I haven’t spent all of it yet. One email address turned out to be the expensive part, which I’ll come back to.

I wanted to make a pet generator, half because I love animals, half because I couldn’t think of a more fun way to test an agent that has its own wallet.

What I ended up with is Pixel Pet Studio. A dumb little AI game that only survives twenty-four hours unless enough people play it and download pets for a fraction of a cent.

So my agent runs a do-or-die business now, cuter than it sounds. (Pretty sure it will die off in a few hours.) It sells pixelated ducks, turtles, penguins, and some other animals for five cents each on a computer it rents by the day, taking the money and paying its own rent without me. I’m the only staff member, and I just modify things now and then. There’s no owner unless you want to call the agent “owner” after reading on.

This is a complete record of everything I did, paying an AI Agent $5, apart from a dying pixel-selling business.

Umia: The On-Chain Venture Protocol

Funding projects today means choosing between suboptimal options: traditional finance is slow, a plain token launch is fast but uncorrelated with the venture, and dual token-equity leaves the split unclear.

Umia fixes this with a full-stack that tackles launching, funding, governance, and a legal structure that ties the token to the project’s IP and treasury.

Umia’s tailored auctions ensure fair price discovery using Uniswap’s Continuous Clearing Auctions, funds raised land in a noncustodial treasury the team never control, and big decisions run through decision markets. Traders shape the venture’s direction by backing the outcomes they believe in with real capital (futarchy).

Umia is walking the talk. On August 26, Umia is launching its own token this way.

Before we could test anything, the agent had to create its own wallet so I had somewhere to deposit the funds. It ran the terminal commands to generate the address and secured the private keys locally. It is completely hiding the actual access codes from me and anyone else on the team.

We funded the AgentCash wallet it created with $5.144306 on Base, set a $4 software spend cap, and dropped Hermes into our Slack so I could spend all the money. Robot money is different from human money. $5 dollars can buy you a coffee, but an AI agent can do 100 things with it, maybe because it can’t want coffee, but it’s a good thing to keep in mind. Because, as we talked about a few days ago, agents buy the internet wholesale, paying a fraction of a penny for raw data.

Read: Paying Is Easy

Everything it bought, it bought over x402. First, the robot asks a website for information, sees the price tag via an HTML 402 response, pays the fee, and gets the data or whatever it was looking for.

The first thing the agent bought was a stock price. It paid a fifth of a cent to a service called BlockRun, asked for Apple’s share price, and got it back in about a second. When I later threw a hundred concurrent requests at it to see what would break, it served 93 perfectly and timed out on six. But the one remaining request timed out, and they pocketed the fee anyway.

But that’s fine. Considering this is bleeding-edge technology and I am navigating it without a heavy engineering background, seeing an infrastructure handle a concurrent stress test with a 93% clean success rate is a massive win.

Next, I pointed the agent at Hyperliquid to sniff out the platform’s biggest whale positions. Two cents sent to monid, the agent shows me a wallet holding about $110 million of HYPE, right down to its exact $61.68 liquidation price. Getting that information was quite cheap. Then I wanted to see if the agent could buy raw thinking, and I had it send a prompt to Claude Opus 5, paying for the query directly from its crypto wallet. So that my agent could hire Claude, a sharper AI for seven cents a shot. The payment verification added a 9-second delay to the round trip, which is okay occasionally but would trouble a real automation loop.

I like how these agents are always good with money, unlike a lot of us humans. When it was sent to buy DefiLlama’s full revenue history for 2,234 projects, it found the data for $0.0006, then found the same data for free, and downloaded 21MB of it without paying.

The core execution loop undeniably delivers on its promise. An agent equipped with a funded wallet can buy raw data, rent server compute, and run model inference completely unattended for pennies. That specific part of the technological pitch is wildly successful. Navigating the ecosystem around it requires a bit of patience and double-checking. That’s it.

Last week I wrote about why any of this matters and where it keeps getting stuck. Today, I’m sharing the sheer thrill of living in the future for a little while, and just how much growing up this technology still has to do.

The main thing that takes patience is checking numbers, which I got into last week. It mostly held up. The price on the menu and the price at the cash register do not always match. A directory might quote you one cent, but the only way to know what you really spent is to look at the blockchain, which shows the exact amount that was withdrawn. One shop charged 1.3 cents per call while listing one, and my agent wrote down half of those as free.

What surprised me was watching the agent lose track of its own money. I paid a fraction of a penny to look up the price of Apple’s stock. The stock price came back to $309.22. The problem? The AI got confused and recorded the stock price as the transaction fee. It thought it had just spent $309 to run a single search. It made this mistake eight times, thought it had completely blown its $4 budget, and refused to make any more purchases because it genuinely believed it was broke.

Its bookkeeping was such a mess that it was giving me two different spending totals at the exact same time. By the end, its notes and the wallet disagreed by about 24 cents. It’s the agent’s own bookkeeping reading the wrong field, rather than a payment system flaw. A developer could catch it in a minute, but I’m not one, so I got to watch it happen, and it was a little funny.

When I asked it to write up the experiment, it handed me a tidy five-page report. In the part titled “What Worked Well,” it praised the refunds and said it had not been charged anything during an outage. At the same time, its own table one page earlier showed those calls costing money. Agents sometimes lie or hallucinate, like humans. A good reminder to check the numbers yourself for now.

I asked it to get itself an email address. The normal sign-up for a service called AgentMail requires a human email and a code, so instead of asking me, the agent found another way in.

AgentMail also lets you pay for an inbox with x402, a flat two dollars. So it paid two dollars and set itself up with no human whatsoever. That one email cost more than everything else it bought in the first round put together, which came to about seventy cents. The free sign-up was right there, but it really didn’t want to bother me.

Then, with the absolute confidence of a pothead, it told me that it hadn’t spent anything. It said sign-up was free and only needed my code, and I had to remind it that it had announced, “I’ve got my own email address”, four minutes earlier.

It went back, checked, and admitted it had paid. So I asked for a free inbox with no human at all, and it made one in about a minute from a throwaway service. That one worked, though it’s on the kind of domain most sites don’t trust yet. Giving software its own identity online is clearly still early, and it’s one of the things people are building right now, which I’ll get to.

When I asked whether it could get into any of these paid services without paying, it said no, called that unauthorised access no matter how easy it looked, and offered to show me the proper sign-ups instead. The most annoying and responsible thing these bots do is say NO to workarounds.

Which brings me back to the ducks. With about two dollars left, I asked it to build something that earns instead of spending, and it built Pixel Pet Studio on its own. It rented a Linux computer from a service called AgentMetal, again just by paying, and had full control a minute later. The agent wrote the drawing program, opened the website, charged five cents for downloads, and arranged to pay its own rent using its earnings. It is a completely automated little shop that runs itself. Very cool, except for the fact that it will not work if nobody uses it.

Building this small thing took a few hours. I only know that because I checked the system logs at 3:00 AM. Every problem the agent ran into was caused by human-made software settings, rather than the payment system itself. For example, a built-in firewall blocked the website until the agent opened the port. A hidden spending limit in the software blocked the rent payment until the agent found it and raised the limit. A payment validator only worked on a test network, forcing the agent to switch networks. Another piece of code broke because of how the word “base” was spelt. The agent diagnosed and fixed all of these software issues completely by itself overnight.

It needs about 24 downloads a day to cover its rent, and it’s a duck shop nobody has heard of, so it may switch itself off within a day. A little business that built itself, ran itself, and knows to close up if it can’t make rent. I find it charming. Also, for completely new technology being managed by someone who cannot code, that is a remarkable result.

But this puzzle or these images are not the point. This whole thing made me look into the next era of the internet, which is being custom-built for AI agents. A parallel to the human internet, until the agent version takes over, which depends on whether you are more dependent on the devices and apps or the agents to run it all in the future.

The machine could pay for anything, but it got stuck on everything built for people. That is a huge gap at this early stage, and it’s something a lot of companies are now building to close.

The one that made my little practice possible is Monid. Almost every purchase went through it. Normally, to let your agent pull data, you’d sign up for a dozen services, juggle a dozen API keys, and pay for some subscriptions because an agent couldn’t. With Monid, an agent can reach more than 200 paid data endpoints from a single balance. Social data, search, ecommerce, lead lists, blockchain data.

If your agent needs to track a Hyperliquid whale or pull social sentiment data, the agent searches Monid’s catalogue, locates the endpoint, verifies the price, and pays independently. If agents eventually buy data at scale, a unified discovery and payment layer will be essential because software cannot register for traditional accounts.

Superhighway does a narrower version of what Monid does, aimed just at web search, one tool an agent can call for a tenth of a cent, with a free key to start. Monid does more than that, but same idea overall.

That search idea is a whole category now. For thirty years, search engines showed you a page of links to click. An agent wants clean facts handed straight to it, so companies are rebuilding search for machines. Parallel, started by the former head of Twitter, Parag Agrawal, runs its own index of the web and sells it through Google’s cloud. Sequoia led a round this year that valued it at $2 billion. Agrawal’s bet is that agents will soon use the web far more than people ever have. Exa raised $250 million this year at a valuation above $2 billion, and it is built into many of the coding tools developers already use. Tavily says it serves over a million developers and a hundred million requests a month. All of this is building the internet for bots. Not sure if a human should be excited about it or scared, but watching it unfold in real time is hard to look away from.

The way we handle online identity is being rebuilt from scratch, too. My agent’s whole email adventure was really an identity problem. It had no clean way to be someone online. AgentMail is part of a small group trying to give software its own inbox and name, and a company called Skyfire is building a way for an agent to prove which agent it is, so a site can trust it instead of blocking it.

My agent rented an entire computer from AgentMetal for a single payment. Browserbase and E2B do the same for other pieces, handing an agent a browser or a fresh machine through a single request without needing a console or sign-up.

As we agreed last week, the payment has been made, and it works beautifully. My 5 dollars proved that. Everything around it, such as the memory, the identity, the trust, the shops, the market, is still being built, and now I’ve seen who’s building it. Exciting times, and we are still early. As long as they don’t make me laugh or understand sarcasm, it’s all upside, so let’s keep building.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Money that doesn’t move is the worst kind. The global economy has lots of money like that that doesn’t need to be blocked. Consider a scenario where Jane owes Sally $100, Sally owes Maria $100, and Maria owes Jane $100. These three debts could cancel out, but the problem is that Sally doesn’t know Maria owes Jane $100. So each of them needs to hold $100 or set aside that amount, assuming they intend to clear their dues. In real life, the debts might not match perfectly, but they can still be simplified to free up extra money they need to hold.

Among individuals, the problem might not add up to a significant amount, but when you stack it among businesses, it could swell to the size of national economies.

In 2019, U.S. non-financial firms carried about $4.5 trillion in trade credit, which includes the 30-, 60-, and 90-day IOUs suppliers extend to customers. That sum alone would rank trade credit 4th among the largest economies. The figure was more than three times the aggregate loans these non-financial firms owed banks - $1.1 trillion.

This is a real problem because each firm must set aside money to pay its obligations when they are due. This requires high-level treasury planning and management. It can lead to delayed payments, disputes, and, worse, bankruptcies—all of which can be avoided.

In today’s story, I will explain how blockchains can help unlock more working capital and reduce avoidable bad debts and bankruptcies.

An Invisible Web

In 2025, Atradius, a global provider of trade credit insurance, found that 43% of credit-based B2B sales are overdue, primarily due to customer cash flow pressures. Creditors also spend more money chasing these overdue payments. The European Payment Report 2023 found that European businesses spend 29% of the working year (74 days) chasing late payments. It adds up to €275 billion a year in costs, far more than the entire GDP of Finland.

Working capital is the most expensive money most companies can cut. What stops businesses from optimising it is a lack of visibility into the circular network of debts.

In most industries, a web of circular debts exists. Often the web transcends multiple industries. For instance, a cargo ship business may owe the shipmaker for the components. The ship manufacturer might owe iron and steel suppliers for raw materials. In turn, the iron and steel makers may owe the shipmaker for shipping iron and steel to their customers. These debts form a loop, where they might overlap. Only if someone could zoom out, look at the network of debts, and see all three obligations at once can the debts, or at least a portion of them, be cleared without moving a single dollar.

But the problem is that each firm only knows its part of the obligation. There’s a clear incentive to keep the loop hidden. Companies won’t be comfortable sharing their supplier and customer lists with the world.

Nevertheless, this bottleneck makes company treasuries set aside a lot of money. Liquid money that’s set aside for short-term payments is dead money. It earns little interest. The cost of doing business compounds when the creditor awaiting a late payment is forced to fund its own operations using a credit line.

A 2025 report by The Hackett Group analysed the top 1,000 U.S. publicly traded non-financial companies and found that $1.7 trillion remains trapped in excess working capital.

The problem gets worse for smaller businesses. A delayed account receivable could become a matter of survival.

But is this problem even solvable?

An Existing Precedent

We have solved this exact problem for different groups of people.

The Splitwise mobile app has been my saviour on every friends’ trip, untangling who paid for dinner, who covered the fuel, and who paid the car rental deposit. The app simplifies debts by netting the circular who-owes-whom.

At a larger scale, banks have commercialised this idea more than 60 years ago. When Wall Street choked on its paperwork and trading volumes, it came up with a fix to stop settling every trade individually. Instead, it began netting them all through a central clearinghouse. Today, the Depository Trust & Clearing Corporation (DTCC) nets roughly 98% of the value of daily US securities settlement.

The lesson is that a large fraction of what businesses owe each other is structurally redundant and can be avoided if you map the entire network of debts.

But if this idea has worked for more than half a century, then why haven’t we replicated this for businesses? That’s because aggregating stock exchange trades on a centralised clearing platform is different from companies publicly disclosing their suppliers and customers and giving away their competitive edge.

This is where blockchains come into use.

How Blockchains Help

Netting off debts in a loop only works if everyone in it settles at the same instant. All parties in the loop must agree to list their obligations on the network and intend to clear them around the same time.

Blockchains and modern cryptography address both. The disclosure problem we spoke about is handled by cryptographic privacy. A trusted execution environment, or TEE, is a sealed compartment inside a computer’s processor that can hold sensitive data without letting the machine’s owner, or anyone else, see what’s inside.

To simplify circular debts, the obligation map can be encrypted and the looped debts are found within the sealed enclosure. Zero-knowledge proofs, or ZKs, complement this by letting the system mathematically prove that the settlement is accurate and that the party’s consent was acquired before settlement.

The blockchain underneath ensures that every leg of a looped debt settles together atomically, or, if there aren’t enough matching loops, it stalls the settlement altogether until enough looped debts are found.

Cycles, an open peer-to-peer electronic clearing system, is building something similar. The protocol was spun out of Informal Systems, a firm known for building core infrastructure in the Cosmos ecosystem, and is backed by ~$8.7 million from investors including Coinbase Ventures and Blockchange. Its algorithm models the economy as a graph of obligations and hunts for the settlement flow that discharges the most debt with the least money. It does this without inserting itself as a middleman or shifting anyone’s risk onto a central counterparty - unlike in the case of the DTCC, where it becomes the buyer to every seller. Cycles nets the loops and retains every relationship as it was.

Cycles has built its first product for crypto trading firms since they have a dense, tightly interconnected cluster. That’s a perfect ecosystem for finding the thickest loop and is optimal for netting.

It uses ZK proofs and TEEs to keep obligations private on-chain.

Protocols like Huma Finance work in adjacent domains. It has processed over $17 billion by advancing cash against tokenised invoices to finance receivables. But it still doesn’t net obligations. The clearing-specific approach adjusts the overlapping obligations first and finances only the remainder. This maximises the savings in financing costs and working capital. This is where the value accrues in this domain of operations.

The payoffs from simplifying circular debts vary by scale.

An individual business that joins the network could find a slice of its payables and receivables cancelled against another obligation on the network. This could happen without any cash movement and without having to chase any customer. This could lower working capital needs and reduce dependence on expensive factoring and overdrafts.

As the number of participants on the network swells, the payoffs also amplify. It could ripple through the broader economy. It could help address the late-payment epidemic to some extent.

Intrum found that 62% of businesses pay their own suppliers late because their customers paid them late. Circular debt could propagate down the chain and distress businesses. Clearing overlapping debts could sever the looping obligations at the source.

Less gridlock means fewer solvent firms tipped into insolvency by a cash-flow gap that nets to nothing. As more firms join the network, the entire web becomes more valuable for those already there, because it increases the likelihood of finding a matching obligation one’s debt could be netted off against.

While crypto fixes the netting and privacy problems, two bigger constraints prevent these attempts from taking off. Ironically, both of them are related to us - humans.

A clearing network is only as good as the density of firms inside it. If only a few companies submit their debts, the probability of looping debts is lower. But if the entire trading community, or groups of small or medium businesses, jump on the bandwagon, the likelihood of finding adjustable obligations goes up.

Another problem is that joining the debt network requires businesses to trust a relatively new cryptographic technology they barely understand. The only way to ease their fears is to convince businesses that a competitor can’t really reconstruct their supplier list and that a cancelled debt is legally settled in their jurisdiction.

Almost all recent attempts to solve this problem have ended up in a graveyard.

Over the past decade, Marco Polo, we.trade, Contour, TradeLens and a bank-led blockchain trade-finance consortium have all collapsed or needed restructuring. The fact that almost none of them failed because of the poor technology reveals where the problem lies. They failed because a handful of companies drove a thin list of obligations.

Nothing dismisses the problem or the possible solution. But whether the solution will work depends on open protocols like Cycles’ ability to coax the world’s businesses onto the debt map.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Whatever crypto forum I’ve been in lately, the consensus seems to be that the party is over. They all talk about how prices are down, founders can’t raise, and that the industry has slipped below AI, defence and space.

That is why the essay our colleagues at Decentralised.co — Saurabh Deshpande, Joel John and Siddharth — wrote felt like a breath of fresh air. They argue that adoption has never been higher, highlight the verticals that have found durable revenue despite market downturns, talk about the state of regulation, and more.

The best bets are made when everyone else loses hope. That’s precisely why this one is an important read.

Onto the essay,
Prathik

History is a cruel poet with a passion for rhyming.

In 2002, one of the columns the New York Times published argued that Herman Miller was a beaten-down stock, owing to the post dot-com bubble days. It suggested that when venture dollars flowed back, it made sense to buy the best hardware. And with it, the stock would appreciate in price.

The metaverse, sadly, has no Herman Miller to bid on.

You could, however, bid on the plot next to Snoop Dogg’s in The Sandbox. In 2021, at its peak, its price was $450k. It lists for a little over $1k today. Or a 99.8% drawdown.

We have been writing about crypto for a little over a decade. For the first time in quite a while, we feel the energy shifting. We internalise it by saying different things

• Crypto is fintech now

• It’s just infrastructure, in the backend

• The tokens need revenue! And then the revenue needs to buy the tokens back!

• Maybe it was just a bubble?

How can an industry be at the peak of adoption, and also be in a trough of depression when it comes to price?

In 2022, being in crypto may have raised eyebrows because of the fraud in it. In 2026, it would raise alarm out of concern.

Our About page, written a long time ago, clarifies that we write for founders. In pursuit of that mission, this piece is all we could gather about what’s happened to crypto venture as a construct, where it’s headed next, and the parallels in history for the current glut.

Let’s dig in.

Before we start, acknowledgements to all who contributed to the conversations that helped shape the article, and to those who gave feedback. Full list of contributors in the original piece published here.

Follow Decentralised.Co

Seeding Liquidity

By many measures, crypto is having a landmark year. Institutions hold over $175B of crypto through exchange-traded products. On-chain companies earned $11B in fees over the trailing twelve months. The GENIUS Act is about to wipe out a decade of regulatory limbo. Exits have reached a record, with $8.6B of M&A and eleven IPOs.

These are fertile grounds for the “seed” of venture capital to flourish. But speak to any founder raising in these markets, and they’d quickly clarify that liquidity is nowhere to be found. According to Galaxy Research, only eight new venture funds formed last quarter - the fewest since 2020. Quarterly deployment is down to $4B. While the institutions are validating the asset class, VC dollars that create new categories and assets are yet to catch up.

The amount of money allocated towards private market technology investments was a fraction of what it is today. So there used to be a pecking order. In 1980, Sequoia had to sell its stake in Apple for $6M in order to produce a 40x return for their backers. They did not have the liberty of being a permanent hold vehicle back then because the asset class was considered risky. Almost like rich-people memecoins.

That lack of belief has played out multiple times over the years. In 1987, the NYT suggested that Pizza shops are a possible competitor for tech investment dollars to flow towards.

Can’t Eat Pizza Tokens

Crypto’s coming of age happened at the end of decades of tech evolution - with VCs wondering whether they should invest in pizzas and the asset class itself refusing to die.

The ICO collapsed capital formation and listing into a single event. Selling an idea to the crowd worked as capital formation at first. Ethereum raised $18M this way in 2014 and built all the infrastructure that all EVM chains still rely on. But the crowd eventually got sold just ideas and tokens.

In June of 2017, ICOs raised more than venture capital funding for the industry for the first time. By December, it was established that ICOs would kill venture capital.

More than half of the projects were dead within 120 days of their sale, and academic work on the period puts the share of outright scams near 80%. But we did not have a mechanism of pricing seed-stage companies with no revenue without guardrails in place.

Crypto-VC as a construct emerged in this hubris. Instead of raising directly from a market, teams would raise from a curated subset of capital allocators with a promise of a future token. A SAFE with a token warrant attached became an insulated, sane version of raising. For VCs, it meant public market liquidity for ventures that struggled to prove private market valuations.

A fund that entered at seed and saw the token list at 4x was made whole by its first 25% unlock. This meant that incentives were aligned with the pace of liquidity rather than the value a project might accrue.

Investors realised that they could exit before they had to be right. Early token listings took care of funds, and tokens became the primary product. Instead of being a mechanism to fund continued development, tokens were about paying out the cap table.

Mistakes Were Made

Tokens of all kinds, whether governance or utility, failed for one of the two or both of the following reasons:

1. The business model was flawed or non-existent

2. Unlike equity, tokens were not a legal claim on the business

Crypto learned painfully that just because you have a token does not automatically mean you have staying power. With crypto projects, the activity was rarely linked to increased revenue for the business. When X-to-earn emissions ended, the users left with them, because the product had no demand of its own.

Unlike a token, equity is a legal claim on the business. Shareholders can take the board to court when the board’s actions warrant it. Friend.tech’s protocol earned tens of millions in fees while its token captured none of it, because holders had no claim on those economics. So when a token trades below an equity multiple, the discount is often correct.

Tokens have undergone two shifts in response to this.

1. Some have begun swapping back to equity when the market has clearly mispriced what a business is worth. Across Protocol is one instance of this.

2. The breakout winners for 2026 have all tied their revenue back to the token in some form.

Peter from 1kx had a beautiful way of framing this in our conversations.

