordinary company and a great one lies not merely in execution or outcomes, but in how precisely it captures the problems of its era and, more importantly, whether it can redefine those problems into a new Zeitgeist that pulls society forward.

From that perspective, one of the clearest Zeitgeists today is financial nihilism among younger generations, accompanied by a growing dependence on dopamine-driven financial behavior. This essay synthesizes prior research and observation to examine the structural origins of this phenomenon, the behavioral patterns it produces, and the directions next-generation fintech must explore to meaningfully address it.

The Three Structural Forces Behind Financial Nihilism

The Collapse of Housing Affordability

The first structural driver of financial nihilism is the collapse of housing affordability. This dynamic is empirically demonstrated in a paper authored by Korean researchers.

The core logic is simple.

  1. As housing prices rise while incomes stagnate, young adults give up on homeownership.

  2. The belief that “I will never be able to buy a home” then spills over into consumption, labor, and investment behavior.

Several findings from the paper are particularly revealing.

The Breakdown of Trust in Time

The second driver is the collapse of trust in time itself. This phenomenon is sharply articulated in @0xsmac’s essay. ' The children yearn for the fiat mines '.

The mechanism unfolds as follows.

  1. Younger generations no longer perceive waiting and compounding as rewards, but as losses.

  2. Long-term planning is not delayed but structurally abandoned.

Several data points highlighted in the essay stand out.

An Addiction-Friendly Environment

While not a primary cause, younger generations are also disproportionately exposed to addiction-friendly environments.

Neuroscience research([ 1], [ 2]) examining the relationship between dopamine D2 receptor availability and economic decision-making provides important context.

D2 receptors function as a dopamine brake, suppressing excessive reward-seeking and incorporating loss signals into decisions. Lower D2 availability, often associated with sustained exposure to high-frequency rewards, correlates with the following traits.

The Phenomenon: Dopamine Finance

These forces converge into a single observable outcome: Dopamine Finance. This manifests most clearly in consumption and investment.

Leveraged Consumption

The clearest example of leveraged consumption is Buy Now, Pay Later (BNPL). Over the past few years, BNPL has transitioned from a trend into a habit.

Alongside BNPL, probability-based consumption has emerged. Services like @coverd convert fixed spending into probabilistic reward structures, effectively gamifying consumption.

Leveraged Investment

On the investment side, younger generations increasingly gravitate toward non-traditional, high-volatility assets.

The Zeitgeist: Overcoming Financial Nihilism

If these trends persist, a large share of this generation will reach retirement without adequate preparation, imposing significant social costs. At its core, the problem is simple: young people have opted out of the game itself.

@sytaylor outlines following possible solution paths in his remakrable article, Financial nihilism has a cure.

The first category focuses on helping young people survive within the existing game.

The second category lowers essential costs structurally by expanding supply and redesigning that expansion as an investable opportunity.

Finally, it may be necessary to accept that Dopamine Finance trend itself cannot be eliminated.

A Trojan horse strategy may be the only viable approach: attract users through speculation, then quietly convert engagement into long-term financial stability through embedded nudges. If a product markets itself as a moral correction to financial nihilism, it may never be adopted. Usage must come first.

Conclusion

What makes financial nihilism unsettling is its rationality. In a world where housing is unattainable, time cannot be trusted, and waiting feels punitive, gravitating toward high-frequency, high-stimulus finance is not irrational. It is logical. The tragedy is that while this path offers short-term excitement, it almost inevitably converges on long-term financial ruin.

Fintech aimed at younger generations is abundant. What is scarce are financial structures that credibly signal that effort and time still matter. Without that signal, no amount of UX innovation will address the problem at its root.

]]>moyed@newsletter.paragraph.com (Moyed)