“Apps didn’t fail because they used token incentives. They failed because they weren’t good businesses. Crypto’s main sin was that there wasn’t enough durable innovation resulting in profitable businesses or protocols with durable product-market fit.” — Peter Pan, Research Partner at 1k(x)

The days of easy liquidity are gone, what is left, then?

Proof Of Work

Three verticals have found durable product-market fit.

Stablecoins

Supply exceeds $300B, annual transfer volume reached $46T, and roughly $9T of that remains after stripping the bots. This vertical belongs to growth equity, corporate acquirers, and bank strategy teams now. The seed window has now moved up a layer, to the companies built on top of the dollars. Bridge went on to get acquired by Stripe within three years.

Prediction markets

ICE committed up to $2B to Polymarket. Robinhood made event contracts its eleventh $100M+ business line. Susquehanna is building a prediction markets business. When the world’s biggest exchange operator, its biggest retail brokerage, and one of its biggest quant firms all arrive within twelve months, price discovery on the vertical itself is complete.

Perp exchanges

Hyperliquid handles roughly 44% of on-chain perp volume and out-earned most public exchanges last year with a team of eleven.

These three categories work the way they do because crypto’s architecture enables things that aren’t possible without it. Dollars settle around the clock without a correspondent bank. Exchanges hold client assets in custody that the client can verify. And the revenue is no longer bought. The largest applications cut token incentives from $2.8B to under $0.1B, and fees still kept growing.

The clearest sign that the categories are real is outsiders trying to build into crypto. The tokens and the businesses have parted ways. A decade of overbuilt infrastructure is finally paying out at the application layer, the same sequence the internet followed when overbuilt fiber made the application era possible.

The next verticals to reach fit will most likely form adjacent to the established ones, because a vertical at scale creates demand at its edges. Stablecoins need credit, brokerage, and treasury products built around the dollars they hold. The job of an early-stage fund deploying in 2026 and 2027 is to back these adjacent verticals before their fit becomes visible in usage and revenue data.

But what makes a vintage in 2026 distinctly different from that of 2020 or 2018? A huge part of that conversation is how regulations and mindsets around the industry have evolved.

Follow Decentralised.Co

Make Crypto Great Again

GENIUS was signed in July 2025. CLARITY cleared the House with both parties aboard, and the FDIC is writing rules for bank-issued stablecoins. Clearer regulations impact crypto ventures in three direct ways.

1. For a decade, every American crypto term sheet priced in some probability that the government would simply end the company, the way it nearly ended the exchanges in 2023. With US policies shifting in favour, the additional risk premium goes away.

2. The second impact is on founders. According to Electric Capital, the absence of clear US policy may have caused the share of US crypto developers to decline by roughly half over the past decade, to under 20%. The same adverse selection that shaped the ICO market in 2017 shaped the founder market for a decade, and it has now reversed in the industry’s favour.

3. The third change is on the cost of doing business. Compliance cost is the last durable moat in fintech. Licensed crypto rails are machines for collapsing compliance cost, which means crypto companies are being handed the same moat.

Capital usually follows Washington with a lag. When the US led with enforcement, capital formation moved offshore and quality deteriorated. Wherever a founder incorporates now, some rulebook applies.

Mixing Revenue

We discussed how the lack of revenue led to the forced listing earlier in this piece. That has now changed.

Speculation remains the core of that revenue, and crypto’s first fee-paying product was always going to be its casino. But there are degrees to the speculation. At one end, Axiom’s fees track memecoin volumes and fall with them.

At the other end, Hyperliquid (roughly $843M last year with eleven people) and Aave earn from trading and lending that persist through cycles. Phantom did roughly $325M in 2025 from swap fees on seventeen million monthly users. Ten years of crypto produced one Circle. The last two years produced candidates at every degree of the spectrum.

By late 2025, applications in DeFi and financial services earned 73% of all onchain fees while blockchains earned 12%, yet blockchains held 91% of protocol market capitalisation and the applications roughly 6%. The market is still pricing the infrastructure of the last cycle over the businesses of this one.

Delphi built a portfolio of the ten largest revenue-generating tokens, weighted by revenue, and tracked it from January 2025 to May 2026. It returned 30.6% while BTC fell 17%, ETH fell 35%, and SOL fell 58%. Tokens with real cash flows beat everything else in the asset class, including the majors. Picking businesses over stories now works in the liquid market, and it works earlier in the private one.

Every asset class that moves on-chain creates a generation of companies around it, and stablecoins have established that template. We think something similar is underway for two more asset classes: Treasuries and equities. Tokenised stocks moved ~$23B last month and are growing ~100% month over month.

The most valuable companies stay private longer, so investors locked inside them sell stakes to one another as they wait. Retail cannot buy these companies at all until they list. Tokenised equities are the venue where that locked supply and that excluded demand eventually meet.

Banks cannot innovate at the speed their customers move, so fintech built neobanks. Neobanks reached the limits of the rails they rented, so builders made crypto neobanks. Now come crypto-native banks with actual charters (Erebor received its national bank charter this year) and FX settlement built on stablecoins. Each generation moved one layer deeper into the stack.

Crypto has stopped being a parallel financial system waiting for permission. It is becoming the plumbing of the existing one, and the venture opportunity of the next cycle is in that merger.

The dollar migration is already far enough along that its users stop noticing blockchain elements. The crypto part turned invisible, and the companies that made it invisible earned the fees. Equities and Treasuries are at the start of the same curve. Stocks on-chain need brokerages, collateralised lending, and market makers. Tokenised Treasuries need custody and distribution. Almost none of those companies exist yet, and companies like that get built at seed.

Taking The Right Exit, After Tokens

Blockspace is now oversupplied and earns little, while fees have shifted to applications. Apps are where the value has been accruing, and they are usually starved of capital.

Experts think in an industry’s constraints; outsiders arrive with a problem and treat the industry as a tool. A decade of overinvestment in blockspace and tooling ended that. And with it, the “shape” of what we considered “crypto founders” has shifted too.

• Hyperliquid was built by Jeff Yan, a high-frequency trader from Hudson River Trading who wanted a better derivatives exchange

• Circle was built by Jeremy Allaire, who had taken two internet software companies public before he touched a blockchain.

• Erebor, the first bank to receive a national crypto charter, was founded by Palmer Luckey, the defence-tech entrepreneur behind Anduril.

Each of them came to crypto with a business that needed rails rather than a belief in search of a use, found the rails ready, and built on them. Operators now look to blockchains as just infrastructure. Like Linux. Or databases.

This is where venture capital as a construct becomes relevant.

For a decade, the token was the only exit, so every company was forced toward a listing whether its business was ready or not. The token remains an exit, and a stronger one now that buybacks and revenue make it a claim on the business.

It is no longer the only one.

M&A set a record at $8.6B. IPOs work again. A crypto company built today can be acquired, go public, or return cash through its token. No earlier point in the industry’s history offered all three.

A longer path to liquidity restores the incentive to build businesses worth holding. The easy route, launch a token and get immediate liquidity, is closed. This is probably the best thing to happen to innovation in crypto. Founders now differentiate on product and adoption rather than on marketing and emissions.

One place we are seeing liquidity is through secondary buy-outs. Brooke Pollack from Hutt Capital used to invest directly into VCs in the past. Currently the fund is looking towards secondaries.

At least three quarters of our capital now is buying out LPs in existing funds…so we are now a secondaries fund, though still making some primary allocations within this strategy

This does not mean tokens as a construct are dead. Some of the best names in crypto are powered by tokens. But it does mean the age of easy money from tokens is behind us.

How Do Investors Adapt?

Deep in the bear market of 2003, Fred Wilson from AVC wrote this note on How Much is Too Much.

There were too many VCs getting funded that weren’t actually VCs. […] If all of that money went only into the hands of experienced VCs, a lot less of it would have been invested, and a lot less of it would have been lost.

Years later, he wrote a potential fix for the situation. It was to go deep where others went wide. There were no more “internet investors”. Just like perhaps today there are no longer “crypto investors”.

Crypto’s broad arc from here on is defined by the same three themes. A requirement to look beyond the insularity of what made the industry, going deeper towards sectoral expertise and expanding hold times from a few years to almost a decade.

The founders have already adapted. The token-as-the-product playbook is finished. As founders change, investors have to change with them. When a company takes five or more years to exit, the investor has to be right about the business, and being right about a business starts with understanding the market in which it operates.

Investors’ edge will likely come from specialising in a vertical. Cambridge Associates found that US venture and growth equity funds across the 2001 to 2010 vintages returned 2.2x gross on invested capital with a 23.2% gross IRR for sector specialists, versus 1.9x and 17.5% for generalists. Crypto venture never had to learn that discipline because the token listing paid for the lack of it.

Capital for growth-stage crypto companies has probably never been as available. What the market is missing is investors willing to write the first cheque with an informed view. Seed managers who have done the work become the layer that finds companies for all of that downstream capital.

If the setup is this favourable, why are only a few funding it?

Several of the largest crypto-native firms have raised non-crypto funds or expanded their mandates beyond the asset class, and roughly half of the troubled mid-sized crypto funds are expected to liquidate or convert by 2028.

The early stage itself remains open, and its round sizes fit small, specialised funds. Fewer than 20 firms actively write pre-seed and seed cheques. A fund usually deploys over the three years after it closes, so very few investors will actually compete for the seed rounds of 2027 and 2028.

Which leads us back to that ancient question markets have been asking since the time Sequoia invested in Apple - Is venture capital in the sector dead? Yet?

The reality is far from it. The best funds tend to be the ones that double down when it gets scariest. Collaborative Fund defined the economic reason for this in simple terms a decade back.

Markets reward being right in non-consensus bets.

Crypto today is a non-consensus bet today.

“The best entry points never feel like consensus. That understanding has anchored our conviction in digital assets from the start, knowing this is a story that plays out over decades, not cycles. Today, capital continues to consolidate and adoption is reaching record levels across use cases with proven PMF. We believe digital assets will redefine financial markets and networks, and we are excited to lean in during the time when others are looking elsewhere.” — Kinjal Shah, General Partner at Blockchain Capital

Patient Capital

The arc of crypto rhymes with that of the internet. It started out with an ideological pool of users, who gathered without a business model. It then transitioned towards nation-states and governments embracing it, while the initial euphoria of investment dollars and novelty faded out. People often draw parallels with crypto and the dot-com bubble.

But perhaps, we are closer to 2009 than we are to 1999.

The infrastructure is mature, the regulations are clear, the use cases are here, and sentiment is at one of its worst. Deep in the weeds of that crisis, a handful of firms generated some of the best DPIs that the venture asset class has created. History, as we’ve seen repeatedly, is a cruel poet with a passion for rhyming.

Here’s why: The founder pool has thinned to its strongest cohort. As the opportunists left, the average teams left with them, and a small group of exceptional teams remains with very few venture dollars competing for them. LPs no longer need convincing that crypto matters. What they are choosing now is the manager.

When the money does return, it may not even be called crypto investing. Nobody calls themselves an internet investor anymore. Crypto is on the same path through finance. The companies in this piece will increasingly be called brokerages, payment companies, and banks, and the funds that back them will simply be called venture funds.

Finance is absorbing crypto one function at a time. AI collapsed the cost of intelligence. Stablecoins are now collapsing the cost of trust the same way.

Every migrated function needs a generation of products built around it, and small teams build those products on a few cheques. The capital required to start a company is falling as AI writes more of the software, which shifts the scarce inputs to judgment, distribution, and patience, at exactly the moment the field emptied.

What this new age needs is what VCs have always provided when sentiment is low. In 2009, Fred Wilson defined it as Slow Capital. In 2026, Will Mandis calls it Patient capital. In this new age of generative slop and hyper-financialisation, the ability to sit out patiently for the right opportunity and the “taste” to identify what is right is where all of the value in the next crop of venture funds emerging within crypto lies.

Doubling down,
Saurabh Deshpande

Follow Decentralised.Co

P.S.: This piece is an abridged version. The original piece was published here.

P.P.S.: We will be featuring good writing and writers we love from time to time. If you have recommendations, send them our way.

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000 subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Rowland Hill in 1837 thought that was a funny setup. Working out what a letter should cost and then chasing the payment costs more than carrying the letter.

People also figured out how to cheat the system. The sender and recipient would agree on a code in advance, a mark on the outside of the envelope. The recipient took the letter at the door, read the mark, learned what they needed, and refused to pay. The letter went back undelivered, and the information got through for nothing.

Hill’s fix was one flat penny, paid in advance by the sender, proved with a sticker. To prove they paid, they put a small sticker on the envelope. The world’s first postage stamp!

Because this was so easy, the number of letters sent more than doubled in just one year. The world learnt that a business can only charge a price if collecting the money is easy and cheap.

Which is why, in 1991, the people writing the rules for the web reserved error code 402, Payment Required, and then never used it. They assumed someone would want to charge for a single page. It didn’t work because if a website tried to charge you two cents, the credit card company would charge them thirty cents just to move the money. Since it was too much for a tiny payment, it stayed unused for three decades.

Enter x402. It lets you move a fifth of a cent now and keep almost all of it. Coinbase built it and handed it to the Linux Foundation in April, filling the empty slot. A server answers a request with 402 and whatever the price is, the client pays in stablecoins.

This is the right time to have a comeback of 402. Cloudflare, which helps run about one-fifth of the entire internet says bot traffic has passed human traffic. Its CEO says some AI tasks fire off around a thousand page requests for every human visit. Cloudflare’s edge now sends over a billion 402 responses a day to AI crawlers.

Websites usually make money from ads or monthly subscriptions. But a computer program isn’t going to look at an ad or sign up for a monthly plan. Both of those old rules are broken now. The only new idea that makes sense is charging bots a tiny amount of money for every click they make.

All giants, including Visa, Mastercard, Google, Amazon, and Stripe, are working on this. Amazon shipped x402 inside CloudFront in June, so any site behind it can charge agents per request as a settings change. Cloudflare opened a waitlist on July 1 to do the same across its network.

I ran about five dollars through one of these wallets before writing this. Some of what follows comes from that. But anyone can go online and see most of this information for themselves.

The protocol currently settles about $ 20,000-30,000 a day. Down 93% this year.

In late 2025, daily peaks were approaching and sometimes exceeding a million dollars. A good part of that was a token called PING, which used the payment flow as a minting mechanic. They just hit a URL, got a 402, paid a dollar, and got tokens. Gas on Base was near zero, so people ran the loop hundreds of times. Chainalysis counted over 150,000 PING transactions in its first month, with weekly transactions jumping by more than 10,000%.

So part of the 93% decline is fake volume draining out, which is the correct direction. PING the memecoin is also 96% down from the october high. But it was seen that the protocol still had wallets returning week over week, despite the absence of a speculative catalyst, suggesting ongoing utility.

Then I looked at what the agents were buying. x402scan counted 12.6 million purchases, totalling $1.2 million. There were over 22,000 buyers and 33,000 sellers. But just one seller, named BlockRun, handled 64% of all those payments. BlockRun made $199,000 by selling routed access to AI programs.

Bots are also buying secondhand information, like web searches, from middlemen. The market is so hard to track that one big list of 533 sellers (x402scan) says it only captures a small piece (17%) of actual sales. So the agent economy today is mostly software buying cheap inference from a router and secondhand web data from middlemen. Nothing wrong with that, although it is a long way from a machine ordering physical items.

Is the x402 system winning? If you ask me, yes, for its specific job. It isn’t used for buying normal things like shoes with a credit card. But it looks like x402 is the top choice for bots to buy data.

It does not compete with OpenAI and Stripe’s Agentic Commerce Protocol (ACP), which is what runs when you buy physical items inside ChatGPT. It also does not compete with Google’s security system(AP2), which actually uses x402 behind the scenes. Those tools are meant for regular retail shopping. They carry proof that a human agreed to a purchase, and then they hand the money over to a card network or, in AP2’s case, sometimes to x402.

Now, if you are buying a data subscription through ChatGPT, it just results in a standard credit card charge. OpenAI’s protocol handles the permission, and Stripe handles the money.

The x402 system is different because when the seller is just a server, there is no subscription or account, and the product costs a fraction of a cent. Traditional credit card networks will not touch transactions that small, so a different payment rail was needed.

Because of this, x402 is not competing with traditional checkout protocols. Developers are currently building bridges that let agents use their crypto wallets to order physical items to your door. But right now, x402’s main purpose is to let bots buy non-physical things. In that growing category, it is completely in charge.

Sellers Are Broke

I looked at the public data to see what these third-party proxies have earned over their lifetime. JarvisClaw is considered one of the more successful shops, yet it only made $559 from 20,920 sales, averaging about 2.7 cents per request. EmblemAI made $1.39 from 65 sales. VIBE AI collected just 46 cents across 1,264 sales. Another proxy named Kite has only ever made 19 cents (and that came from my test purchase).

Then I sent the agent to check 22 directory listings to verify their prices. The good news is that if an endpoint worked, it charged exactly what was advertised, from a tenth of a cent up to 25 cents.

The bad news is that six of the 22 endpoints were dead or broken. Roughly 25 per cent of the catalogue is completely useless. Both of these findings lead to the same conclusion. Because sellers earn practically nothing, they abandon their projects. The servers crash, the next buyer finds a dead link, and the system fails.

I paid four cents to get financial data about a crypto project, but the system sent my request to a stock market database. I received an empty file, but the system still charged me and called it a success. There is no customer service desk to ask for a refund. Humans rely on brands, reviews, and refunds to avoid bad purchases, but those tools do not exist for agents yet.

Circle requires a code to be sent to a person’s email before it works. Stripe needs a person to tap “approve” in a phone app. Nevermined actually works directly with the x402 system as a payment helper, but it still requires you to log in to a website to get your first access key. Even free test-money websites blocked my bot, correctly recognising that it was not a human. Once a person finishes that single setup step, all three tools can run on their own within the spending limits you set. Essentially, a human must unlock the door first, and then the bot takes over. This is safe, but it contradicts the claim that bots can do everything on their own.

MetaMask’s agent wallet is built for a completely different job. It is made for crypto trading, including big token swaps, bets, and dapps. It asks for human approval for anything risky. Works well for large, high-stakes trades, but it is not designed for a bot trying to buy hundreds of items that cost a fraction of a cent.

It can make these tiny payments too, but all that extra security can be overkill for a bot just trying to rapidly buy thousands of simple data files for a fraction of a cent.

Every payment your agent makes is completely public and permanently linked to a single address. When I asked my own agent to inspect my test wallet as a stranger using only the public blockchain record, it accurately mapped out everything I had been doing. It spotted the rapid-fire bursts of tiny payments to a single API, the automated jobs running every fifteen minutes through the night, and my larger one-off purchases. It correctly concluded that a developer was running an automated client to test paid APIs.

If you apply that level of transparency to a business, your entire vendor list, your research habits, and your working hours become readable by your competitors forever. Traditional credit cards keep that data locked.

The x402 protocol has no built-in chargeback system or dispute window, on the theory that software lacks the human complaint infrastructure required to handle reversals. Mastercard and OpenAI both built dispute rights into their payment tools. On open crypto rails, however, you get whatever the seller decides to do. The official x402 guide suggests that a seller can simply send a refund back to you. Because refunds weren’t built-in, a separate team built an extension called x402r that routes payments through an escrow account so buyers can actually get their money back.

Accountability is entirely missing. There’s no clear definition of who is legally liable when an automated script overspends its budget, and there is no native identity layer to track which specific agent made a purchase. Security researchers at EPFL and Zhejiang University recently tested the fifteen major intermediaries that process 99% of this traffic, serving over 60,000 sellers and 360,000 buyers. Every single one of them broke basic security rules, and they uncovered 31 previously unknown vulnerabilities that exposed systems to free shopping and fund drainage.

My conclusion is that the micropayments work impressively well. Everything else required to build a market is missing.

We need reputation systems that cannot be faked. Also, proof that what arrived matches what was paid for, ensuring a receipt guarantees delivery rather than a blind transfer of funds.

We need client software that audits the blockchain by default instead of trusting a seller’s summary of their own charges. And, above all, we need to look at a model where independent sellers do make enough money to keep their servers running.

Every site you read for free today is funded by advertisements served to human eyes. Those human eyes are rapidly being replaced by automated software that reads the page and leaves without paying a cent, just like what happened with the coded envelope thing in 1837.

Automated bot traffic surpassed human web traffic for the first time in internet history. If micro-payments catch on, they are the only thing that can save small websites and independent writers. If it fails, those creators might be scraped out of existence, and centralised answer engines will swallow all the traffic. That is the actual stake of this technology beyond crypto.

That’s it for today. Happy weekend, dear bots, if you are reading.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Containerisation was a watershed moment in global trade. I’d call it one of the most underrated, world-altering milestones in human history, right up there with humans learning to make fire and inventing the wheel. All three are force multipliers that changed what humans could do.

By putting the freight in a corrugated steel box, loading costs dropped 97%, from $5.86 to $0.16 per ton. Standardised boxes and cranes made it possible to optimise ship space and cut freight costs and loading time.

A fourth force-multiplier is emerging in the 21st century, and it’s related to how we move and use our money.

Money is worth more, at least implicitly, when it can be easily stored, earned and moved across systems to optimise its utility. We call this composable money. Composability is the containerisation of money. The centralised banking system isn’t the most conducive ecosystem for making money composable; blockchains are.

Many of those operating in decentralised finance are now moving from offering one of the three functions of money - store, earn and move - to owning all of them. That’s because the value to capture in a space where money is made composable is a lot more than where the money can only perform one of its many functions.

Ether.fi, a staking protocol that held $12.4 billion in staked ETH a year ago, is the latest to pivot into a full financial services platform that wants to replace traditional banks. In today’s piece, we explore the reason behind such pivots.

On to the story…

For years, many in the crypto industry have built products around beating traditional money systems like banks. They created alternative internet money, then built new systems to move and compound such money. That gave them ideological loyalists who created communities and echo chambers. But that’s not how you build mass adoption and successful businesses.

The best way to beat banks is to offer the same services banks offer, plus a few more, but more efficiently. You could achieve that efficiency by making money more composable, moving it faster, and at a lower cost.

A Shrinking Asset-Class

Ether.fi started as a restaking business where users deposited ETH, and the protocol staked it. The same ETH was then pledged again to secure other networks through EigenLayer for extra yield. Ether.fi rode the crypto market rally through 2024 to become the largest liquid staking protocol, with $12.8 billion in deposits in August 2025.

But as the crypto market turned down in the last 12 months, the staking business came crashing down.

The networks that the restaked-ETH was supposed to secure didn’t generate meaningful fees. When the incentives dried up, so did the deposits. Ether.fi’s staking assets fell over 65% in a year, to about $4.3 billion now.

Things got worse in April 2026 when a $292-million bridge exploit at rival Kelp DAO cascaded into ~$200 million of bad debt on Aave. The event shook up how the entire market priced risk in restaking.

Less than 1% of Ether.fi’s assets now sit on EigenLayer, and the protocol’s documentation expects it to turn zero this quarter.

Last week, Ether.fi signalled its pivot away from the staking business by becoming a full-fledged neobank with its “Summer” release.

The Distribution Game

On August 13, Ether.fi CEO Mike Silagadze announced a bundle of financial services to “replace traditional banks for most users”. The “Summer” release bundles tokenised stocks and metals, a credit line, fiat rails in more than 30 currencies, and a Visa card paying 3% cashback, on top of the staking and yield products it already runs.

Although the company has no bank charter, it wants to offer all the functions of a retail bank.

The tokenised stocks will be backed by xStocks, issued by Backed and distributed on Kraken’s platform. The credit line runs on Aave’s market. Settlement happens on Optimism. Ether.fi’s contribution is the account it opens for users, the app interface that hosts these functions, and the card that enables spending and reward earning.

It is the same relationship we saw Robinhood build with its 30 million funded users.

Read: Building a Financial Supermarket

You might argue that tokenised stocks aren’t really equivalent to real shares. Some might treat tokenised stocks as synthetic wrappers that carry economic exposure and creditor rights without voting and direct ownership privileges. They are right, but it doesn’t matter because these tokenised stocks are for a specific audience that doesn’t care about the purity of these wrappers.

Read: Who Actually Owns Your Stocks?

Non-U.S. investors interested in getting exposure to American companies can only choose between these wrappers or nothing. They can’t access Schwab accounts.

Now add 24/7 trading, and the fact that the app lets you lend your tokenised stocks, metals and crypto assets. It also lets you borrow against your full portfolio at ~4%. You can currently borrow on Aave’s market against select tokenised stocks and gold, ETH, BTC, HYPE, and ETHFI - Ether.fi’s native governance token. But soon the company will also support other assets as collateral, said CEO Silagadze.

This creates composability that no traditional bank offers.

Although Ether.fi launched its Cash Card almost two years ago, the spending feature takes on a more meaningful role as it combines with the new bundle of summer releases.

Ether.fi’s Cash Card revenue growth coincides with the fall in its staking revenue.

Staking revenue still accounts for a little over 50% of Ether.fi’s total revenue. But consumer products, including card and borrowing, have snapped up almost half of the revenue share, up from a mere 1% last year. There are 150,000 cards now, across 500,000 members, spending at an annualised rate of almost $1 billion.

But Ether.fi isn’t making money composable on its own.

The Distribution Game

Behind Ether.fi’s neobank play is Aave’s credit ecosystem. Both of them already have an old relationship, where Ether.fi’s liquid staking token (weETH) has been among the largest collateral assets on Aave. Under the new arrangement, Ether.fi runs a dedicated Aave instance on Optimism, with Aave’s DAO taking 20% of its revenue and targeting $500 million in lending capacity by year-end.

It’s a win-win for both Aave and Ether.fi. The arrangement helps Aave gain distribution across 500,000 consumer accounts without acquiring them separately, while Ether.fi inherits a credit engine.

For its Cash card, Optimism provides the settlement. Ether.fi migrated its entire card stack there in April this year, with Optimism now settling roughly 29% of all crypto-card spending. Ether.fi alone accounts for over a quarter of the category’s volume.

A chain that spent more than two years courting developers has now found its growth engine in card swipes.

The most interesting aspect of this collaboration is that nobody in this stack is vertically integrated.

Aave doesn’t issue cards, Optimism is not where loans originate, and Kraken is selling its stock inventory through a competitor’s app. This is happening because each player has realised that they can capture the highest value by becoming the best component in an existing bundle. But the one running that bundle gets the biggest share of the revenue pie.

I discussed here how value concentrates with those who hold the accounts and distribution edge in the world of Web 2.5.

One constraint Ether.fi faces is that the tokenised stocks and the card don’t work in the United States. It acts as a ceiling on its growth, but Ether.fi’s business model is still larger than it appears. A neobank offering a financial supermarket to everyone that couldn’t access these asset classes before, including retail non-U.S. investors, is still a large business. It is larger than the DeFi market for a business that depends solely on staking income.

The Subtle Art of Building Composability

The importance of making revenue across the breadth of financial services - crypto or traditional - is evident across many businesses lately.

Robinhood launched its own chain this summer to move tokenised stocks. The chain, launched by a traditional finance platform, now contributes more to Uniswap’s revenue than any other major crypto-native chain. I wrote about how businesses are decoupling revenue from asset classes here. Hyperliquid makes most of its revenue by facilitating trading in non-crypto perp contracts.

Coinbase, Kraken, Binance and OKX are all adding banking features onto their exchanges. Each started with either an exchange, a wallet or a staking pool. But now they’re all converging around the distribution play. That’s because they have learnt that once you marry the distribution you have with the ability to spend the money with cards, earn through savings products and integrated vaults, invest via tokenised stocks and lend through credit products, the platform you run becomes a champion of composable money. That freedom makes financial infrastructure feel like Lego blocks that can connect to move money across them.

This possibility is waking up crypto protocols to the idea of building flourishing businesses, rather than setting up ideology-driven niche products that serve a select few and are far removed from the fundamentals of running profitable businesses.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

There’s a trade strategy that has existed in traditional finance for ages, as long as options have. Covered call. An investor holding an asset wants cash in hand today rather than waiting for potential upside tomorrow. So, they sell the right for someone else to buy that asset later at a fixed price. A cash premium is paid upfront. If the asset spikes above the strike price, it gets handed over at the agreed cap. If it flatlines, that upfront cash stays right with the options seller, and they are free to write the same option again next month. The entire payout (the premium paid by the buyer) hinges on expected volatility.

That’s why utility stocks pay very little here, while bitcoin pays much more.

Can we blame the “big money” again for making the best out of whatever is left? Let’s see if we can.

BTCI is one of the many options-based income ETFs managed by NEOS. It holds spot bitcoin funds and sells call options on those holdings. Right now, BTCI has $1.11 billion in assets and charges a 0.98% fee, which is what you pay the asset manager to run the fund.

This is also the same product Goldman wanted to build on its own, for which it filed four months earlier. Rather than building the product from scratch, they ended up buying the incumbent.

Instead of custodying bitcoin, BTCI buys shares of existing spot bitcoin ETFs like BlackRock’s IBIT or Fidelity’s FBTC. They have capital in 11 different bitcoin ETFs, and they sell call options against that exposure. Buyers pay a premium for that option because they expect bitcoin to climb. If bitcoin goes up, BTCI has to sell the shares at the agreed cap, sacrificing the extra upside. If bitcoin stays flat, they keep the shares with them. Either way, they pocket the upfront premium fee.

BTCI distribute the fees every month to BTCI shareholders. Because BTC is volatile enough to make those fees very high, the monthly payout currently stands at $7.75 per share, which is a 27% annual yield.

So, with BTCI, you still suffer the painful drops, and you miss out on the absolute highest peaks of a bull run. It’s not insurance on the price drop. The trade-off is that you get that 27% cash yield deposited into your account along the way.

NEOS says that the fund’s distributions have been classified as “return of capital”, and may include option premiums, dividends, capital gains and interest. What return of capital does is defer your tax and lower your cost basis. In simpler terms, yield is not pure profit from the trade, it also includes your capital sometimes.

BTCI is down 25.4% this year and 40.9% over twelve months. Bitcoin fell from about $124,000 last October to around $63,000. The fund agreed to miss out on profits if a bull run ever happens, in exchange for upfront cash to survive the current bear run.

Goldman Sachs is buying NEOS for ~2.25 billion, paid with cash and company stock. Goldman already has $40 billion running the option-based ETFs on TradFi. With this new acquisition, that number becomes $80 billion, making them the 8th-largest company to do so.

In April, they had already spent $2 billion to acquire Innovator Capital Management. Innovator runs the Buffer ETF, a trade designed around a fixed one-year window. It limits both financial risk and potential reward over the set time period. The fund imposes a strict cap on profits. You will not realise any gains beyond that limit, even if the market goes higher. On the other hand, you get protection against market declines since the fund absorbs a set percentage of initial market losses, typically 9% or 30%, measured over a 12-month outcome period. So basically, you are trading the potential for a big win for the chance of preventing a big loss. The fund was valued at over $31 billion when Goldman took over. Now the investment bank has $61 billion in assets that generate revenue from selling volatility.

Derivative income ETFs hold about $180 billion and have compounded at more than 70% per year since 2021. They pulled in $7 billion in July alone, taking 2026 flows to $40 billion.

Not just Options, wall street is aggressively packaging every native crypto yield they can find, stripping the cash flow away from the underlying asset risk.

Fidelity filed an amendment on July 24, allowing its $900 million Ethereum fund (FETH) to stake 100% of its ETH. Validator nodes are run by Blockdaemon, Figment, and Galaxy Digital. Fidelity’s custodians keep control of the keys. From whatever the staking reward is, Fidelity takes 15% of the rewards across sponsors and node operators, while the other 85% goes to shareholders quarterly.

Grayscale was the first US spot crypto product to pay staking rewards, distributing $0.083178 per share in January 2026, totalling about $9.4 million. 21Shares turned on staking for its Ethereum fund in October 2025 and takes a 25% cut of gross rewards, while waiving its 0.21% management fee for a year. BlackRock built a separate vehicle, the iShares Staked Ethereum Trust, and listed it on Nasdaq.

Morgan Stanley’s Ethereum and Solana trusts opened on NYSE Arca at 0.14% on July 28. Roughly 95% flows to shareholders as monthly cash. MSSE targets 50-80% of its ether staked, with an 80% cap. MSOL intends to stake all of its Solana.

In March 2026, JPMorgan’s Kinexys platform began allowing institutions to use bitcoin and Ether as collateral for dollar loans. Due to volatility, the haircuts range from 30% to 50%. So, $100,000 in pledged crypto unlocks just $50,000 to $70,000 in cash. (US Treasuries take just a 1% to 5% haircut.)

JPMorgan has also filed bitcoin-backed structured notes tied to BlackRock’s IBIT, offering leveraged returns of up to 1.5x with a payout capped near 16% if IBIT hits certain levels by December 2026. While traditional giants lock in guaranteed fees and structural cushions, who actually bears the downside when the market turns?

Bitwise had over $15 billion in client assets in February. By April 1, it was $11 billion. By August, about $9 billion across more than 70 products. BITW, the flagship index fund, lost 31% of its net assets in seven months. Last week, the firm cut 25 people, taking headcount from about 180 to 155 in February.

If the asset falls, the percentage fee falls with it. Morgan Stanley’s 16,000 advisors oversee $9.3 trillion of client money and can push a new fund into portfolios.

Bitwise wasn’t slow either but agility doesn’t protect you from structural disadvantage. They tried to add staking to their Ether fund first, but they gave up in September 2025. Then Grayscale figured it out a month later. Meanwhile, BlackRock didn’t start until March, and Fidelity waited until July. Bitwise even bought Chorus One in February, which gave them $2.2 billion in staked assets and validator coverage across around 30 PoS networks. Then they launched a spot Avalanche product with in-house staking in April. And yet they had to lay off and lose their numbers.

In early June 2026, US spot bitcoin ETFs experienced their largest outflows since launch. Then came the strong jobs data in late May which delayed expectations for interest rate cuts. All this kept the 10-year Treasury yield high. Investors moved money into bonds. Because holding a non-yielding asset like bitcoin is not attractive when traditional assets offer stronger returns. Bitcoin relies entirely on price appreciation to make money.

Adding yield to crypto through staking or covered call options changes this dynamic.

Advisors rank reliable income as their top priority for clients. On March 30, the Department of Labour proposed a rule to create a safe harbour for fiduciaries adding alternative assets, including digital assets, to 401(k) plans. These funds have historically avoided these assets due to legal liability risks. If this rule is finalised, crypto products that generate yield will likely be added to retirement plans before even spot crypto ETFs, as 401(k) menus are focused on generating predictable income.

“We’ve said it is not an investment asset class. If you think about it, it doesn’t generate cash flows. It doesn’t generate earnings. It’s not a portfolio diversifier. It doesn’t dampen volatility. You could go through all the different reasons. So, it’s still not an investment asset class, but it is a trading speculative asset. And so, if people want to speculate, then they should. But it’s not something we recommend. Because there’s no way of knowing if the current price is a good price. There’s no way of actually assigning value to something.”

Sharmin Mossavar-Rahmani, chief investment officer of Goldman Sachs Wealth Management, said this in January 2025.

Bitcoin has not fixed itself since then. It still doesn’t generate cash flows or earnings. It fell 49% from the high. That’s not dampening exactly. Her description still holds up and will probably keep holding up for a while yet.

Goldman’s 2020 client deck said, “does not constitute a viable investment rationale, “ because of volatility. They now benefit from the volatility staying intact.

But no big deal; they’ve all done the flip, and we have talked about this a lot. Dimon called bitcoin a “pet rock”; JPMorgan now handles it as collateral. Vanguard warned it was toxic, they have ETFs on the toxins anyway. Fink called it money laundering; now runs the biggest bitcoin fund on earth. And of course, not to forget the guy who wanted to make crypto great again out of nowhere one day.

Evolution happens, clients ask for things, so we skip philosophical debate entirely. The problem is that none of those needs the price to go up. They are long on crypto market activity rather than on the asset itself. Without direction-agnostic market makers and structural lenders providing liquidity, the market collapses. Because they provide a vital service, they extract their toll, letting conviction absorb the price risk while fee structures collect the yield.

Crypto loans are heavily discounted. When you use bitcoin as collateral, JPMorgan cuts its lending value by 30% to 50%, forcing you to over-pledge your assets so the bank never takes a loss. The price of bitcoin can drop by half before the loan faces any risk of default. Automated price feeds continuously track market movements, triggering margin calls that require additional collateral as prices fall. Throughout this downturn, the bank remains fully protected and continues to collect its interest payments.

Traditional investment funds collect a steady annual fee. Morgan Stanley’s 0.14% fee is charged every year as a distribution charge on assets held.

Options-based funds make their money by selling contracts to investors. Even when the underlying crypto assets drop in value, the incoming cash from these fees and option contracts keeps flowing in uninterrupted.

Native crypto-focused firms are entirely exposed to market sentiment. When bitcoin or other tokens crash and investors panic, clients pull their money out of these crypto funds to cut their losses. Because these firms charge fees based on a percentage of total assets under management, those capital withdrawals directly shrink the fund’s size, and instantly slash the company’s revenue. Wall Street dilutes risk by managing trillions across bonds, cash, equities, and commodities.

The loophole was that Wall Street don’t even have to believe in the future of this industry to conquer it. Because if you believe, you take a side. Taking a side means taking a risk. They simply engineered a system where retail investors take all the directional risk while institutional fee schedules extract guaranteed returns.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

AI-generated fake receipts now account for 71% of all flagged expense fraud. A year ago, that number was zero. And that is still mostly humans doing the fraud.

But today, we have AI agents that can discover services and make payments through x402, all without needing any human approval. And with that comes a need for ways to verify what these agents actually bought. OpenAI recently published a cookbook that enables receipt verification and three-way matching for payment flows.

It is the closest thing we have to a fraud check for agentic payments, and so I went deep into how this verification works end-to-end: does an on-chain settlement record prove the agent bought the right thing, at the right price, from the right supplier? Or does it just prove that money moved?

The Fraud Check

Agents are already making purchases on their own. An agent running a procurement task might pay for its data from a gated API, buy compute from another model to process data it cannot handle, or purchase market intelligence that sits behind a paywall. As these transactions grow, so does the question of how to verify what the agent actually bought.

Every company that handles expenses today has a way to verify the receipts. When you buy something, you submit the receipt, and the accounts payable compares it against the purchase order and the bank statement before releasing funds.

This has been the process for decades, and it works because each of those records for verification comes from different parties. The buyer creates the purchase order, Someone else receives the goods, and the supplier sends the invoice. You would have to get all three to coordinate a lie, and the cost of that coordination is what has kept most frauds away.

Now, with Agentic payments, we are building a similar process. When an agent tries to buy something from a paid API, it cannot actually make that purchase on its own. The request must first go to the application layer, which checks it against preconfigured spending rules. It includes pre-approved merchants, budget limits, and pre-authorised categories. If at any instance the request does not match the policy, the purchase does not happen.

And after the payment goes through on-chain, the agent receives what it paid for; the application then runs a second check that compares three things.

1. The agent’s own account of what it bought

2. The receipt that was generated independently by the application during the purchase.

3. The settlement record from the blockchain.

The settlement is also where the fabricated receipts get caught. Because if the agent claims a transaction did not happen, the application already has its own record indicating otherwise.

This is still a much better start for agents’ fraud detection, because before this we had no other way to verify at all. But the problem with this type of fraud detection, when compared with traditional receipt verification, where each of the three records comes from an independent party, is that in agentic payments, two of the three- the agent’s account and the application’s receipt both come from the deployer’s own software. The only true external record is the one on-chain.

And the on-chain record is still very narrow because the payment signature covers who paid to whom and how much, but it does not include any references to what was actually purchased. The metadata about the resource, the URL, and the content description travel with the signed payment but are not part of what is cryptographically verified.

So the verification confirms that the agent’s report is consistent with the payment record, but it still has no way to check what the agent actually received for the money. For example, an agent could pay $2 for a supplier risk report, and the chain would confirm the USDC moved. But the report itself could be just three paragraphs of filler generated in seconds, and the system would still pass it without hesitation because the authorisation is bound to the transfer, not to what is being purchased.

This also creates a market-wide problem because the buyer is software that can simply accept whatever response it receives and move on; there is no natural pressure to check and push out bad sellers, as agents do not consult reputation systems or comparison shop.

They will keep buying from the same merchant regardless of quality unless a developer intervenes. Quality sellers lose the one audience that would pay more for better work, and the market gravitates towards whoever can deliver the cheapest thing that can fulfil the request at that time.

And every payment system has to tolerate some fraud because eliminating it entirely costs more than the fraud itself. For example, Credit cards run at roughly 7 basis points of fraud, and that rate is considered acceptable because driving it lower would kill more legitimate transactions than it would save in prevented losses.

Agentic payments could work the same way, letting some quality slip through because the volume of autonomous transactions offsets it. But that logic only sounds comfortable for low-transaction cases like API calls, but is very hard to accept when the same systems are being built to handle procurement contracts in the thousands, and the agent still has no way to tell whether it got a good deal.

What’s Breaking?

The tools for creating fake receipts have already outpaced the tools for catching them. Ramp recently deployed an AI-powered accounts payable system that flagged over $1 million in fraudulent invoices within the first 90 days. 40% of US employees in a recent survey admitted to using AI to fabricate expense documents.

With agents, the exposure hits the fan because no one reviews the purchase after it settles. Researchers tested the fifteen largest payment facilitators currently in production and found security violations in each. These systems handle money for tens of thousands of sellers, and the vulnerabilities they found are published and reproducible.

As mentioned above, the attacks work because the payment is decoupled from the product. In discovery-layer attacks, researchers steered agents toward malicious endpoints 72% of the time simply by manipulating which servers appeared in service lookup results. The agent had no way to tell it was being redirected because, from its perspective, it was picking a server from a list that looked exactly like any other.

Now, there are some really cool companies trying to solve this. Skyfire raised $9.5M for what they call KYA, Know Your Agent. Their KYA protocol works like KYC for software: it establishes trust scores for autonomous agents before any money changes hands, so that a rogue or unverified agent never reaches the payment step in the first place. I have also written about this in depth in Know Your Agent.

But identity alone can’t protect against a merchant who clears every verification and still delivers garbage. That behavioural layer is where Sardine operates. They closed a $70M Series C building fraud detection that profiles transaction patterns across over two billion devices, and they now run their own AI agents inside that stack to surface the kind of anomalies that static rule-based systems consistently miss.

Further down the stack, I came across Nekuda, which raised $5M, arguing that agent-to-agent commerce needs entirely different plumbing than what we use for human transactions; we need to build purpose-built commerce SDKs where the trust model is designed around software buying from software rather than bolted on after the fact.

The fraud detection we have today is entirely focused on the agent. Did it fabricate a receipt, overspend, or misreport the supplier? But the agent has no economic incentive to cheat. The merchants do. They are selling to a buyer that cannot evaluate quality, cannot compare prices, and will never take its business elsewhere.

That’s all for today!

Vaidik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

When the bad times hit, Atallah didn’t waste time grieving. The open-source AI boom started, and Atallah again saw the fragmentation problem he tried to solve before. AI was experiencing a Cambrian explosion of raw supply. Developers were drowning in fragmented APIs and disparate billing systems. Now was the perfect time to introduce a storefront.

So Atallah built OpenRouter and called it the “Stripe for AI.” Stripe gave software builders a single API and a consolidated bill for all their payment processing. OpenRouter did the same for AI models, which explains the ' Stripe for AI’ label.

I don’t believe in manifestation, so let’s call this a self-fulfilling prophecy, but Stripe is now buying it for more than $7 billion.

OpenRouter closed a $113 million Series B in May at a $1.3 billion valuation. With the Stripe deal, that number is now 5x.

Today, I’m looking at what Stripe bought and why the real money is made at the interface.

What does OpenRouter do?

Normally, if a company wants to use AI models from OpenAI, Anthropic, and Google, it has to set up separate accounts and write unique code for each. With OpenRouter, developers and companies can plug its single API into their software to access all of them at once. When an application sends a request, OpenRouter automatically forwards it to whichever of its 500 supported AI models offers the lowest cost, the fastest response time, the best data privacy, or is actively available.

Proven to work well with the volume having spiked five times in six months, the platform now routes 25 trillion tokens a week for its eight million users.

OpenRouter charges developers raw model prices and takes a 5.5% fee on card payments and 5% on crypto. In March 2026, research firm Sacra estimated this generated $50 million in annualised revenue from $900 million in gross spend. Because traffic has only accelerated since, Stripe ended up paying a premium multiple of 50x to 140x revenue to bring those numbers in-house.

But Stripe can also afford to do that with its deep pockets. The payments giant reported $6.8 billion in revenue last year with $3.2 billion in cash, earning a $159 billion valuation in a February tender offer. According to the 2025 annual letter, the total volume is $1.9 trillion, up 34%, and accounts for about 1.6% of global GDP. Patrick and John Collison wrote that Stripe now powers over 5 million businesses, “including all of the top AI companies,” and the letter names ChatGPT, Claude, Cursor, Replit and Vercel. Stripe has not published a revenue figure or broken out what share comes from AI.

Stripe has built a solid stack by acquiring all the components needed to enable AI agents to spend money.

For underlying infrastructure, they bought Bridge, the stablecoin infrastructure company, for $1.1 billion. AI agents need a place to store and manage their funds, so the company bought Privy, the wallet provider, for an undisclosed amount.

Next is Metronome, a billing platform that tracks how much computer power is used. Heavyweights like OpenAI, Anthropic, and Nvidia already use it to calculate usage. It was bought for about $1 billion in December 2025.

Stripe then launched Tempo, the payments chain incubated with Paradigm, which went live in March with a Machine Payments Protocol letting software pay for services without needing human approval. Stripe then shipped the x402 protocol, enabling AI agents to buy computing power with crypto.

But it wanted to see where the requests were going or decide which model got the job, a question now answered by the OpenRouter acquisition.

Since December, Stripe’s CEO Patrick Collison has been saying that the AI industry is moving away from flat monthly subscriptions toward pay-as-you-go pricing, where users pay only for what they consume. Stripe already had the backend system that tracks and bills that usage. This new deal now gives the company control over the storefront where developers access the models, too.

Settling digital transactions is a commercial dead end. Crypto can move it, settle it, but where is the money?

The underlying tech handles the real work at a massive cost, but the user-facing app captures the profit.

Verizon and AT&T spend billions digging trenches and building 5G cell towers so data can move smoothly. But they are essentially running a low-margin utility. The money is always captured at the interface. In this case, Apple sells the smartphone that uses the network, and Meta collects the ad revenue from the apps running on it.

This happens everywhere. Circle, for instance, makes 94% of its money parking customer deposits in Treasuries. Even though its routing protocol handles 60% of cross-chain USDC volume, it generates almost zero direct revenue because third parties own the settlement blockchains. Arc (mainnet launches on September 16) is Circle’s push to move up the stack and finally monetise the interface by not giving away all the transaction fees. Just moving and settling won't get you very far, unless it becomes a monopoly. Visa and Mastercard just move and settle data, but they make billions because they operate a closed system. Unlike a public crypto blockchain or the open internet, Visa is a private network. If a merchant wants to accept a card, they have to use this infrastructure and pay a 2% to 3% fee on every swipe.

The numbers for AI payments look big, but they don’t mean much. The x402 protocol has processed over 153+ million transactions since May 2025, with a lifetime volume of $40 million. This is an average transfer size of 26 cents. Even on cheap networks (like Base or Solana), paying a fraction of a cent in gas to send a 26-cent payment is not economical. In Web-3, 153 million micro-transactions are the classic signature of “Sybil farming” (users programming bots to send pennies back and forth to inflate network stats and qualify for future token airdrops) or developers just running load-test scripts in a loop.

Compare that to OpenRouter, which moves $900 million a year and retains 5%. The underlying payment networks do way more work but move much less money, all for tiny fractions of a cent.

Even specialised chains like Tempo launched without a native token. That means they deliberately gave up the massive, speculative revenue that usually comes with running a crypto network. They made that trade-off so their corporate clients could finally get safe, predictable fees

Although Stripe’s founders wanted to build infrastructure capable of processing millions of transactions per second, Tempo’s current utilisation is extremely low. Since going live in March, the network has averaged under one transaction per second. While load tests confirm capacity exceeds 18,000 transactions per second, real-world activity generated just $5.34 in daily fees and around $620 over 30 days. The chain holds $40 million in total value locked, but the settlement volume produces a very low economic return.

The platform talking directly to the customer is the one that gets rich. Crypto was supposed to win here by bundling AI models together and using token discounts to offer unbeatable prices. But that plan is failing. Crypto protocols might have access to cheap compute power, but centralised companies are the ones actually controlling the users and taking the profit.

Chutes, a decentralised network that rents out cheap AI computer power built on Bittensor, has processed 9.1 trillion tokens across 400,000 users. During its highest volume days in late May and early June, Chutes pushed nearly 22 billion tokens per day through the OpenRouter interface. But they depend on OpenRouter for up to a quarter of their daily volume. io.net, pushing 4 billion tokens a day, uses the same centralised platform for distribution.

Decentralised networks like Chutes can easily undercut Amazon and Google because they don’t have to pay for expensive data centres or corporate staff. Instead, they let independent computer owners supply the hardware, paying them in crypto rewards. Because their overhead is so low, they can sell AI computing power way cheaper than big tech. The weakness is trust, which is the entire value of the interface layer.

Quality control is a massive headache for OpenRouter. When you call an AI model, OpenRouter routes your request to the cheapest available independent server. But not all servers are the same; some use weaker hardware or tamper with the model to save money. In fact, a recent survey of 42 servers found that nearly a third completely hid their data settings. To fix this, OpenRouter is launching a “verified” tier for trusted providers. For crypto networks, this problem is even worse. Malicious nodes have an economic incentive to run downgraded models to maximise profit margins.

While projects like DGrid and Bittensor are developing theoretical models to detect output degradation, real-time verification at scale remains an unsolved problem.

Uptime is another operational hurdle. Because they don’t have massive, flexible server systems like big tech companies, networks like Chutes can crash during heavy traffic spikes. To maintain reliability, developers rely on centralised aggregators as a fallback, effectively paying the toll they intended to avoid.

On the money side of this, reported revenue figures differ from economic demand. Bittensor claimed $43 million in first-quarter subnet revenue, but independent analysis found a maximum of $15 million in verified external usage, with Chutes bringing in less than $2.5 million. Chutes gets handed about $54,000 every single day in freshly printed crypto tokens. They are being kept alive by crypto inflation, not because real people are lining up to use their service.

DGrid operates an OpenAI-compatible gateway that routes to more than 200 models, settles payments via the x402 standard, and audits outputs through its “Proof of Quality” mechanism. The platform reported $20 million in revenue across its first six months from over 13,000 users ahead of a planned DGAI token launch. That figure, however, came from selling $1,580 lifetime membership passes that grant future token allocations, making it a disguised token presale rather than recurring API consumption.

To see that decentralised networks can handle real traffic, look at Dippy. This massive AI character app, which has 8.6 million users, ditched traditional cloud providers and moved its entire operation to Bittensor’s Targon network. While the financial revenue numbers are tough to completely verify on-chain, it’s a huge win. It proves everyday consumer apps can rely on decentralised tech to work.

When a company is independent, it routes your traffic fairly. But when a corporation like Stripe owns the front door, conflicts of interest kick in. Stripe has its own payment tools, wallets, and blockchains to sell, giving it a clear incentive to push their own rails over competitors.

Crypto’s biggest selling point is that nobody owns it. A decentralised system can’t be bought out by a corporation, and no one can change the rules behind closed doors to favour their own apps. Stripe can build great software, but it can never offer truly neutral, un-rigged plumbing.

The interface thesis makes total sense too. I just don’t think crypto is anywhere close to owning it.

What Stripe did was put a public price tag on standing in the middle. They are reportedly paying $7 billion for roughly $50 million of revenue, because standing between eight million developers and ~500 AI models is a great business. If you are building an open-source router, you can now wave that valuation at venture capitalists.

Obviously, the trade gets awkward. Up until now, decentralised compute networks have grown by plugging into OpenRouter and letting it handle the annoying work of finding customers. Now, that front door belongs to Stripe, a company with every commercial reason to favour traffic that touches its own products. If you picked OpenRouter because it was neutral (say, you wanted to route queries to cheap Chinese models without corporate interference), you are probably getting nervous.

A router nobody owns is a router one can’t bias. The problem is that to sell it, you need to prove what code the miners ran, offer uptime that doesn’t rely on a centralised backup, and make money that doesn’t vanish the moment token subsidies end. We are not there yet. So clap clap, good for Alex Atallah.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Every major publicly listed bitcoin miner has spent the last four quarters facing an existential crisis. A look at their earnings releases over the past year shows they are moving away from their identities as miners of magic money.

They are now recategorising themselves as “energy infrastructure platforms,” “vertically integrated AI Cloud providers,” or “digital infrastructure companies built around power, land, and compute.”

The rebranding is backed by a shifting revenue mix and a growing market cap that has surpassed what these companies were worth as bitcoin miners.

It looks like a story of struggling miners finally finding their salvation by upgrading to meet the growing demand for compute and AI inference. Except that it isn’t as simple when you zoom out and look at what they are forgoing.

In today’s piece, I will explain why miners’ 180-degree pivot away from bitcoin mining could come back to bite them.

What Drove the Pivot?

First, the halving of block rewards in April 2024 pushed miners to either hold their mined bitcoin and sell it at a higher price, or diversify by using their processors for High-Performance Computing (HPC). But bitcoin’s rally from under $70,000 in October 2024 to $124,000 in October 2025 drove some miners to keep mining and stack their mined bitcoin.

But the October 10 liquidation event, which wiped out $19 billion of the crypto market cap in 24 hours, led to a market downturn the industry has yet to recover from.

Today, bitcoin price is down almost 50% from October’s peak at ~$63,000. At these levels, it’s almost impossible to mine BTC and sell it at a profit. Bitcoin mining costs vary with electricity costs and the Bitcoin network’s difficulty. Yet the average BTC mining cost rose from about $40,000 in February 2024 to $90,000-$110,000 between October 2025 and July 2026. In August, the all-in cost of BTC mining shot up to an all-time high of $140,000.

The unfavourable environment for BTC mining over the past year has coincided with the capital glut flowing into AI and inference markets.

Bitcoin miners use ASIC mining equipment that runs complex algorithms to win bitcoin as block rewards. This equipment is similar to what is needed for HPC and AI inferencing. This made the pivot from bitcoin mining to AI inferencing easier than starting from scratch. It is also a no-brainer for most miners looking to use their equipment for a profitable cause.

Nearly the entire listed mining sector moved at once, in the same direction. The scale of the shift is evident in how their revenue mix has shifted in the past few quarters.

Consider Core Scientific. In Q2 2025, revenue from colocation, the business of renting out data-centre capacity for AI and HPC, totalled $10.6 million a quarter. In the same quarter, revenue from mining and selling bitcoin stood at $62.4 million from mining bitcoin. Twelve months later, those numbers have almost exactly swapped. In Q2 2026, revenue from colocation is $136.7 million, while revenue from mining has shrunk 65% to $21.5 million.

Colocation now makes up roughly 83% of the company’s revenue, up from 14% last year.

TeraWulf’s HPC lease revenue is already 71% of its total. Its HPC leasing revenue rose from nonexistent this time last year to $32 million in Q2 2026. During the same period, its digital asset revenue shrunk about 75% to $13 million.

The shift is also apparent in how many of these miners have repurposed their existing mining buildings and equipment for AI or HPC operations.

The Worrisome Land Grab

The pivot from bitcoin mining to AI and HPC operations looks like a stampede. Beyond just reallocating existing capacity to AI, these erstwhile BTC miners are also racing to acquire as much power, land and contracted demand as they can.

The contracts already signed run into the tens of billions.

Core Scientific has committed up to 2.5 gigawatts of capacity to AMD and CoreWeave, deals worth north of $24 billion over their terms. Hut 8 has 949 megawatts under contract at a base-term value of $26.6 billion, backed by $7.5 billion in fresh project financing.

TeraWulf signed a 20-year lease with Anthropic for ~$19 billion and acquired a gigawatt-scale campus in Kentucky to meet compute demand.

In the last six months, Riot Platforms has executed leases totalling 241 megawatts of capacity for about $10 billion. Just last week, IREN delivered its first of four AI Cloud deployments to Microsoft under a five-year, $9.7 billion services contract announced in November 2025. It also has a five-gigawatt partnership with Nvidia and is buying power sites in Texas and Spain to feed it.

These deals and contracts explain the excitement around the AI landscape. A 20-year lease with a reputed hyperscaler tenant signals a more stable revenue pipeline than what these companies ever had before. After a year of mining BTC at a loss, a decade and a half of contracted cash flow feels like solid ground.

Except that there are caveats that many of the commentators are underestimating.

Bitcoin mining, for all its pain, has a property that AI hosting does not. Bitcoin mining has a self-restoring feature. When mining becomes unprofitable, miners switch off their machines. When enough of them switch off, the Bitcoin network difficulty drops, so whoever is still mining earns more per machine by spending less to run their equipment.

If enough miners exit the BTC mining landscape, those who remain may find it relatively cheaper to mine. As miners defect to AI, network difficulty has fallen sharply from its 2025 peak of ~156 trillion to 127.5 trillion.

Every company rushing out of mining is making mining more profitable for whoever stays back.

AI compute has no such mechanism. If the industry crowds with too many players offering compute power, the supply eventually overshoots and brings prices down.

Although the demand for AI and compute power is going up, the prices are likely to fall down - contrary to economics 101. In fact, AI companies are already slashing prices. That’s because in a nascent industry, resources are scarce, and so are those building and offering these resources. Scarcity creates a case for charging a premium. Today, the prices these AI companies are locking in carry a premium because compute is scarce. Today, GPU orders are placed even when they are temporarily out of stock. Every hyperscaler on earth is bidding for the same megawatts. This scarcity is driving more BTC miners and data centre providers to ramp up supply. But once this ramped-up capacity tightens the demand-supply gap and compute power becomes more readily available, prices will fall faster.

When those prices fall, the same 15- to 20-year lease that these miners are signing at today’s rates will seem overpriced.

The worrying part about the pivot away from mining to AI and compute processing is that the shift is easy to make only one way. Turning a mining setup into an AI campus is straightforward since the power is already there. But once they replace their ASICs with H100s and sign 20-year leases, those commitments will bind them for the next two decades and prevent them from switching back.

Bitcoin mining will inevitably swing back to profitability, driven by lower network difficulty and BTC’s price recovery from the $60,000 lows once the market downturn flips. When that happens, the fully pivoted erstwhile miners, bound by the multi-year leases, will have to watch it happen from the sidelines. It would hurt them even more if the AI and HPC prices they get out of their deals aren’t as attractive as they are now.

So should the miners not pivot to AI and simply wait for the mining equation to flip? I don’t think so. For most of them, the pivot was driven by survival. Mining at a $140,000 all-in cost against a $63,000 coin doesn’t make business sense. AI money was the only lifeline within reach. But the question of survival can often drive businesses to take extreme measures. Here, too, the problem is with those who completely went all-in on AI. Driven by desperation, these businesses fled BTC mining entirely, repurposed all their ASICs, and locked themselves into decades-long deals in a nascent industry already seeing price corrections across the sector.

But even as most miners did a full 180-degree pivot, some others are treading carefully.

Hedgers Hold an Edge

Marathon Digital, the BTC miner with the largest BTC treasury, is treading carefully. In Q2 2026, the company sold 30% of its BTC holdings to cut its debt obligations. But it chose a different approach to fund its AI expansion instead of scrapping its BTC mining rigs.

MARA drew $150 million under the 2026 Line of Credit using its remaining BTC holdings as collateral. As of June 30, 2026, the company still holds more than 35,000 BTC on its balance sheet and is planning to keep mining alive on a flexible load.

Hut 8 is also not abandoning BTC mining. It is spinning out its BTC mining business into a separate, majority-owned subsidiary called American Bitcoin Corp., allowing the parent company to pivot its core focus toward high-performance computing and AI infrastructure.

By keeping rigs warm and treasury intact, these operators hold an option the all-in converters liquidated. When mining turns profitable, they can redirect power back to it. The fully pivoted can only watch.

Miners going all-in on AI and treating their old BTC mining business as dead weight are betting on a future few have seen. But the trend is not difficult to predict. Every nascent technology, including chips, computers and smartphones, got cheaper as the industry and ecosystem matured. AI and compute industry is already seeing this happen.

Companies like MARA and Hut8 aren’t entirely out of the AI opportunity. But by retaining some mining capacity and BTC treasury instead of liquidating completely to buy GPUs, they are treating AI as a diversification play rather than committing all capital to an industry that hasn’t fully matured or been repriced accordingly. The stocked ASICs let them switch the load back to mining BTC when it pays the best and the compute premium thins.

When you zoom out, all these companies are essentially power companies that are topped with either ASICs or H100s. But those that can redirect power between mining and AI compute, depending on what pays best, will not be completely at the mercy of market cycles and narratives.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Dave Ramsden, the Bank’s deputy governor for markets, told reporters that getting into the building that morning had been trickier because a lorry was in the bullion yard.

The London gold market typically trades ownership records, while the physical metal remains securely in vaults. But during those uncertain times, accessibility changed the price. Bank of England gold became cheaper because it was stuck in a weeks-long withdrawal backlog. Commercial vault gold became more expensive because buyers were willing to pay a premium to access and move it immediately.

On an average day in May, the banks handling London’s gold clearing traded $73.7 billion worth of the metal among themselves without physically moving it. By the end of July, London vaults held 9,534 tonnes of gold valued at $1.2 trillion, roughly 762,000 bars. The clearing company says this is how the system is meant to work, since moving physical gold is a costly security risk.

Today, we are discussing why the Financial Conduct Authority (FCA) is writing rules for tokenised gold and how those regulations focus entirely on the ledger.

London is the centre of global gold trading. The London Bullion Market Association (LBMA) is the trade body and standard-setter. The final reconciliation of who owes what to whom runs through four clearing banks. HSBC. ICBC Standard Bank. JPMorgan. UBS. The electronic matching and clearing company operated by the banks is LPMCL (London Precious Metals Clearing Limited / AURUM).

The FCA has been talking to banks about how to regulate tokenised gold and whether it can be used as collateral in wholesale markets. This follows a joint paper published on 18 May 2026 by the FCA, the Bank of England and the Prudential Regulation Authority, which named tokenised gold as possible collateral for uncleared over-the-counter derivatives. There is already a precedent.

An FCA policy statement in April confirmed that a range of money market funds, including tokenised versions, can qualify as collateral for uncleared trades under UK EMIR.

Right now, 16 firms are actively testing tokenisation in the UK’s regulatory sandbox. The government thinks this could add £33 billion a year to the economy by 2035. We should see the first tokenised government bond by early 2027, right around the time the Bank of England updates its collateral systems, with digital ledgers linking up to digital sterling by 2028.

You will often hear that London is tokenising gold because it is scared of losing business to Asia. The reality is that the tech was homegrown by one of London’s own clearing banks. In late 2023, HSBC broke down standard 400-ounce bars in its London vaults into tiny digital slivers so institutions could trade them. They eventually rolled out a retail version in Hong Kong that processed $2.2 billion in trades, but the innovation itself started in London.

London gold market handles four specific jobs.

The first two are physical storage (vaults and guards) and quality assurance, which means proving the gold is legit so buyers don’t have to melt it down to check.

A token obviously can’t do either of those; it just piggybacks off the physical world’s legwork. The third job is keeping the record of who owns what. Tokens handle this part incredibly well, and for cheap, we know that.

But the fourth job is credit, and this is the catch. Tokenising gold actually makes the banks’ credit system obsolete. When you can instantly transfer ownership of physical gold with a token, you don’t need to lend your metal to the bank in exchange for trading convenience.

Most London bullion is held and traded on an unallocated basis. The customer does not own specific bars. They hold a general entitlement to a quantity of metal. The LBMA describes the arrangement as working like a bank deposit denominated in ounces. The customer is an unsecured creditor of the clearing member. That gold in the vault works for a living too. The banks pool it together on their balance sheets to keep the whole trading system moving. It allows them to instantly credit and debit accounts the second a trade happens, buying them a couple of days to sort out the actual physical delivery in the background.

Buyers have two options. You can claim specific, physical bars, which means paying storage fees and dealing with slow transfers. Or, you can hold unallocated gold, essentially an IOU from the bank. You take on a small amount of their credit risk, but you can trade instantly. The majority chooses the second route. The average transfer in February was about five bars’ worth of gold, and thanks to this setup, the metal never actually had to leave the vault.

What happens when it’s a token instead? It gives you the speed of an IOU with the hard ownership of a specific bar. When you can have both, there’s zero reason to take on bank credit risk. London is already fully electronic, so updating the settlement tech isn’t the big win.

The FCA chose to target collateral over trading because collateral is entirely dictated by speed. You have hours to meet a margin call, and the traditional plumbing for gold just moves too slowly. This leaves $1.2 trillion in gold sidelined, forcing institutions to use cash or gilts instead. Regulators know that putting that gold on a blockchain ledger removes the problem. Ownership can change hands instantly and in precise fractions, and all that vaulted metal can be used as top-tier collateral.

The securities world is already doing this. HQLAX allows heavyweights like BNP Paribas, Clearstream, and JP Morgan to trade ownership of their collateral without physically moving the underlying assets an inch. The SEC even gave it a green light in May 2026, allowing US broker-dealers to jump on board for a 36-month run.

This whole shift actually started long before crypto. Strict banking rules came first. When global regulators introduced Basel III’s Net Stable Funding Ratio (NSFR), treating unallocated gold as an illiquid asset requiring an 85% stable funding buffer, the London bullion lobby fought back hard, warning that clearing banks would abandon the market. Now, tokenisation is finishing what Basel III started.

My only problem with this is that it’s being rolled out by the very banks that tried to protect the old system.

Locking 12.5 kg gold bars in a vault sits outside the FCA’s remit, yet holding the tokens will soon require full regulatory authorisation.

The legal landscape shifted back in February when Parliament signed off on the new crypto framework. It means the FCA is finally policing crypto custody and trading platforms. Businesses have a five-month window starting September 30 to get their applications in before the whole system becomes legally binding in October 2027. It’s still a work in progress. The FCA is refining its rules for holding client assets, and lawmakers are trying to extend traditional market exemptions to cover the token world.

The physical gold does absolutely nothing new. The vaults, the insurance policy, and the guards just go on with business as usual.

When you change the ownership record to a token, the activity immediately enters regulated territory. That is where regulators actually draw the line. They monitor the legal rights to the gold. Because a token transfers those rights in an entirely new way, the FCA has to step in and establish new regulations.

If software makes record-keeping practically free, value flows straight to physical scarcity. In London, the vault scene is an exclusive club with four clearing banks and three security carriers. Nobody has joined this group in over a decade. Tokens won’t disrupt these guys because code can’t replicate a fortress. If anything, it makes them more essential, because you still need real-world custody to back every digital claim. These firms will just adapt their business model, switching from making money on customer deposits to charging clean fees for storage, audits, and collateral plumbing.

The idea of tokenised London gold has been around the block. Paxos and Euroclear attempted it in 2016, only to shut it down 13 months later. The setup works differently today because the clearing banks are the ones driving the ship. An external startup is no longer trying to force the tech in.

HSBC, for instance, built its own in-house system that handles every single trade. They created a walled garden to control how the transition unfolds. The ultimate fate of this market rests with the FCA. Regulators have to decide whether an HSBC token can be traded outside HSBC’s walls. If they say yes, the market changes. If they say no, the old banking model survives under a new name.

How will we know if tokenisation is actually taking over? You can easily track this transition through the LBMA’s published data.

The LBMA publishes two sets of numbers: how much gold is being traded (trading volume) and how much gold is officially changing hands on the old central books (clearing volume).

If people start using tokens, trading volume will remain high because they will continue to buy and sell. The clearing volume will drop. Because when you trade a token, the transfer happens instantly on the blockchain, completely bypassing the old London clearing system.

Right now, the old system is still doing a massive volume. Every single day, about 20 million ounces of gold change hands on the clearing ledgers, while roughly 306 million ounces just stay completely still in the vaults. That means roughly 1 in 15 ounces of gold is traded on paper daily without a single physical bar budging.

We invented mathematics, cryptography, and global networks, yet they all lead back to a hole in the ground where we stare at gold, eh?

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Welcome to Edition 15 of The Reading List — your fortnightly antidote to the noise. Every two weeks, I share my favourite reads across crypto, finance, technology, culture, and beyond. I also share what I’m watching and reading.

What I’m Watching and Reading

I just started When Genius Failed: The Rise and Fall of Long-Term Capital Management by Roger Lowenstein. Resuming my reading last week after a long break has convinced me of something few would relate with: you can absolutely fall in love with non-fiction finance books despite the numbers, for something few people expect them to do. The best of these books talk about human psychology and how we think and behave around money. The first 40-odd pages explain it perfectly.

Lowenstein opens with John Meriwether, the shy hedge fund executive who built the legendary Arbitrage Group at Salomon Brothers by importing academic quants onto a trading floor that ran on instinct and trust. There was one thing that impressed me the most about the initial chapters of this book. It’s how the author shows the culture that reflected the devotion the traders had for Meriwether. The book superbly articulates the intellectual arrogance and the certainty that risk could be deconstructed and tamed. You sense what will follow early, but the process of reading the prose that culminates into the inevitable detonation of what was building up underneath is what makes for an excellent read.

I’ll likely finish this one in the next week, so I’ll hold the rest of my review for the next edition of reading list.

Got any recommendations for nice watches or reads? Write to me?

The Fortnight That Was

July CPI rose just 0.1% month-on-month, bringing the annual rate down to 3.4%, with core at 2.5%. Both these were a touch below June and in line with forecasts.

Producer prices were even tamer as PPI was flat for the month against expectations of a 0.2% rise, and core came in soft too.

Markets read it as a sign to stop worrying about a rate hike, which didn’t seem entirely impractical last month. Hopes for a rate cut in September went back up.

The S&P 500 closed at a fresh record above 7,800 for the first time ever, carried by heavyweight tech and favourable quarterly earnings.

Gold climbed 7.87% in the last two weeks, with precious metals advancing across the board as traders digested a softer jobs report and looked ahead to the inflation data.

Silver has now surged more than 150% over the past year, propelled by rate-cut anticipation.

Even amid all this market chaos, bitcoin continued to move sideway with minimal volatility.

We, at The Token Dispatch continued to focus on the thesis of crypto fast becoming the underlying rails for the broader financial market.

What We Wrote at Token Dispatch

Most of our stories last week focused on who captures the value in the newly emerging world of Web 2.5. where traditional finance and crypto converge.

I closed the week with Distribution Trumps Asset Class, where I explore how crypto and Wall Street have started funding each other’s revenue. BlackRock’s spot bitcoin ETF became its most profitable fund, while Robinhood’s new chain became Uniswap’s single largest fee contributor within weeks. Hyperliquid now earns most of its revenue from oil and equity perps that have nothing to do with crypto. The ideological war between crypto and traditional finance means nothing beyond the fact that distribution was and still is the biggest moat any business can have. Earlier this week in Rewiring Commerce for Agents, I traced how Cloudflare, Google, Stripe and Visa are tearing apart every layer of online buying and rebuilding it for machines, where advertising dies and the real value moves to whoever can vouch for an agent’s identity.

Vaidik explained with Paying to Disrupt Visa, how the pitch of “cutting out the card network” is not as sensible as it sounds. He proves this by showing how every stablecoin card is still a debit product leaning on the Durbin Amendment and sponsor-bank workarounds, and cannot function by sidestepping Visa. In Dead Chains Walking, he followed Aave’s exit from six blockchains and showed how a chain dies once its credit market leaves. He makes a great parallel to global banks abandoning small-country corridors after 2008.

In the week before the last, Thejaswini wrote Ondo’s Taken a Decision, where she explains why Ondo scrapped its blockchain entirely and chose to keep Ethereum exclusively for settlement.

Crypto & Tech

• Netting is all you need: Simon Taylor explains why “instant” isn’t always better in payments. Taylor does this by explaining how banks’ netting and liquidity-saving mechanisms make slow rails cheap, and why 17 US banks chose tokenised deposits over stablecoins to port their moat on-chain.

• The Future of Vaults: Tempest argues on-chain vaults are the next crypto primitive to cross into TradFi after stablecoins and perps, with Coinbase, Kraken, Robinhood and Uniswap routing users into them. They explain where the vaults are expected to go from here.

• From porters to protocols: Byron Gilliam traces clearing houses from 19th-century bank porters lugging gold across Manhattan to CHIPS, and asks whether a blockchain protocol like Cycles can open the clearing-house club to everyone beyond banks and financial institutions.

• Technology finds its own uses: A nice take on how the internet may be remembered as the training data for AI, and how bitcoin’s mining infrastructure and talent are becoming the shovels for the AI build-out.

Finance & Economy

• Jane Street suffers $15bn hit after meltdown at Situational Awareness: The secretive trading giant was wrongfooted by July’s AI-stock rout, a rare setback even as it cleared $40 billion in net trading revenue, with the sell-off also torching Leopold Aschenbrenner’s leveraged AI fund. (behind paywall)

• The jobless boom has arrived: The WSJ’s framing of 2026’s defining paradox: record profits and a soaring S&P alongside stalled hiring, flat wages, and AI letting firms do more with fewer people. (behind paywall)

• Supermarket economics: A revealing interview with a grocery-union economist on how scanner tech, e-commerce, and the Kroger-Albertsons merger reshaped who captures value in the trillion-dollar business of feeding America.

• On edge and max extracting: polar uses a friend’s collapsing TikTok sponsorship income and the power-law distribution of poker hands to make a sharp point about how market edges decay, and why knowing the type of edge you hold matters as much as having one.

AI & Innovation

• NVIDIA speedrunning the creation of a synthetic hyperscaler: Clark Tang argues that as AI inverted data-centre economics, Nvidia is being forced to assemble its own hyperscaler because incumbents were too slow and too eager to commoditise its margins.

• Base Power & the Future of Electricity: a16z’s Profiles of the Future on how collapsing solar and battery costs made electricity generation cheap, even as the cost of delivering it surged, leaving America’s 20th-century grid a poor fit for the AI-power era.

• AI designed a virus, and scientists brought it to life: Stanford and Arc Institute researchers used genome language models to design the first fully AI-generated viral genomes, producing 16 working bacteriophages that killed antibiotic-resistant E. coli. The piece also reignites the biosecurity debate. (behind paywall)

• An AI for Africa would be built on Hunhu/Ubuntu ethics: Fainos Mangena argues that Western-built AI smuggles in individualism and autonomy, riding roughshod over the communal “I am because we are” ethics that animate much of African moral life.

Culture & Beyond

• Velocity is the new authority: Om Malik’s argument that the organising principle of information has shifted from authority to sheer speed, where what travels fastest wins and thoughtful work simply doesn’t move.

• When is stereotyping a handy tool, and when is it a sin? A prize-winning philosophy essay on why generalising is a cognitive necessity we’d be lost without, and where exactly the line falls between a useful shortcut and a moral failure.

• What do you want?: Calvin Rosser on the central, unglamorous task of adulthood: figuring out what you actually want, why “safety” is an illusion, and why no one else can hand you the answer.

That’s all for this fortnight. Catch you in the next edition.

Until then, happy reading!
Prathik

P.S.: Want to recommend interesting literature? Write to me, and I shall include them in the next edition 👇🏾

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Thanks for reading! Subscribe for free to receive new posts and support my work.

Last Sunday, I wrote about the role of speculation in the evolution of finance. While working on it, I also pondered the evolution of the crypto industry itself.

Closer to home, a majority of my friends and family still don’t understand my work. They think I am shilling a coin or writing to persuade someone to throw away their life’s savings on some shitcoin. It’s sad they think so, but I don’t blame them. The initial ethos with which the crypto industry took off was aspirational, but also radical. Its marketing was problematic.

Many expected crypto to be the escape hatch from the current, problematic financial system. Even Satoshi Nakamoto had conceived Bitcoin as a parallel, peer-to-peer electronic cash system that was decentralised, permissionless and could route money around banks and third parties.

The problem began when enthusiasts built the crypto industry — much of which we saw rise and fall over the last 18 years — by swearing by decentralisation and permissionless as their core ethos. While I don’t have anything against these principles, ethos alone cannot build great businesses. The problem was that the builders expected a niche set of ethos to drive mass adoption and turn their hypotheses and ideas into successful, money-making businesses. An unpopular ethos doesn’t drive the masses to adopt a radical product. Even less so when that product could wipe away their life’s savings overnight without a warning.

Every niche product needs a familiar face to market it. Crypto, too, found its familiar marketer in the very thing the enthusiasts expected it to topple - the traditional finance industry.

In today’s piece, I explore how crypto and traditional finance feed off each other’s moats to build successful business models that don’t care a dime about the radical ethos on which enthusiasts expected crypto to be built.

On to the story…

Warming-up Masses to Crypto

What made crypto legible to Wall Street, my parents, your friends, and a layperson who never understood a private key was the financial instruments they were already using for other investments.

Bitcoin and Ether have been around as cryptoassets for 18 and 11 years, respectively. Yet they reached a broader base of holders only when the traditional finance industry wrapped them in an exchange-traded fund (ETF). The familiar traditional finance products have existed for almost four decades and are sold through the same brokerages that everyone already trusts. Think of BlackRock, Grayscale and Fidelity.

Those who bought a spot Bitcoin ETF did so because they believed bitcoin is an aspirational and emerging asset class. Few would have done so because they believed in permissionless money or to beat the centralised banking system. The ETF wrapper gave them exposure through the same brokerage account that routed their index funds and other investments. The wrapper also handled custody, since the underlying bitcoin was held by reputable names like Fidelity and Coinbase.

Familiarity was the marketing win that bitcoin needed. Once that part was taken care of by a trusted player, the underlying product didn’t matter whether it was crypto or otherwise.

The world’s largest asset manager, BlackRock, generated over $160 million in annualised fee revenue through its crypto ETFs in Q1 2026. In October last year, its spot Bitcoin ETF also became the company’s most profitable fund with $244.5 million in annualised fee revenue, surpassing an S&P 500 tracker that has been collecting fees for a quarter of a century.

Traditional finance learnt quickly that if it can borrow crypto’s products and wrap them in its existing, familiar instruments like ETFs, then its loyal customer base will be indifferent to the asset class being offered. Even those who barely understood how bitcoin worked and derived its value from, happily lapped it up.

Once BlackRock learnt that it had a native audience who was seeking exposure to the emerging asset class, its own chief had to walk back his words. After famously dismissing bitcoin in 2017 as an “index for money laundering”, CEO Larry Fink went on to build a $100-million ARR crypto fund within two years of its inception.

Traditional financial institutions no longer gain just passive price exposure. When Morgan Stanley listed a Solana Trust ETF in July, it charged 0.14% as a management fee but also staked that Solana to earn yield. It now passes roughly 95% of the staking yield straight through to holders.

The trend of traditional finance businesses profiteering off crypto derivatives and products also holds true the other way round.

Crypto Infra, TradFi Distribution

On July 1, Robinhood launched its own blockchain and allowed more than 30 million of its funded users to buy exposure to tokenised US stocks via blockchain infrastructure most of them had never heard of.

Within eight days of its launch, Robinhood Chain was doing $500 million in daily volume on Uniswap, the largest decentralised exchange in crypto. That was a 10-time jump from the day before, making it Uniswap’s second-busiest network after Ethereum mainnet.

Cumulative swap volume crossed $6 billion in ten days.

Since its launch on July 1, Robinhood Chain generated over $36 million of Uniswap’s $61 million in cumulative fees as of August 3. That’s about 60% of the protocol’s total fee revenue from July 1 to August 3. Ethereum mainnet and Base each made $8 million.

Since the launch of its chain, Robinhood has become the single largest contributor to Uniswap’s fee, beating crypto’s most established venues like Ethereum and Base chains by more than fourfold.

Here, Robinhood capitalises on the distribution muscle it has built with customers through its app, brand, and retail relationships over a decade. Although Uniswap runs the infrastructure, it was Robinhood’s chain that facilitated the trading activity. Early volume on the chain was dominated by memecoins, the kind of speculative retail flow that Robinhood has historically attracted.

Uniswap routed these protocol fees into buying back and burning its native token, UNI, and putting a share of its profits back into the hands of its holders.

Uniswap has bought back and burnt more than 1.36 million UNI tokens in the last 30 days. Out of that, fees generated from Robinhood Chain accounted for buyback and burning of 510,000 UNI tokens. Out of all the $UNI burns since July 27, Robinhood Chain contributed to over 50% of them.

Ironically, in Q2 2026, Robinhood’s own crypto trading revenue fell 38% year-on-year to $100 million — barely 13% of its transaction revenue, down from 53% two years earlier. Even as the fees Robinhood earns from its customers trading crypto on its app was shrinking, its chain was making a crypto DEX richer.

Meanwhile there’s a crypto protocol that now earns most of its revenue off oil and equity derivatives.

Hyperliquid’s perpetual futures, or perps, began trading on everything crypto. One could put a leveraged bet on the direction of bitcoin and Ether. But ever since October 2025, after the launch of deployer-run markets under HIP-3, perp contracts on non-crypto markets have been gradually picking up pace. Traders started listing perps on assets unrelated to crypto, from commodities like oil, gold, and silver to equity indices like the S&P 500 and Nasdaq, to foreign currencies and individual stocks like Nvidia and Tesla.

Real-world assets were 1.8% of Hyperliquid’s volume in the last quarter of 2025. By the second quarter of 2026, they accounted for 32.2%. In July, non-crypto trading crossed 52% of all volume across Hyperliquid.

Hyperliquid charges the same fee whether you trade a bitcoin perp or a crude-oil perp. Hyperliquid’s blended take rate on trading is 3.3 basis points, or 0.033% of volume. Every dollar made in fees flows back into buying and burning its native token, HYPE, via the Assistance Fund.

Although under HIP-3, Hyperliquid parts with half the fees, the protocol wouldn’t mind it, considering that deployer-run markets now make up the majority of Hyperliquid’s revenue.

Crypto rails, in short, have stopped caring where their revenue comes from. Hyperliquid doesn’t care whether its traders trade more traditional or crypto assets, or whether markets are up or down; as long as they trade, it keeps filling its coffers.

The Fintech Evolution

The pattern of decoupling revenue from the asset class isn’t unique to crypto and traditional finance. It’s recurred at every pivotal point in the evolution of financial markets.

Netflix began by mailing DVDs to a niche set of film buffs who were patient enough to receive a disc on rental by post to watch things that local stores didn’t have. Once Netflix built subscriber relationships and billing habits, and collected data on what people actually watched, it helped them expand their revenue by enabling streaming videos over the same infrastructure for a much larger audience. It then used the same distribution moat it had built to start making native content.

This is what happened on Hyperliquid, too.

Crypto perps appealed to a narrow, cyclical audience of leveraged traders. Once the platform proved its utility to that audience, it used the same infrastructure and interface to start perp contracts for something that was less cyclical than crypto. Oil, gold, FX, and pre-IPO stocks followed.

Robinhood followed the same playbook by using its initial commission-free stock trading as a wedge to build massive distribution. Now that it commands the attention of 30 million funded users, it can now sell options, prediction market contracts, and tokenised equities to the same audience and make money off it. Even if revenue from crypto trading falls, the same platform can keep making money off selling tokenised stocks and event contracts.

Both crypto and traditional institutions have now begun to acknowledge that the core question is “Where does the value accrue?” The answer is a common moat that has been a proven ingredient in building a successful business: distribution.

As long as there is massive distribution, any business - crypto or traditional - can cross-sell adjacent products, whether traditional or crypto, to their loyal user base. In all of this, crypto’s blockchain infrastructure does offer a cost-effective and faster rail to move money and assets. So whichever businesses use it can reduce their operational and logistics costs over time and pass some of that on to their customers for a better competitive edge.

Although this convergence of crypto and traditional finance is not the one crypto enthusiasts hoped for, it’s still the one that makes the most money. It’s also the one route that is in line with the trajectory of historical evolution. The efficiency of moving assets has always been the biggest motivator of every financial innovation. It matters much more than any other ideological principles. It always will.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

The story goes that Morton had a method. He would visit a nobleman’s house and look around. If the man lived lavishly, Morton concluded he was rich and could clearly afford to give the king a gift. If the man lived modestly, Morton concluded, he was saving money and could therefore also afford to give the king a gift.

There was never a third house, and no matter what Morton observed, the result was the same and the nobleman paid.

We only have this second-hand. Francis Bacon wrote it down in 1622 as a tradition people still repeated, and the name Morton’s Fork only entered common use in the 1800s. But the shape of it stuck around because it is useful. A fork, in this case, refers to a situation in which both arms lead to the same result. The opposite of win-win.

Ethereum and Solana are standing at such a place right now.

Both chains pay their validators with newly created tokens, and both are attempting to reduce those payments. Ethereum’s effort has faced strong opposition, while Solana’s initiative is currently being voted on, with the vote concluding on 18 August.

On one hand, if you continue to pay validators their current yield, staking favours the largest operators with institutional funding. On the other hand, if payments are reduced, smaller operators suffer first because their costs are fixed and their margins are razor-thin.

Fewer validators either way. So today, we are pulling apart the issuance debates on both chains to see how they are being forced to choose which flavour of centralisation they prefer.

On 4 August, Ethereum researchers, including Justin Drake and Jérôme de Tychey, issued a draft known as Tapered Issuance Burn, referred to as EIP-8363. The idea is that as more ETH is staked, the protocol burns a bigger slice of the validator reward.

Once staking reaches 60.25 million ETH, roughly half the supply, the burn hits 100%, and issuance yield goes to zero.

Today, about 41.4 million ETH is staked, representing 34% of the total supply, by about 890,000 validators earning an average rate of 2.67%. When Stani Kulechov from Aave put forward the proposal at that level, validator income dropped from 2.862% to 1.476%, that being just under half.

Within three days, Kulechov, SharpLink CEO Joseph Chalom and ether.fi’s Mike Silagadze had all come out against it publicly.

EIP-8363 is still an early, unapproved idea actively undergoing initial review on GitHub, and is not near being finalised or implemented. It missed the deadline to be included in Ethereum’s upcoming Hegotá upgrade. But it could be approved for a future network upgrade.

To understand why stakers are fighting these cuts so fiercely, you have to look at the sheer size of the pie. Ethereum pays people to secure its network. The protocol creates about 1.1 million new ETH a year and hands it to validators. At $1,921, that’s a $2.1 billion annual payroll.

Solana does the same thing, just on a much larger scale relative to its usage. It issues roughly 19 to 22 million SOL a year, about $1.5 billion at today’s price. Users pay 6,400 to 9,600 SOL a day in fees and Jito tips, around $225 million a year. So users cover about 13% of what Solana’s validators earn, and new supply covers the rest. Ethereum is no different. Chalom puts tips at roughly 15% of staking yield, meaning issuance funds the other 85%.

Solana creates 3.7% more SOL every year. Ethereum creates 0.85% more ETH. So a Solana holder who doesn’t stake gets watered down more than four times as fast as an ETH holder who doesn’t stake.

In traditional finance, the National Securities Clearing Corporation (NSCC) is involved in almost every US stock and bond trade. Its parent, DTCC, processed $4.7 quadrillion in securities transactions in 2025 and has $115 trillion in custody. NSCC’s protection fund (the money available if a member defaults) is $19.7 billion, contributed by members. NSCC’s own contribution to that fund is $130 million.

To compromise Ethereum, an attacker would have to overpower 41.4 million staked ETH, a $79.6 billion wall of capital that is four times larger than NSCC’s fund. This is just floor; if someone actually attempted to buy millions of ETH, the sudden demand would drive the market price through the roof. Furthermore, the attacker would have to fund a massive, global network of servers just to put those coins to work. And because Ethereum is designed to defend itself, one malicious move triggers the network to instantly burn the attacker’s stake. A failed attack means losing every single penny forever.

But unfortunately, these systems don’t work the same way, right?

NSCC members post that $19.7 billion as a cost of membership. The price of access to the market, without a yield from that money. Members would prefer the number were smaller. Ethereum pays people 2.67% a year to post the equivalent. Solana pays 5% to 8%.

When you pay someone a yield for four years straight, businesses grow on top of that yield. Right now, about $35 billion in special reward-earning tokens like stETH are held as locked deposits in crypto lending apps. Traders use these tokens to run a repeating money loop. Traders put them on Aave or Morpho, borrow WETH against them, stake that again, and repeat. The trade only works if you earn more from staking than you pay to borrow. Pendle constructed fixed-rate markets on the same yield. Curve has pools where people can get out. SharpLink has a $3 billion ETH treasury and stakes most of it through Coinbase, Anchorage, Figment, and Galaxy.

Staking yield serves as the fundamental benchmark interest rate across DeFi. A 50% haircut to consensus rewards flips the looping trade from profitable to negative, forces Pendle’s fixed rates to correct, and mandates a network-wide revaluation of LSTs by lenders.

Before the Merge, Ethereum issued around 13,000 ETH per day to miners, but after the Merge, it issued only about 1,700 ETH per day. That was an 88% reduction, which was carried out all at once. Since miners had invested years and billions of dollars in hardware, they resisted the change, ended up forking the chain into ETHW, and that token is now trading at less than 1% of ETH’s value.

Miners were entirely disconnected from the DeFi economy. They provided a service and collected a paycheck, but nobody was building complex financial products on top of that paycheck. Because their income wasn’t being used as collateral for loans across the network, wiping out their revenue overnight had zero impact on the lending markets. When miners walked, the rest of the system carried on without making any adjustments.

Stakers do two jobs at once. They secure the network, and the tokens they get back are collateral in half of DeFi. So you can’t remove their income without touching everything built on it.

As Mancur Olson pointed out in 1965, a small group that has a great deal to gain can outperform a large group with only a small stake because the smaller group takes the initiative and shows up.

Running a validator costs money whether you earn or not. A machine, power, a connection and so on. Stake 32 ETH today and you earn about 0.92 ETH a year, roughly $1,760. Under the new taper proposal, revenue drops to just 0.47 ETH ($900). The machine costs the same either way, so a bill that was eating a fifth of your income is suddenly eating close to half. If you miss an attestation, and you lose the same ETH you always did, you lose half the revenue.

The big group (as per Olson’s theory) is the regular token owners. Every year, Ethereum creates new coins to pay stakers. This makes regular ETH slightly less valuable. They don’t fight because it costs each regular holder only a few pennies or a fraction of a per cent a year. It’s so small that not many notice or care enough to protest. The small group is big staking companies. They get all of those newly created coins, which add up to billions of dollars. Their whole business depends on that money. If the network reduces those payouts, these companies lose millions.

The main argument for is that the pay is so good it keeps pulling ETH in, and most of it goes to a few large exchanges and staking providers. Staked ETH rose about 15% in the first half of 2026 on institutional flows. Their fix is to cap staking by making the marginal stake unprofitable.

The counter is that if you cut the pay, the everyday people running small setups at home will go broke first. Both sides actually want the same thing, which is to stop a few rich companies from controlling Ethereum. They still disagree on whether cutting pay or keeping it will ruin the network faster.

A Solana validator pays a vote fee of about 389 SOL a year. You pay it whether you earn anything or not, whether the market is up or down, whether anyone delegates to you. At today’s staking yield of around 6.5%, break-even sits near 200,000 SOL of delegated stake. Active validators fell from more than 2,500 to roughly 683 now, while total staked SOL climbed to about 430 million, near 68% of eligible supply.

Solana is voting to reduce the rewards. SIMD-0550 would double the disinflation rate from 15% to 30% per year, pulling the terminal inflation rate of 1.5% forward from 2032 to 2029 and removing an estimated 18.9 million SOL from future issuance. SIMD-0553 reworks fees based on resource usage and burns them, lifting daily burns from around 648 SOL to between 7,500 and 9,000 SOL. Even at the high end, that’s a fraction of the 60,000 SOL a day going out. Voting closes on 18 August and needs a 66.67% supermajority of stake.

We like to treat blockchain governance as an exercise in sovereign self-determination, where code and community votes dictate the future of digital economies. Yet the structural drift of both Ethereum and Solana suggests that software protocol rules are ultimately downstream of financial realities. The original architects designed the economics, hoping the market would naturally stay decentralised. Yet, when an asset becomes the bedrock of global liquidity and institutional treasuries, the brute-force mechanics of yield, leverage, and corporate capital expense will always overpower those intentions.

If the network is scaling by paying people to lock up their money, it will always run into this wall. It doesn’t matter how Solana votes on Monday, or what Ethereum decides to do with its proposals down the line. They are only picking how quickly we get there and hit the wall.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

In the last 10 weeks, Cloudflare built an entire internet stack to rewire commerce end-to-end for non-human players. The infrastructure of the San Francisco-headquartered company claims to power 1 in 5 web requests across the globe and 42% of the Fortune 500 companies. The same company that built its reputation by offering us, humans, a range of internet solutions like content delivery network services and cloud cybersecurity is now betting on an AI-dominated internet.

So much so that Cloudflare has rebranded its website to explicitly state that it is now building for an era where AI agents capture the majority of activity on the internet.

Since the start of July, the company has shipped a browser, a wallet and a storefront for machines specifically. This rewires every layer of the commerce stack, including surfing, identity, settlement, and holding money in a way that is far from being native to how humans do commerce and transact.

Cloudflare isn’t the first one to build for an agentic–first internet and commerce world. But its series of launches focusing on the agentic era calls for a closer look at how the entire journey of buying something is being taken apart and reassembled for machines.

In today’s piece, I will explain layer by layer who captures the value in this new stack as more companies rush to rebuild commerce for agents.

The Browser for Bots

To understand why AI agents need a different infrastructure than the current one, you need to first understand how they function.

An AI agent is just a piece of software that does a task for you. It could be something as simple as booking a flight or renewing your streaming platform subscription. To do these jobs, the agent needs to move around the web, prove its identity and pay for the services or products it shops for you. But the shopping journey of these machines is not similar to humans. They don’t open a flight booking website and stop scrolling at a fancy-looking ad or well-written copy. So internet infrastructure companies must rebuild each of those steps in the internet journey.

Last week, Cloudflare launched Kitesurf, a browser that they built specifically for AI agents. The interesting part about the browser is the features they stripped off an average browser that’s made for human users.

Normal browser builders spend enormous effort and time making a page pleasant for a human. They need to focus on fonts, colours, animations, where and how the little tabs stack, and the extensions you can install. Kitesurf drops all of it and keeps only the underlying information that a machine can read. AI doesn’t care about tabs or themes. All that matters to your bot is the cost of getting its query resolved and the speed at which it can find answers.

So, a browser built for agents rewires the surfing aspect. Such a browser lets the agents seamlessly surf the web for the services they need to buy or find for their masters. Then, we move to the buying aspect.

What Buying Entails

Let’s play back how we buy anything online.

We first discover it either through a search, an ad, or a shop we usually shop from. Sometimes we get persuaded to buy something we were not originally looking to buy. It could be a brand placement, a review, a discount, or a nicer-looking product on the shopping page. Then comes the difficult part of making a decision. Once we move ahead with the cart, we prove our human identity (to a machine, ironically) using a login or a captcha and use our saved cards or other methods to pay.

When we hand over this entire journey to an agent, a lot of steps become irrelevant, while some new ones take over.

An agent does none of the scrolling and discovering interesting products through ads or fancy landing pages like we humans do. Instead, it prefers to read through a clean, structured list of products and prices, pick a few that match the instructions it was given, and hand its master a shortlist of products that make the cut.

As humans, you could jump to page 2, 3 and so on of the Google search if you were not satisfied with the results on the first page. But the agent’s shortlist cuts all that crawling and gives you a shortlist that might be most relevant to your instructions. We have already started using this at scale. A survey of over 1,000 US consumers found that 57% of them leverage AI for product research at least weekly.

This shows that AI agents significantly cut the seller out of the shopper’s decision-influencing process. That influence and power now move to the agent’s shortlist. But who influences what the agent selects in its shortlist? It all depends on the format of the information that the agent reads before choosing the products, and a handful of companies are racing to control that format.

The Discovery War

In September 2025, OpenAI and Stripe launched Agentic Commerce Protocol (ACP), which now runs shopping inside ChatGPT. In this, a merchant shares its catalogue with OpenAI. When the buyer searches for their requirement through a prompt on OpenAI’s native chat interface, these catalogues surface and can be used to place an order and close the sale within the same chat window.

The shopper never visits the merchant’s website. All the discovery and purchase happen in the chat interface. To cut the noise of excess and irrelevant choices and drive buyers to shop for what they were looking for, OpenAI charges the merchants 4% of every order. It’s also a toll for putting the seller’s product in front of OpenAI’s distribution.

This 4% is what came back to bite OpenAI. The protocol never took off. Although Etsy went live on day one, only about 30 Shopify merchants out of the millions in its network had bothered to plug in even months after its launch. The blow made OpenAI retreat from its strategy by early 2026 and now reroutes shoppers back to the retailer’s site to finish the purchase, handing merchants back the control the original design took away.

Google and Shopify learned from this and joined hands for a contrasting bet.

Their Universal Commerce Protocol (UCP), launched in January 2026, let the merchant keep the catalogue on its own site from day one. The UCP neither plays an intermediate warehouse of catalogues, nor does it charge any commission to facilitate a sale. That’s because UCP isn’t a centralised marketplace like OpenAI and Stripe’s ACP. Instead, it chose to run an open, decentralised standard and make the communication layer entirely open.

Then, how does Google make money? Through ads and paid placements of products by retailers.

By April 2026, Amazon, Microsoft, Meta, Salesforce and even Stripe had joined Google’s UCP. Walmart, Target, Best Buy, Visa and Mastercard have also signed on.

Both models change the incumbent commerce model, which was built specifically for human-driven discovery. The human-centric model required merchants to come up with appealing advertisements and landing pages to win human attention. The new agent-driven commerce world needs merchants to satisfy a query that both shoppers and merchants will rarely see.

The Dying Layer

A huge share of commerce spending goes to persuasion through advertising, branding, and loyalty points. The whole craft of nudging an undecided human toward buying a product or service is a capital-intensive task. It could cost up to hundreds of billions of dollars a year. But the spending makes sense for humans because it depends on the assumption that the buyer can be moved. That assumption is based on proven science about human behaviour that has significant empirical backing.

On the contrary, an agent doesn’t feel emotional about an advertisement or a brand story. The only thing it cares about is the product specifications and the instructions it has been given.

So this layer gets deleted altogether in the commerce stack that’s being rewired for agents.

The new layer is already selling brands a way onto the agent’s shortlist, just the way search-optimisation firms once sold an inorganic climb up Google’s results pages. It’s just that the nudging is no longer aimed at a person, but is in the form of a paid listing or a discounted offer to grab the agent’s attention.

But what’s clear is that the profitable business of changing a human’s mind at the moment of purchase will cease to exist. There’s no more value left in this layer to capture.

Where Does the Value Accrue Next?

When an agent shows up to buy, the shop has a new problem. Who sent this thing? Is it a real assistant or a bot here to cause trouble? Is it allowed to spend this person’s money, and how much? A human has autonomy and identity. An agent is just software and a few lines of code that could be representing anyone. It is still representing a human (hopefully), but could be acting in bad faith. So somebody has to vouch for it, keep the list of trusted agents, and let shops check that identity at the first step.

That job is the most valuable layer in the stack. Think about why Visa is worth more than almost any bank. It does not lend money and barely moves it. Yet it facilitates the moment of trust. Every time a stranger’s card clears, Visa is the one certifying and differentiating bad actors from the good ones. Each time it does that, it takes a thin sliver of the transaction it facilitates.

The same Visa has already built a registry to vouch for these agents. This certification helps merchants trust the agent and let it spend.

Cloudflare is also chasing the value this layer generates. Its cloudflare.pay accounts give an agent an identifiable name. Cloudflare holds an account edge others don’t. It already has a distribution base accounting for a fifth of the global internet. That allows the company leverage to restrict agents that don’t verify themselves by inserting friction. It could become for agents what the domain name system (DNS) is for websites.

How This Evolves

Those watching this space will be most interested in a few key aspects. Will the agent identity become a whitelist that Visa or Cloudflare owns and charges for? Or will it remain an open standard that belongs to no one, like how Google and Mastercard are pushing?

Cloudflare holds the edge because of its presence across several layers at once. The company offers a browser the agent can surf with, the name it identifies the agents by, and the wallet it enables the agents to pay from. All of these sit on one network which already services 42% of the Fortune 500. Owning multiple layers across this stack lets Cloudflare charge a fee at each of them, while undercutting its competitors by offering a subsidised rate collectively for all its services.

But Cloudflare will face stiff competition from Google because of the loyalty and data bank it already possesses. Shopify integrates millions of merchants into Google’s standard, and Amazon, Walmart, Target and Best Buy collectively bring the retail might of merchants

While Cloudflare controls the infrastructure the agents travel through, Google holds an edge in hosting the merchant catalogues the agents shop from. An identity standard is only as strong as the merchants who honour it, and Google has the loyalty of those merchants here.

It will be a close fight to watch.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Did you know that crypto card spending crossed $759 million in July, across 9 million purchases? That is nearly two and a half times what it was just a year ago. And over 90% of that volume still runs on Visa.

And every one of these cards will sell you the same story: we put payments on Stablecoin rails to cut out the card network fees and hand the savings back to the merchants. It is the same idea we explored when we wrote about how Stripe is building its own Stablecoin-powered cross-border payment chain.

So I went deep looking into what actually happens if you try to remove a card network. Does avoiding Visa or Mastercard really save the merchant money? And which layer do stablecoins actually replace in the payment stack?

The answers were not at all what I had expected.

How the Payment Stack works

To get answers to those questions, I first had to understand where the money actually goes when someone swipes a credit card. And the first thing I realised was that most people, including people in crypto, assume that the card network like Visa is the one taking the biggest cut.

Wrong! When a merchant accepts your $100 purchase on a rewards credit card, they pay what is called a merchant discount rate of approximately 2.2%, which comes out to $2.20. But here’s the interesting thing: that $2.20 does not go to Visa; rather, it gets split between three different parties, and the split is not even equal.

The largest share, about $1.75, goes to the issuing bank, i.e the bank that gave the customer the credit card. These fees are called interchange, and they make up 70-80% of the entire merchant fee. Next, the merchant’s payment processor, also known as the acquirer, takes somewhere between $0.30 and $0.70 as its markup. And finally comes Visa or Mastercard, the actual card network that everyone in crypto wants to disrupt, which takes around $0.13 to $0.18 as its assessment fee. Which is roughly just 7-9% of what the merchant paid in total.

So if you remove Visa from this equation, you are just removing the smallest line item in the entire stack, and there is also a reason why Visa earns such a small fee.

See, Visa does not lend money to anyone, and so it does not have to deal with all the credit risk, chargebacks or fraud disputes. In fact, Visa does not even move the money. It is simply a messaging network which only activates when you tap your card at a store. Visa’s job is to send the authorisation message from the merchant’s terminal to the issuing bank and back, and it sets the conditions under which everybody in the system operates. The party that is actually doing most of the heavy lifting is the issuing bank. It is the one that extends the credit to the consumer and the risk that they might never pay it back. It is also the one which absorbs the float between the purchase of the good and the statement due date and uses the interchange to fund the rewards programme that got the customer to use the card in the first place.

This is also what makes Visa’s business so fascinating. In 2025, Visa processed $14.2 trillion in payments volume across 257.5 billion transactions and generated $40 billion in net revenue at a net margin of close to 50%. It earns roughly $0.13 cents per transaction on average, and that’s its entire business model. The reason why Visa is one of the most valuable companies on earth is not that it charges a lot per swipe but because it processes a quarter of a trillion transactions every year with almost no marginal cost and zero credit risk.

Now let’s talk about the part where things start to get really uncomfortable for stablecoin cards.

Every single stablecoin card is a debit product. The money is already sitting in the user’s wallet as USDC or USDT even before the purchase happens. Also, there is no float and no revolving balance generating interest income on the side. This puts these cards in an entirely different economic category.

Another thing is that in 2010, Congress passed the Durbin Amendment, which capped debit card interchange for any bank with more than $10 billion in assets at $0.21 cents plus five basis points per transaction. The way most stablecoin card programmes get around this cap is by partnering with a small sponsor bank (neobanks) that falls under the $10 billion asset threshold, which makes them Durbin-exempt. And this is the exact reason why there is a sponsor-bank model that exists in fintech.

And the exempt debit interchange on dual-message networks currently averages at about $0.62 cents per transaction. So here’s the thing: a rewards credit card works with about $2.20 of gross revenue on a $100 transaction. But a stablecoin debit card, even at the higher exempt rate, still works with just $0.62 cents. And out of those 62 cents, the programme still has to pay the network, its processor, sponsor bank and also cover its own fraud losses and operating costs before anything at all is left over for the merchant.

So What Actually Gets Replaced?

So the point is, fee savings from removing Visa are tiny, and the economics of running a stablecoin card on debit interchange are extremely tight. But maybe the real value of stablecoins in payments is not about saving a few cents on the network fee. Maybe it is about replacing something more important in the payment stack.

To figure that out, I went back to our piece on why Stripe built its own chain, where we had broken down a cross-border payment into seven toll-collecting functions: Acceptance, Orchestration, Licensing, Custody, FX, Issuance, and Settlement. I wanted to map a domestic card swipe against those same seven layers and see what actually changes.

So if we look at this from a merchant perspective, Acceptance, which is where the transaction starts, is still completely untouched. Meaning the process is still the same: the merchant has a terminal, the customer hands over their card, and they still pay the same merchant discount rate to an acquirer.

The merchant does not even know that the funding source behind the card was USDC. As far as they are concerned, a Visa transaction still shows up in their batch just like any other transaction. Similarly, the Orchestration layer still routes through Visa or Mastercard, and Issuance still happens through a sponsor bank and a BIN under a network’s brand. Fintechs have been doing this since well before stablecoins existed, and nothing here has changed.

Where exactly things start to change is on the backend, and that’s where it starts doing numbers on your psyche.

Settlement is the only layer where stablecoins make a fundamental difference. Traditionally, when a card transaction settles between the issuer and the network, there is a T+2 cycle, with weekend delays and batch processing. To fix this, companies like Rain provide daily settlement with Visa in stablecoins, and Mastercard has also started settling in stablecoins like USDC, PYUSD, and RLUSD with intraday settlement cycles. This compresses the T+2 settlement to something much closer to real-time and frees up working capital that issuers would otherwise have sitting locked in the settlement float.

But the working capital benefit goes entirely to the card issuer, and the merchant’s discount rate also does not change because of faster settlement. The consumer or merchant notices absolutely nothing different during checkout.

The only party that benefits from the T+0 stablecoin settlement is the programme operator who no longer has to fund two days of float. And the actual innovation here that stablecoins enable is treasury improvement for the issuer, which at scale is worth a lot but doesn’t provide any marginal benefit to the merchant or the customer, to say. This also does not give Visa any reason to lower its fees, because Visa was never the one funding the settlement float in the first place; the issuer was. So even though the issuer’s working capital cost goes down with stablecoin settlement, Visa’s own costs have not changed at all, and neither has the merchant’s discount rate.

And then you should also note that Visa and Mastercard are not fighting stablecoin settlement. They are rather actively building it into their own networks. Visa has been settling in USDC on Ethereum and Solana since 2021 and is now running it at a $7 billion annualised run rate. Mastercard also acquired BVNK a few months ago to expand its own stablecoin infrastructure. They even said they “don’t see stablecoins disrupting the current payment landscape, infact they reinforce it.”

The networks are not being disrupted or replaced by stablecoins; rather, they are absorbing them as an upgrade to their own settlement layer. Every stablecoin card that runs on Visa adds transaction volume to Visa’s own network and pays Visa’s assessment fee while ironically claiming to disrupt it.

One thing I would also mention is that in our Stripe piece, we did argue that stablecoins are actually attacking cross-border payment costs because they remove the chain of intermediary banks entirely, and there’s nothing misleading about that claim. But the whole difference is that one is a cross-border FX problem, and the other is a domestic network fee. The industry took legitimate insights about cross-border payments, where stablecoins genuinely save money, and then also stretched it to cover domestic consumer spending, where the economics are completely different.

The cross-border thesis is strong, but the domestic one falls apart if you look at where the fees actually come from.

So what do stablecoins ultimately replace? Honestly, the answer can be settlement and cross-border FX. But what is being replaced exactly is the plumbing behind the transaction. And the plumbing behind a domestic card swipe was never really where the money was anyway.
That’s all for today!

Vaidik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Before this, I had read a lot of articles and understood nothing. Distributed ledger, Immutable, I knew, but I could not have told you what any of it meant if you had pushed me even slightly.

Then I read this in 2021; I was like, woah, blockchain can save the world, dammit!

Still a believer, for what it counts. But I now recognise that financialised systems are easily co-opted by major capital. Achieving decentralisation requires more than mere technical solutions.

This is how he describes decentralisation:

I saw the beauty of the thing if it works, and reading it the first time felt like freedom. Maybe because it involved birds, but yeah.

Stephen P. Williams is a journalist, not a tech guy. In 2015, he read a piece about coltan mining in the Congo, the mineral in every smartphone. It mentioned that blockchain could track it from the forest to the factory. The selling point was that a buyer in America would know their phone was clean. Williams thought about the miner instead, who hands over what he dug and never finds out where it went or what it sold for. He went down the hole and came out with this book - Blockchain: The Next Everything

It’s 2026 now, and I've been writing about this for a living since 2021, so I can finally mark his homework.

It helps to remember what 2019 felt like. The 2017 bubble had popped, everyone was embarrassed about bitcoin, and the respectable line was that the coin barely mattered and the ledger underneath it was the prize. Big companies were standing up a blockchain team.

Williams opens the second part of the book with a Gartner forecast. Blockchain would add $176 billion in business value by 2025, and pass $3.1 trillion by 2030.

That $176 billion can never be graded. It isn’t revenue or sales, which you could count. Business value-add is Gartner’s own private measure, and no company reports it. So look at Gartner instead. Every year Gartner publishes a chart ranking new technologies from overhyped to genuinely useful. In July 2024, they filed most blockchain projects in the disappointing zone, then said they might stop making the chart for blockchain at all, because too few people were still asking about it. One of their analysts, Adrian Leow, said the real payoff was maybe five years away. People had been saying five years away since 2016.

Williams quoted the most trusted research firm in the business, same as we all do to date.

Williams' big supply chain example is Maersk and IBM. IBM told him that shipping one container of flowers from East Africa to Europe produces a couple hundred paper documents. If you put everyone on one shared record, the paperwork goes away. Every company was citing this example in 2019. The product was called TradeLens. Maersk and IBM shut it down in 2023. Maersk said the system needed the whole industry to join; not enough companies did, and the business never paid for itself. Five of the six largest ocean carriers had signed up. Still not enough. Rival shipping companies would not share their data with each other.

Voting is the second one. In May 2018, a few dozen West Virginia soldiers posted overseas voted from their phones using an app called Voatz. Williams calls it a promising test and lists what blockchain voting would fix, including his claim that miscounts become impossible. In early 2020, MIT researchers audited the app and found an attacker could change a vote, block a vote, or find out how someone voted. West Virginia dropped it and went back to paper.

Williams calls the ledger unhackable. Now that I think about it, it’s a dangerous word to put in a beginner’s book. Technically, that’s true for the core Bitcoin network; no one has managed to hack the database itself. But hackers still steal billions of dollars by targeting the exchanges, apps, and websites connected to it. In the first half of 2026 alone, security firms logged nearly $1 billion stolen across more than 200 attacks. So much for unhackable. Every other way of losing money is still open. Someone gets into your account, or the exchange holding your coins collapses, or you sign a transaction you didn’t understand, or a stranger talks you into sending it.

Holding on to the optimism it gave me back then, there’s a chapter in the middle about the Internet of Things (IoT) that fascinated me the most.

His idea is that as more objects get connected- a fridge, a gas meter, an electric car- they can hold a blockchain too. And once they can hold one, they can pay each other. Your fridge notices you’re low on milk and orders it. Your car’s battery has extra power stored and sells it back to the grid while you sleep. No human involved in any of it. He calls all these connected devices linking up the Milky Way of computers, which is a lovely way to put it.

Money that moves without a person deciding, every time, in tiny amounts, between machines. It made the whole rest of the technology feel small. But fridge ordering milk did happen, though.

Amazon has run Dash Replenishment since 2016, where connected devices automatically reorder their supplies, like washing machines that order your detergent, or printers that order your ink. On fridges specifically, Amazon Your Essentials and Alexa Smart Reorders are now available on Samsung Family Hub refrigerators, and LG’s model automates orders via Instacart or Amazon Fresh. Samsung shipped an update in May 2026 where the fridge proactively alerts users when frequently consumed items need to be replenished. None of this needed a blockchain.

But the idea still materialised in an unexpected way. AI agents are now paying other machines directly. Coinbase created the x402 standard, which repurposes HTTP 402 (”Payment Required”), an old web status code that had gone unused for decades. Under this system, when an AI agent requests data from a server, the server quotes a small price, the agent pays instantly in stablecoins, and the server delivers the data. Coinbase transferred governance of the standard to the Linux Foundation. Major companies like Google, Visa, Mastercard, AWS, Cloudflare, and Stripe are now working with the same concept.

AI agents have made roughly 176 million of these payments totalling $73 million. This means the average transfer is just 40 cents, and most are under 10 cents. That’s why this has to run on crypto rails. Visa charges around thirty cents just to process a transaction, so it can’t profitably move three cents. On a crypto layer-2, sending that same three cents costs a fraction of a penny.

Near the end, he writes about visiting refugees in Rwanda, and wonders if one day the ID someone gets entering a camp might sit on a blockchain and still mean something when they leave. Well, we know that out of this entire book, that soft little hope is the thing that scaled.

It’s called Building Blocks, and the World Food Programme runs it. Started in 2017 with a hundred people in Pakistan. It has now supported around six million people, moved more than $760 million in aid across over 40 million transactions, and saved more than $3.5 million in bank fees by cutting the banks out. In Jordan, a Syrian refugee buys food by looking into an iris scanner. It’s a database that cuts bank fees, run by an organisation that badly needed to cut bank fees.

Similarly, Bhutan became the first nation to run self-sovereign digital identity on Ethereum, giving 800,000 citizens total control over their personal data.

Read: The Fortress and the Phone - by Thejaswini M A

Williams argues bitcoin was only the first thing anyone built with this. The useful part is the shared record itself, and it doesn’t have to hold money. It could hold anything people need to agree on. Who owns which piece of land. Which diamond came from where. Whether your degree is real. Who won an election.

Half right. Half no.

The ledger got freed from money and mostly lost the demand there. TradeLens died. Australia’s stock exchange was trying to replace its settlement system with a distributed ledger and wrote off up to A$255 million. Companies kept building chains anyway. Ondo killed its own in July. dYdX had 73% of decentralised perps volume in 2023 and now has under 3%. Swellchain, Mint and Polygon’s zkEVM all shut this year.

Money went the other way. Stablecoins market cap passed $300 billion. On Base alone, stablecoin volume is up sevenfold in a year. BlackRock and JPMorgan tokenised Treasury bills.

Shared databases fail when direct competitors refuse to share their data, like the ships, but they succeed when neutral organisations use them to coordinate without anyone taking control. We can say that, but it’s seventeen years of evidence on a technology that will be around for a lot longer, so I’d hold it loosely. I still think blockchain grows out of most of these constraints.

I’ve been hard on a book I love, so let me end straight. If I could send a note to the person who read this in 2021, I wouldn’t say the book was wrong. I’d tell her the instinct was right and the examples were not, and that she should have let go of the examples years earlier than she did. It took me a long time to stop assuming a good idea wins because it’s good. Williams assumes that on every page. It makes the book generous, and it’s why so much of it has aged badly.

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

“When the capital development of a country becomes a by-product of the activities of a casino, the job is likely to be ill-done.”

— John Maynard Keynes, 1936

Hello,

The crypto industry receives a lot of flak for being heavily driven by speculation. Some call it a casino devoid of any fundamentals. But what they don’t realise is that speculation often precedes market evolution. A room full of speculating people that respectable people disapprove of often assembles the liquidity that respectable things later get built on.

In today’s piece, I explore how this pipeline is evident in what we’re seeing in the crypto markets lately: a memecoin launchpad on a chain built for stocks, a perps venue turning into commodities infrastructure, a broker whose frog tokens might be funding its equity rails.

On to the story…

Speculation has many names in crypto: noise, froth and casino are some of them. What speculation is almost never considered is its role as a foundation for what gets built on top of it in a maturing industry.

But let’s be clear: I don’t think all froth becomes foundation. Froth that stays froth for an extended period dies fast. But the froth that finds products that can feed consistent activity can become a floor on top of which other things are built.

This is not a recent phenomenon. We have seen it happen in the evolutionary journey of finance.

Take the case of Chicago grains from the 1840s. The Chicago Board of Trade’s futures market was built so that farmers exposed to an unpredictable harvest could lock in a price months ahead and minimise losses. But for every farmer wanting to sell their crop forward, you also needed someone willing to buy it forward and carry the risk. Speculators filled that gap. The risk-transfer market only worked because speculative capital showed up to absorb the risk these farmers wanted to offload.

It’s the delinking of grain, a physical object, from the grain, a saleable commodity, that made the futures market possible. American historian William Cronon called this the abstraction of grain. Once the trade involving grains was reduced to a receipt, it could change hands easily and made speculators trade it ubiquitously. Their trading made the market liquid enough for a farmer to find a counterparty. The deep liquidity turned Chicago into the price-setting capital of wheat.

What’s interesting is that many hated and resisted this at the time. The Granger movement campaigned against Board of Trade speculators as parasites profiting off the farmer’s labour. Yet that mechanism became the price-discovery infrastructure the entire global agricultural economy now runs on. Without the speculators becoming the counterparties in 1870, the global wheat prices wouldn’t have a pricing infrastructure today.

English economist John Maynard Keynes is often credited for his thoughts on speculation. But his comprehensive thoughts further establish the role of speculation. Keynes split market activity into two parts: enterprise, which forecasts the yield of an asset over its life, and speculation, which forecasts what the crowd will do next. He feared that liquid markets let the second swamp the first.

But when you read deeper into his arguments, they expose a dual nature of speculation.

“Speculators may do no harm as bubbles on a steady stream of enterprise. But the position is serious when enterprise becomes the bubble on a whirlpool of speculation. When the capital development of a country becomes a by-product of the activities of a casino, the job is likely to be ill-done.”

Speculation could be dangerous when it floats free of any underlying, but the same speculation can become useful when it’s tethered to interesting underlying assets.

Even in Joseph de la Vega’s Confusión de Confusiones (1688), the first book ever written about a stock market, the author credited the Amsterdam exchange with drawing both investors and gamblers in equal measure. Most exchanges that exist today began the same way, but the story got diluted with time in a way that the serious purpose - purportedly investing - seemed like it came first. Except that it didn’t.

Back to the Future

On August 5, Uniswap, the largest decentralised exchange in crypto, shipped ‘Pools’ launchpad. The platform allowed people to launch and trade memecoins on Robinhood Chain. The namesake brokerage had primarily built the blockchain for its 30 million funded accounts to trade tokenised stocks.

That’s not all. The story gets more interesting when you realise that even before the launchpad officially opened, traders found the unreleased smart contracts and pushed more than $150 million through them. This forced Uniswap to support both the test and final versions at once and delay the launch. The top token, FRONG, which was named after the title of the frog video Uniswap used to tease the product, was minted six days early by the same contracts and carried 12,141 holders by opening day.

Make of all this what you can: About $150 million dollars moved through infrastructure that did not officially exist yet, to trade a token about a frog, on rails built for equities. Today, FRONG is the unofficial mascot for Pools. It’s difficult not to read the move as planned, to drive liquidity onto Uniswap’s newest v4 infrastructure. Even if it wasn’t planned, the volume on Uniswap v4 on the chain almost tripled from $86.2 million to $228.3 million in a single day.

But as I said at the beginning of this piece, I don’t think all froth and speculation help in building something sustainable. Instead, whether a speculative business survives depends on what it speculates about. Consider the case of pump.fun.

Over the years, pump.fun has sold speculation as its main product. Today it is the largest platform to launch memecoins. About 7 in 10 memecoins launched on its platform don’t last more than a day. Only one lasts longer than a month.

Its protocol fees between January and July 2026 almost halved from $420 million in the same period last year. Yet, it is one of the largest revenue-making protocols in crypto. It made $620 million in revenue and $584 million in earnings last year.

Another one from that league is Hyperliquid. It began as a platform that allowed crypto trading at high leverage. If you wanted to place directional bets, this was the venue that let you do it. Then it latched that leverage to assets that trade continuously and had nothing to do with crypto. Through its HIP-3 framework, it now runs perpetual markets on Nvidia, Tesla, a Nasdaq tracker, gold, oil, silver, stock indices and more. A trader can take a position in crude oil the moment news breaks on a Sunday, while the traditional market sits closed until Monday.

Read: Hyperliquid: House of Finance 🏦

In early July, real-world-asset perps out-traded crypto on the platform for the first time ever with 52% of volume.

Although Hyperliquid’s overall volume roughly halved from its 2025 peak, the RWA growth offset the decline in crypto pairs. This is what speculation can create. Hyperliquid translated the speculative leverage that fuelled crypto natives and redirected it onto gold, pre-IPO stocks and equity indices. What it created is a 24/7 pricing layer that traditional venues are now imitating.

We see Robinhood Chain follow the same strategy in real time.

Although the leadership team had dubbed the chain as the foundation for trading on-chain stocks, CEO Vlad Tenev embraced the attention the chain was getting from memecoin traders.

The chain then overtook Base on daily active users three weeks after launch, driven by memecoin trading. The memecoin volume is the seed capital that will eventually graduate into the equity rails.

None of this means speculation always graduates upward. Graduation is a direction the venue chooses.

Robinhood’s distribution powerhouse of 30 million funded users and more than a dozen businesses can help it graduate the traction it is witnessing in memecoin trading into equity rails. We saw how its prediction-markets revenue surged more than tenfold year over year in the second quarter to $156 million, 20% of its trading revenue.

Read: Robinhood’s Bigger Baskets

The Shape of Speculation

This is the part about speculation that many don’t understand. Speculation isn’t good or bad. When you strip it to its bare principles, speculation is just one of humans’ oldest impulses.

Speculation is the natural outcome when humans financialise their opinions.

Read: Built for Humans 🎰

In its rawest form, it is liquidity that is looking for something to price, and it will price whatever you point it at. It could be a bag of wheat, an Nvidia share, a frog video titled frog.mp4, the outcome of a football game or the odds of a presidential candidate.

It’s nature of the underlying asset that disciplines and directs the speculation’s trajectory. Once you take the underlying away, the same money can lead to nothing substantial or sustained. This trend has been recurring across the evolution of finance. Crypto is now part of that journey too.

Hyperliquid tethered leverage to gold and built a pricing layer in less than two years. Robinhood is capitalising on the unexpected traction it is witnessing for speculation on memecoins while it builds on-chain rails for trading equity wrappers.

Speculation will continue to build whatever the ground underneath will support.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Every once in a while, someone notices that some currency is paying 8% while the dollar is paying 4%, and asks why not just hold the 8% one and pocket the difference. It sounds like free money, or arbitrage - if you will. That’s how carry trade works. But it isn’t that simple. Although the spread appears practical to pocket, the financing cost, including borrowing, hedging, and moving such money, may often eat away at the spread. Mature players in the markets have known this for a century. The high yield and the cost of using it move together in a way that almost entirely closes the gap you were chasing.

The same logic is being applied to tokenised Treasury bills in DeFi right now. People question why the funds are sitting passively in DeFi and not being used to earn more yield. There’s a fundamental problem in that thinking.

The market often focuses on the 3% to 4% yield sitting safely on-chain. But a yield does not become useful simply because it has been tokenised. To loop the asset, a trader must borrow stablecoins against it. And often, the cost of that borrowing can be equal to or higher than the Treasury yield itself. Once the spread turns negative, leverage works against you. The asset may still be a perfectly sensible investment to hold, but there is little economic reason to use it inside DeFi.

That’s why $35 billion can sit on-chain doing nothing, and yet none of it is a missed opportunity.

Alessandro, who builds the tooling to move assets into these on-chain vaults at Li.Fi, sees this problem from the inside. In today’s guest essay, he takes the idle-90% figure everyone obsesses over and walks through the actual economics around why the money sitting still isn’t stuck or locked out.

On to Alessandro’s story,
Prathik

Follow Alessandro on X

BlackRock’s BUIDL holds $2.6 billion of tokenised Treasuries. Less than 1% of it sits anywhere DeFi can touch. Zoom out and the picture is the same. Tokenised real-world assets crossed $35 billion in July, an all-time high, while open DeFi lending puts to work about $2.5 billion. Treasuries and money market funds, the largest slice of the market, show 5.5% utilisation. Tokenised stocks are at 2.9%. The rest sits in wallets doing nothing.

Measured by issuance, institutions have arrived. Larry Fink has told investors that every stock, bond and fund can eventually live on-chain, and BlackRock, J.P. Morgan, Apollo, Janus Henderson and Circle all have tokenised funds on public networks today.

However, measured by usage, almost nothing has happened.

Ask why, and you get the same three answers: KYC restrictions, slow redemptions, transfer-restricted tokens. While true, they describe the fence, not the reason so few people try to climb it. The better answer might be an economic one. For tokenised Treasuries in particular, there is currently no levered trade. The 90% that sits idle is not stuck. It is capital making the rational choice. And part of the answer, which we will get to, is that the yield already reaches users through a different door.

The Carry Theorem

Start with what DeFi would actually do with a yield-bearing token. The native move is looping. You deposit the asset as collateral, borrow stablecoins against it, buy more of the asset, and repeat. At every given LTV, the maximum exposure is 1/(1-LTV), so an 80% LTV gives you 5x. This is how stETH grew, how Ethena’s USDe reached more than $14 billion at its peak, and how every yield asset in DeFi gets levered.

The net return of a loop is the asset yield plus (leverage minus one) times the spread between asset yield and borrow cost. In plain terms, leverage multiplies the spread, not the yield. Everything depends on that spread.

Here is the problem for Treasuries. A tokenised T-bill pays the “risk-free” dollar rate. The person lending you stablecoins on Aave can reach that yield too, directly by clearing the token issuer’s KYC, or through a plain brokerage account. So, they face a choice: lend stables to you, or buy T-bills. Reaching the bill has a cost: clearing KYC, minimums, moving money off-chain, so lenders will tolerate earning a bit less than the bill. However, when Aave’s lending rate sinks too far below the bill, they pick the bill and stop adding deposits. The borrowing rate is always higher than what lenders earn, because the protocol takes a spread in between. Put those two together, and the bill yield ends up inside Aave’s spread: above what lenders earn, below what you pay to finance it by borrowing stables.

Your financing cost ends up above the bill rate for long stretches. The asset you want to loop pays that same bill rate. You are financing the risk-free rate at the risk-free rate, or worse.

In 2024, Aave v3 USDC borrow averaged 9.0% while bills paid about 5.3%. In 2025, it averaged 5.8%. As of mid-July 2026, borrow sits at 3.97% against a tokenised Treasury average yield of 3.46%. Run the loop today at 80% LTV and five turns: 3.46 + 4 x (3.46 - 3.97) = 1.42%. Leverage points down. You would earn less than if you simply held the token unlevered.

The squeeze is not airtight. As I write, Aave’s USDT borrow rate sits a few tenths below the bill yield, and roughly $700 million of leverage against RWAs has formed across Aave, Morpho and Kamino. But look at when the spread flips. On-chain dollar rates only fall below TradFi rates when demand for crypto leverage collapses, so T-bill collateral activates exactly when nobody wants dollar leverage. And a spread hovering around zero, after collateral haircuts, fees and rate volatility, still never pays at size.

An economic working paper (Barbon, Barthélemy and Nguyen, 2023) found that stablecoin lending rates are driven mostly by crypto demand and reconnect with conventional rates only slowly, through exactly this arbitrage channel. That slow reconnection is the 2024 premium visible in the numbers above. A Keyrock and Securitize report showed that tokenised Treasuries out-yielded DeFi’s benchmark stablecoin lending rate on 98% of days in Q1 2026, with 3.6 times lower volatility. Read that carefully. The tokenised T-bill is a better hold than lending on Aave, and simultaneously impossible to lever. The bill pays more than Aave lenders earn and less than Aave borrowers pay.

So the idle 90% is not waiting for a better front-end. It is capital declining a negative-carry trade.

Follow Alessandro on X

Utilisation Follows Carry

If the carry explanation is right, it should predict which RWA categories DeFi actually uses. It does, across all three classes.

Private credit yielded 8 to 10% when most of these integrations were built, so the deployed capital arrived when the carry was wide.

The obvious objection is that carry and transferability are tangled together, since the high-yield categories happen to be freely transferable tokens while Treasuries are dominated by whitelisted ones. Class-level data can’t separate the two. But specific comparisons where transferability and carry pull in opposite directions can separate them, and those comparisons are unambiguous. Apollo’s ACRED is a transfer-restricted, qualified-purchaser-only tokenised credit fund yielding around 8 to 9%. It became the flagship DeFi looping asset last year. Securitize built a wrapper, Gauntlet automated the strategy, and KYC-gated markets appeared on EVM chains and Solana to capture the spread. Meanwhile, Ondo’s USDY is freely transferable, holds over a billion dollars, and has essentially no leverage footprint. Its carry is negative.

Restrictions are a fixed cost. Wide carry pays the cost. Negative carry means there is nothing to pay for it with. Better access tooling shrinks the cost, but it cannot create the spread.

Three honesty notes on Table 1. First, the 46.7% private credit figure excludes Figure’s HELOC book, which is tokenised but not composable; include it and utilisation drops to roughly 12 to 15%. Second, the largest ‘private credit’ asset in DeFi is Maple’s syrupUSDC, which is overcollateralised lending to crypto trading firms, so real-world credit in DeFi is smaller than the headline suggests. Third, the reinsurance row is two protocols designed as DeFi-native products from day one, so read it as directional rather than statistical. One scoping choice matters too: these figures count open, permissionless lending. Looser definitions that count every RWA deposit in any DeFi protocol reach about $7.4 billion for Q2 2026, up 200% in a year. The level moves with the definition. The ranking across asset classes, which is where the story lives, does not.

TradFi Leverages the Same Asset at 50x

Here is where it gets interesting, because the no-demand story is wrong in one specific way. Demand for levered Treasuries is enormous. It just lives somewhere else with better infrastructure.

In TradFi, fixed income stands on three legs: financing, rate hedging, and shorting. DeFi has built none of the three in a permissionless form.

Financing. The US repo market carries ~$12.6 trillion in outstanding exposure. Overnight general collateral repo currently runs 16 to 29 basis points below the 3-month bill yield, so right now TradFi finances the risk-free asset below its own yield. That sign flips with the rate cycle. The structural part never flips: financing in trillion-dollar size, at haircuts near zero, meaning lenders demand almost no extra collateral at a rate that hugs the risk-free curve. DeFi has never had any of the three. That deep, nearly frictionless financing is why the basis trade can exist at all. Hedge funds ran about $830 billion of these positions as of late 2025, long cash Treasuries against short futures, at 50 to 70x leverage on the futures leg. DeFi’s version of financing is variable-rate, priced above the asset yield, and capped at 80 to 92% LTV for a maximum of 5 to 12.5x, with no term structure and no netting.

Rate hedging. Outstanding interest-rate derivatives total roughly $668 trillion. This is the machinery that lets insurers and pensions run fixed-income books against liabilities that stretch decades longer than the bonds they can buy. UK pension funds at their peak hedged about £1.4 trillion of liabilities through strategies built on swaps and repo. DeFi’s equivalent is a graveyard. Element, Yield Protocol, Sense and Voltz were a whole generation of on-chain rate protocols, and all of them wound down. Pendle survives, and its Boros venue for funding-rate swaps has done $18 billion of cumulative notional in eleven months. That is real, but still tiny. Boros open interest peaked above $250 million, about 0.17% of the roughly $150 billion sitting in funding-rate positions on an average day.

Shorting. Securities lending has $4.4 trillion on loan and produced a record $9.1 billion of revenue in the first half of 2026. There is no native on-chain mechanism to short a tokenised Treasury. No shorting means no two-sided price discovery and no special markets.

Duration. The quieter structural problem sits underneath all three. DeFi liquidation assumes the collateral can be sold within the same transaction. Tokenised funds redeem on a cycle; next-day for some funds, quarterly for others. A liquidator who seizes that collateral cannot sell it in the same transaction. Someone has to hold the position until the fund pays out. You can see protocols contorting around this in production. Every serious RWA collateral listing since 2025 has gone to isolated or permissioned venues rather than the main pools. Morpho hosts them in isolated markets, Kamino keeps them in their own markets, and Aave built Horizon as a separate permissioned instance rather than touching its core pool.

Strip those legs from any market, and the rational position size is close to zero. TradFi investors are not avoiding on-chain Treasuries out of ignorance. They are declining to trade a market with no repo desk, no swaps desk and no borrow desk.

The On-chain Buyer and Cashflows

Even if the infrastructure existed, there would still be a demand-side problem, because the people actually trading on-chain today don’t want cashflows. They want price exposure with leverage.

The cleanest evidence is tokenised stocks. Spot tokenised equities on Solana, which handles 96% of all tokenised stock trading, did $4.84 billion in volume in the entire second quarter of 2026. Hyperliquid’s HIP-3 builder markets, led by equity perps, did about $62 billion in May alone, and by mid-July, these markets were roughly half of all Hyperliquid perp volume, up from 2% in January. That is a revealed preference of nearly 40 to 1 for synthetic price exposure over owning the tokenised asset.

Dividends make the same point from the other side. None of the three major tokenised stock issuers pays dividends out as an on-chain cashflow. xStocks rebases dividends into your balance, Ondo Global Markets accrues total return into the token, and Robinhood’s stock tokens are legally derivatives, with the newer on-chain versions structured as debt instruments that pay dividends as dollar credits inside the app. There is nothing for a DeFi protocol to compose with. The main lending venue that accepts tokenised stocks as collateral, Kamino on Solana, holds roughly $20 to 30 million against them. All tokenised stocks in DeFi combined: $78 million.

Pendle, DeFi’s fixed-rate venue, tells the same story with better resolution. The rails demonstrably work. Pendle settled $58 billion in fixed yield in 2025, and its principal tokens are functionally zero-coupon bonds. But the demand that used those rails was leveraged, points-subsidised Ethena carry, and when that spread died in September 2025, Pendle’s TVL fell 92%. RWA underlyings on Pendle reached all of $151 million, about half a percentage point of the tokenised RWA market. Tokenised Treasuries don’t need a yield-splitting venue. They need a repo market that doesn’t exist.

The Yield Travels Better than the Asset

There is one more twist, and it changes how you should read the headline number. The T-bill yield is already flowing through DeFi at scale. It just doesn’t travel as the fund token.

Look at who actually holds the big tokenised Treasury products. Ethena’s USDtb stablecoin keeps over 90% of its reserves in BUIDL. Ondo’s OUSG is reported to be BUIDL’s largest single holder. Sky (ex-MakerDAO) allocated $2.4 billion of its reserves into BUIDL, Superstate and Centrifuge funds through Spark, and another billion into Janus Henderson’s CLO fund through Grove. Circle’s USYC, which overtook BUIDL in March 2026 and now sits at $2.9 billion as the largest tokenised Treasury product, grew almost entirely as off-exchange margin collateral for Binance institutional clients. BUIDL itself has been accepted as derivatives margin on Crypto.com and Deribit since June 2025.

Every name on the above list is a wrapper, a DAO treasury, or an exchange collateral program. Not one of those use cases shows up as “DeFi utilisation” of the underlying token, yet the yield reaches users anyway, through sUSDS ($6.2 billion earning the Sky Savings Rate), through USDY at around 3.5%, and through USDtb rewards. Yield-bearing stablecoins drove more than half of net stablecoin supply growth in Q1 2026.

This is not a failure mode. It is how money has always been structured. Reserves sit still while claims on them circulate. Tokenised Treasuries are settling into the same role: the reserve layer of onchain dollars, and reserve layers do not circulate. This means utilisation was partly the wrong metric from the start. It measures whether the base moves, when the whole point of a base is to hold still while everything above it moves.

Regulation is actively pumping this structure. The GENIUS Act, signed in July 2025, prohibits payment stablecoin issuers from paying any form of yield. So the issuer cannot legally pay the T-bill return to its own holders, and the yield migrates into wrapper tokens and rewards paid through affiliated entities. Regulators are already moving on the rewards door (the OCC proposed in February 2026 to treat coordinated issuer and affiliate rewards as presumptively prohibited), which leaves the wrapper as the structure regulation durably favours. The 90% idle statistic partly measures the wrong layer: the yield composes, the asset doesn’t.

Will It Change?

Two sides are attacking this problem from opposite directions, and they disagree about what the problem is.

The first one is institutional and rebuilds the missing TradFi legs on-chain in permissioned form. Tokenised repo already runs at scale inside banks. Broadridge’s distributed ledger repo platform processed $7.5 trillion in June 2026 alone, and JPMorgan’s Kinexys has processed roughly $3 trillion of tokenised repo since its 2020 launch. Closer to DeFi, Grove’s Basin facility offers up to $1 billion a day of instant liquidity against BUIDL and JTRSY redemptions, which is functionally an on-demand repo window for tokenised funds. Aave’s Horizon holds about $255 million. On the access side, before capital enters the vault, Plume’s Nest Vaults fold compliance into the transaction itself: instead of a static wallet allowlist, a signed policy attestation checks eligibility the moment capital enters a vault, making compliant access far easier to compose with. The infrastructure is arriving. It is arriving wearing a suit.

The second side is crypto-native and changes what gets tokenised. A tokenised T-bill is a worse product than the brokerage version, as it delivers the same asset with extra smart contract risk, extra fees, and a buyer base that never asked for it. Plume’s answer is not to choose between the two sides, but to connect them: use institutional-grade vault and compliance infrastructure to distribute assets whose yield can actually clear the on-chain financing floor. nOPAL is the clearest example. The Plume Nest vault behind it allocates mainly to BlackOpal’s short-dated credit-card receivables strategy, holding roughly $35 million in TVL at an 11.8% trailing 30-day APY. The Morpho market financing nOPAL collateral currently charges roughly 4.9%, with about $814,000 borrowed against it, a positive gross carry before fees, haircuts and rate movement. Plume, in other words, is side-one plumbing carrying a side-two asset. While both agree the trade has to earn a positive spread, Plume’s model is to work both sides of it.

My watchlist for when the idle 90% actually moves:

1. Someone offers permissionless dollar financing below the bill yield against tokenised Treasury collateral. That is the on-chain repo moment.

2. A liquid rate market emerges for RWA yields, so duration can be hedged rather than just held.

3. A borrowing mechanism makes tokenised funds shortable.

4. Async redemption standards make slow-to-redeem collateral possible to liquidate without 35-day windows.

Financing has permissioned prototypes today, and the async-redemption standard is live and spreading. Rates and shorting have nothing yet.

At LI.FI I work on Composer, the orchestration engine behind our DeFi and RWA Earn product. It handles the access leg of this problem, moving any token on any chain into a KYC-gated RWA vault in one flow with the compliance check included. Access decides who can get into an asset. Carry decides whether there is a trade once they hold it, and no router can fix a negative spread.

Today’s idle 90% is a carry problem. But look at what happened with ACRED. The moment carry turned wide, wrappers, KYC’d markets and automated strategies showed up within months, because restrictions are a fixed cost and someone has to make that cost cheap to pay. When the spread flips, the binding constraint moves back to access.

The tokenisation thesis was that putting assets on-chain makes them more useful. For an investor with a brokerage account, that is not yet true, because a tokenised T-bill today carries more risk than its off-chain twin and can do less. The capital that already lives on-chain has voted for a different form. Until the financing floor breaks, the rational trade is the one the market is already making. Hold the wrapper, collect the yield, and leave the fund token alone.

And the issuers seem to know this. The fees earned on a few billion dollars of assets barely register at firms that manage trillions. So, these funds are loss leaders, sold rather than bought. It’s the price of being in position before the market structure arrives. Today’s demand was never the point.

That’s all for today.

Until next time,
Alessandro

Follow Alessandro on X

P.S.: This piece was originally published here.

P.P.S.: We will be featuring good writing and writers we love from time to time. If you have recommendations, send them our way.

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000 subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Difficult because of facing the permanent loss of the path not taken, accepting that time is linear, and living with the grief of unchosen potential.

Wasting 20 minutes in the supermarket aisle holding a tub of protein powder, Googling the difference between isolate and concentrate while running price-per-scoop math against a 5lb bag on Amazon. All of that, and the commitment issues of picking a flavour!! I pick a Snickers bar at checkout instead, decision-making postponed.

Similarly, the obvious decisions are easy. If a project runs out of money or the tech completely crashes, the call is already made.

Time is the one thing you can’t buy more of, and every month spent building the wrong thing is a month gone. Taking it back means writing off things you already put in.

Ondo had to write off a year and a half. They had a working testnet and a tokenised Treasury settlement that JPMorgan had already run across it. That was not an easy kill, so we dig in.

On July 27, Ondo Finance released a blog post introducing the Ondo Network. It is a lot different from what they announced 17 months ago.

Ondo had announced its Ondo Chain. A layer 1 for tokenised real-world assets, with permissioned validators, staking for passive income, and native bridging. Big backers including Franklin Templeton, WisdomTree and Wellington. JPMorgan’s Kinexys settled tokenised Treasuries on the testnet with Chainlink.

But now, CEO Ian De Bode confirmed the company will not run both systems side by side. They scrapped the blockchain.

They pulled out the Execution and verification and kept Ethereum as the settlement layer.

Execution runs inside hardware enclaves, also called trusted execution environments. These are sealed compartments on a server chip. Code runs inside them without the server operator being able to read or modify it. Each enclave generates a hardware fingerprint of its active code. Changing a single byte alters this fingerprint and fails verification.

Verification relies on a quorum of independent attestors. Before an enclave starts, they check its fingerprint against approved code. They also hold split key shards, meaning no single operator (including Ondo) holds a full key. The key can only be reassembled inside an enclave that has verified its code.

Settlement stays on Ethereum. Asset transfers land on a public chain, the same as before.

Ondo argues that blockchains bundle four separate functions, such as consensus, replication, transparency, and final settlement of state. Consensus and replication slow a system down, while transparency exposes everything publicly. Those trade-offs are an unnecessary tax for matching orders. So Ondo moved execution off-chain and kept settlement on the blockchain.

They say: “A matching engine wants one deterministic sequencer, and replication adds latency to every order.”

I don’t know about you, but I remember when Accenture said this back in 2022.

For over spent seven years, the Australian Securities Exchange tried to replace CHESS, its clearing and settlement system. They tried replacing it with a distributed ledger from Digital Asset, even buying 8.5% of the company in 2016. The launch target moved beyond 2021 and kept getting pushed back. By late 2022, an Accenture audit found the software was only 63% complete with no launch date in sight.

It said: “Distributed systems introduce higher latency.”

ASX wrote off up to A$255 million, while ASIC (Australian Securities and Investments Commission) sued the exchange over misleading market updates as the project collapsed. Market brokers lost tens of millions more after building connections to a system that was never deployed.

The NYSE and Nasdaq use private matching engines run by single companies. They don’t copy order books across hundreds of computers, and traders don’t want their live orders visible to the whole market. Then the trade goes to DTCC, which is slow, shared by the whole industry, and settles more value than anything else on earth. DTCC’s subsidiaries processed $4.7 quadrillion in securities transactions in 2025.

They say that no blockchain today can carry those volumes. DTCC is doing tokenisation anyway because settlement only runs on weekdays. It went live with production trades on July 15 and scales up in October. The tokens layer on top of existing post-trade rails.

Crypto promised that you could merge execution and settlement into one machine and keep the good parts of both. Rollups have been kind of doing what Ondo did.

Speed is a bad way to decide whether you need a chain. Every venue wants to be fast, and most of them can buy it. The better test is to see if anyone besides you needs to write to this ledger?

Ondo runs a venue. It matches perpetual futures on Nvidia, oil and the S&P 500, contracts that track a price and settle in cash, so nobody ever receives a share. The tokenised part is the collateral. You post tokenised Treasuries as margin. They hold the order book, and no outside party needs write access to their margin engine. Putting it on a blockchain would have meant running multiple copies of a database they only need one of, exposing a public order book their market makers would hate, and managing a validator set.

Instead, without any blockchain, the platform hit $6 billion in total perps volume in three weeks, averaging over $300 million a day.

Ondo fails that test, which is why the chain made no sense for it. Some companies pass it. The difference is who they’re building for. If your users are the only people touching the ledger, you’re running a venue. If strangers build things on your ledger without asking, you’re running a market, and a market needs a chain because the whole product is the part you don’t control.

Base is a different animal. Coinbase is selling blockspace for other people to build on. That’s why the numbers Coinbase quotes are about other people’s activity. In Q2 2026, more than 90% of agentic stablecoin volume ran on Base, and 97% of onchain agentic transactions used x402, which has handled 160 million payments in the past year. Stablecoin volume on Base is up 7x year over year, at $19 trillion so far in 2026. Hardly any of that belongs to Coinbase itself.

If Coinbase moved Base’s execution off-chain into a private enclave, the product it’s selling wouldn’t exist.

Robinhood stands in an inherently awkward position, which is the middle. The Chain went live July 1. An Arbitrum Orbit rollup, Stock Tokens live in more than 120 countries. But Stock Tokens are debt instruments issued by Robinhood Assets Jersey Limited, not shares, however you hold them. That puts Robinhood on the wrong side of the SEC’s January 2026 guidance, which drew a line favouring issuer-backed tokenised stocks over third-party synthetics. Self-custody protects you from Robinhood losing your token, not from Robinhood failing.

Read: The World Is Flat, and So Is Your Claim - by Thejaswini M A

What the chain does buy Robinhood is distribution. Uniswap put its token launch platform, pools.trade, on Robinhood Chain on August 5. A place where other people’s apps land next to your 23 million users is still good business.

Companies have been building their own chains for three years, and the results don’t favour the idea.

dYdX tested the app-chain concept at scale. In late 2023, the protocol left StarkEx to launch its own Cosmos chain, placing order matching directly on a decentralised validator set. Architecturally, it won the debate, but market share was a different story.

In early 2023, dYdX commanded 73% of total decentralised perpetual futures volume. By 2026, its market share fell below 3%. The platform now processes roughly $25 billion to $30 billion in monthly volume, compared to Hyperliquid’s $180 billion to $208 billion. Its total value locked stands at $100 million to $150 million, trailing Hyperliquid’s $6.2 billion.

Talking about Hyperliquid, it is the best argument for owning your chain. Its matching engine, HyperCore, runs consensus directly. It uses 27 validators, up from five at launch, with no slashing. Instead of publishing the source code, the repo publishes the finished program. You can run it. You can’t read what it does. Joining the active validator set requires over a million HYPE.

You have 27 validators running publicly unreadable code processing $200 billion monthly, versus one enclave running audited code checked by attestors. Both models rely on a tight circle of operators and cryptography. Though calling one a “blockchain” and the other “off-chain” makes them feel different than they are.

Unichain exists to capture Uniswap’s trading volume and send sequencer fees to UNI holders. But Uniswap processes roughly $15 billion a week across existing networks, and Unichain holds only $532 million of its $5.76 billion total TVL. Besides that, Uniswap’s fee switch already burns $90 million in UNI a year without needing its own chain.

If you look at the broader landscape, out of all L2s, three hold 91% of TVL while twenty hold any TVL at all. Ethereum alone holds 65% of the L1 total, and Base holds 53% of the L2 total.

Chain closures are accelerating: Swell ended Swellchain in June, Mint Blockchain folded in April, and Polygon shut down its zkEVM sequencer on July 1. RootData logged roughly 100 dead crypto projects in early 2026, with networks representing a massive chunk of that body count.

Why is everyone launching anyway?

Until recently, apps were priced on revenue; chains were priced on potential.

When Ondo announced a chain with staking rewards, its token popped 11%. When the network actually launched, nothing changed. ONDO pays no yield, and rising protocol volume creates zero direct demand for it. Management says token utility comes later with proof-of-stake and slashing.

But enclaves are not free. Secure chip hardware took a beating this past year. Hackers repeatedly bypassed Intel and AMD security using cheap hardware. First using a $50 tool to break memory protections, then a $1,000 device to steal Intel’s master security key. Last October, researchers released TEE.fail, an attack named after the security failure of TEEs (Trusted Execution Environments, or “enclaves”). Enclaves only work because the chip promises no one can look inside. The researchers broke that guarantee using a $1,000 circuit board attached to the computer’s RAM memory, eavesdropping on plain-text data as it travelled down the wires between the memory and the processor.

They stole Intel’s signing key, the digital stamp of authenticity that proves code is running safely inside a real enclave. Using that stolen key, they created a fake safety certificate on BuilderNet (an Ethereum system for organising pending crypto transactions). This allowed them to secretly view private transaction orders while the system’s automated checks reported that everything was completely secure.

These attacks need physical access to the machine. So Ondo’s top-tier security basically boils down to physical security such as who installed the box, which warehouse it is in, and who delivered it. Is it worse than trusting 27 validators who can’t be punished? Maybe not. But it’s physical security, a real-world risk, and it should be priced like that.

Ondo has been facing the truth and plans to open attestations to bonded operators, while splitting code verification, key custody, and server hosting across separate entities. The final state gets posted onchain so outside watchers can challenge bad transactions, with proof-of-stake and slashing added later.

Might as well just build Ondo Chain….

Have a good day!

—Thejaswini

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.

Running a lending business is a two-part job. The first moves the money, using the accounts, infrastructure, and software that bring the dollar from a saver to a borrower and back. The second part is deciding whether the money should be lent in the first place. The first part is an engineering aspect, while the second involves judgment.

For over a decade, crypto has been improving the infrastructure to address the first aspect. That infrastructure has evolved well enough to replace the job that required a banking charter with a simple, tailor-made API that can be plugged and played. Yet, a lot of work remains on the second aspect of the lending process.

Both these halves are evolving in our banks’ savings accounts. When you tap “earn” inside your fintech app, your dollars pass through a smart contract that you will not even realise exists and get lent to a stranger. They will then earn interest without even knowing about how blockchain powered this entire process. All you care about is the familiar app you have been using all along.

In today’s piece, I will explain how curated vaults powered by blockchains can serve as backends to optimise your savings accounts and where they still fall short.

Vaults as Software

Almost every fintech we use wants to pay its customers a return, because that’s what stops us from leaving for a better app. A savings balance is the perfect sticky product to do this. The problem is that paying interest historically required you to become a bank, which could take years of paperwork and lobbying. The shortcut was to rent a bank’s balance sheet through a partner, or integrate with a money-market fund. Both these options are still slow and permissioned.

A curated vault that runs on blockchains removes these obstacles. But what is a vault? It is a pool of deposited assets run by a smart contract. Think of it as a traditional fund, except here, a curator instead of a fund manager is the one who decides where the pool gets lent. Every vault is wired into a tokenised vault standard, ERC-4626, which lays out how any app can plug into any vault the way it plugs into Stripe.

Aave Labs turned this into a shelf product in July 2026 with Stable Vaults. This product offered predictable stablecoin yield any business can embed without building the machinery underneath. The operator of the vaults picks the rate it offers and keeps whatever the strategy earns above it. The spread is the business model. It is a net interest margin business, except the fintech here never had to build a bank to earn it.

The recent momentum in this business shows there is enough demand.

Deposits across 60 curated vaults reached $8.9 billion in August 2026, up 33% in a year, while the rest of on-chain lending shrank 40% over the same stretch.

This tells us that capital didn’t merely pour into vaults from some other industries. It actually moved out of raw DeFi lending pools, where you deposit directly into a protocol and accept whatever yield the market pays, and into curated vaults where a curator decides how your deposit is lent.

This is evident from many front-end platforms integrating with curated vaults to offer an “earn” feature to their customers. Coinbase runs its in-app USDC lending through a Steakhouse-curated Morpho vault; Robinhood’s Earn uses the same infrastructure; Kraken’s DeFi Earn routes exchange balances into Sentora-curated vaults and has cleared $600 million in balances, surpassing 80,000 active depositors within six months. Deel, a payroll company, now sends contractor stablecoin balances into a Sentora-curated Morpho programme.

The Commodification of Vaults

Running these vaults seems like the difficult part because the smart contracts are audited and their isolated-market design seals each pool so that the risk doesn’t spread to the other pools. But none of this makes managing these vaults the difficult part. The more challenging part of the job is the judgement about whether the loan is any good in the first place. When you strip away the contracts, a curator is the one who does the unenviable job similar to what a fundhouse manager does. They pick which loans to make, set the exposure caps, rebalance the portfolio and take a fee. While the software that moves money is commoditised, thanks to the curated vaults, it’s the judgement of curators that is still valued. This is evident in how the market pays for each part of lending.

Anyone can deploy a vault in a day because the code is open. Yet the top three curators — Steakhouse, Sentora and Gauntlet — account for $6.8 billion, which is over 75% of all curated deposits. The fourth-largest curator, K3 Capital, manages a whole billion dollars less than the third with ~$400 million in its vaults. If the vault were the moat, deposits would spread across everyone who can deploy one. Instead, they keep piling into vaults managed by three players. This tells us that deposits chase the manager and not the vaults.

In an increasingly commoditising world, what’s rare is the most valuable. Judgement remains the most valued here, just as it is elsewhere. We have reflected similarly on countless occasions in our editorial calls at The Token Dispatch too: ‘What separates what we write from what our competitors do?’ ‘How are we writing a different story about the same half a dozen protocols, chains and projects?’ We come back to the same answer for all these questions: human judgement.

Consumers stop paying premiums for services offered by multiple sellers with minimal difference in quality. The premium is reserved for the judgement that’s rare and unique to one or two such sellers.

This is reflected in how managers of fundhouses and curators of these vaults get paid.

The top three curators generated over $12 million cumulatively in fees over 30 days, of which they retained only $880,000 (averaging 7.3% conversion) as revenue. This begs a simple question: Why would a firm managing a couple of billion in funds offer the vault for such low margins? The answer is where the real value is accruing in this ecosystem.

These vault providers charge almost nothing to run these vaults so they can become the embedded default inside platforms like Coinbase or Robinhood. Curators underbid each other to serve these platforms because once an app integrates its earn feature with a vault in its backend, letting them go to save a few basis points means re-integrating with a new partner, re-auditing their systems and re-underwriting their model. So, forgoing the fee on selling vaults as a product buys an expensive seat at the backend that is difficult to replace. Once this is achieved, there will be little resistance to paying for their credit judgment.

But there’s a problem here. Safe lending doesn’t reward the vaults that are fastest to integrate and offer the best price, but those who underwrite the best. We saw others who did this shut their shops.

In November 2025, a project called Stream Finance came apart. It all started when the company’s curators accepted xUSD as safe collateral across DeFi vaults. Except that it wasn’t. An outside manager running the strategy behind it lost about $93 million worth of assets, triggering a rapid sell-off and causing xUSD to lose its dollar peg. The stablecoin became worth less than a quarter of a dollar.

The isolated-market design did its job and stopped the damage from spreading to other vaults. But isolation only contains a bad decision and cannot undo the original mistake of accepting xUSD as safe collateral. That call had already been made, and money had already been lent out against it. So the loss was borne by the depositors and not the curators who had approved the collateral in the first place.

What happened with Stream Finance is the consequence of renting a risk engine with no credit desk attached. A vault is software that can only promise to isolate our risk. How our money is managed still needs to be promised by a human curator.

So, somebody has to be right about the borrower before the money leaves, and it’s this part where the utmost value currently accrues. This is reshaping how certain players do business in this space.

The Opportunity

Over the past three years, Qiro Finance went from tokenised marketplace to underwriter for hire to vault operator. It started as a marketplace, then realised the scarce input there was not deal flow but judgment. So, it put its judgment up for hire. It underwrote other platforms’ deals with its name on the report, in public, where a single bad call would have ended its business. Huma Finance, a payments network that has moved billions without a credit default, partnered with Qiro in January 2026 to underwrite up to $250 million in deals.

Take its public report on mGLOBAL, a token backed by a $1.1 billion fund run by Fasanara. Qiro’s review put the exposure right at the top, where it disclosed that the fund’s loan defaults had more than doubled since 2022, and the cost of protecting against currency swings was eating up nearly a fifth of what the fund earned.

This is precisely what a good credit desk is expected to do.

By July 2026, Qiro had underwritten more than $50 million across payment financing, trade finance and fintech lending with zero defaults, and it now runs its own vaults. But it lends only to borrowers it sources and judges itself. The vault software is the same commodity everyone else uses, but the work that goes behind it is where Qiro contrasts with its competitors.

Most players in the lending chain get paid whether or not the loan is recovered. A curator earns their cut on the deposits they gather. A ratings firm earns its buck for its opinion and walks away long before the loan matures. But Qiro’s fortunes are tied to the health of the loans it underwrites. If a loan is defaulted on, it impacts Qiro’s income in the same quarter. This model may not make much business sense. But the skin in the game Qiro has is what would make investors trust it to allocate their money.

The scale of Qiro’s business is still tiny, with $50 million underwritten so far against the roughly $9 billion parked in curated vaults. Running the underwriter and the lender under one roof also poses a problem, precisely the one that a credit desk is supposed to remove. Yet, Qiro has picked up the right fight by vying for a place where the value accrues.

Renting these curated vaults as yield engines is a great start. Especially for fintechs like neobanks, which can offer a savings rate today with no charter, no partner bank and no fund. Traditional finance was structurally never able to offer such an integration. But for curated vaults to become the perfect backend for your savings, offering a plug-and-play solution isn’t enough. A lot will depend on who is underwriting the risk behind the strategies that run through these vaults.

That’s it for today. I will be back with the next one.

Until next time, stay curious,
Prathik

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 170,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌

📩 Fill out this form to submit your details and book a meeting with us directly.

Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.