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.

The first thing I got reminded of when I read today’s guest essay for the first time was what insurance reveals about human behaviour. Ask most people what insurance they have, and chances are high you might hear phone insurance. Not health cover or a safety net to cover income for dependents. But phone insurance was offered at checkout for a few hundred rupees, and opting for it required the least decision-making.

This reveals how we make decisions around risk. Nobody sat down, listed what could go wrong, ranked it by damage, and decided that their phone screen should be at the top of the list. The risk coverage was offered by default, and the protection got shaped around what was easy to sell. Little did the risk exposure itself matter.

Yet, it works. The claim goes through, the screen gets replaced, and the coverage provides for what it promised.

This brings me to today’s piece.

Lauris works on how states of the world become financial markets — event contracts, derivatives, corporate risk, and the legal and market structure sitting between them.

In today’s piece, Lauris argues that a hedge isn’t a thing to buy at random. It should be a curated relationship built around an exposure you’ve already identified. If you get that order backwards, then you will walk away holding something that pays out on a proposition while your actual problem persists unattended.

Follow Lauris on X

On to Lauris’ story,
Prathik

Anyone can hedge a trade. That does not mean anyone can sell a company “a hedge.”

For a trader, any position that reduces risk elsewhere on the book can reasonably be called a hedge, whether the trader sits on a macro desk, trades crypto, or uses a retail brokerage account.

Buy an election contract against a portfolio exposed to the result and you have hedged the portfolio. The fit may be rough and the protection partial. That still counts as a hedge.

Corporate finance uses the same word for a more demanding relationship. The exposure comes first: a cash flow, liability, or operating risk. The instrument has to match the amount, tenor, and risk factor. Whatever it misses is basis risk. Credit, collateral, documentation, and accounting sit around the trade because the relationship, not merely the payoff, is the product.

A swap, forward, option, or event contract can hedge one account and express a view in another. Its economic function depends on the exposure elsewhere on the holder’s balance sheet. For a company, the relevant questions are which exposure the instrument offsets, in what amount, and for how long.

Both meanings of the word are valid. The category error is treating them as interchangeable.

I have now been in several conversations where experience in a quantitative-trading or market-making seat was assumed to carry over to structuring. The confusion is especially common in tech, where a quant background carries a broad competence halo, often deserved: traders have built excellent exchanges, and many are formidable operators.

But technical horsepower does not make role-specific knowledge portable. Quantitative trading and market making are organised around price, information, and inventory. Structuring a product or a swap that may serve as a hedge is a service operation around a client’s exposure. Not isomorphic.

Prediction markets operators, when thinking about this, make that error frequently. Their natural unit is the contract: list an event, attract liquidity, and find the flow. The same order works poorly for corporate risk transfer, where the exposure comes first. Yet the habit has spread outward from trading and crypto: take the available yes-or-no contracts, map them onto a company’s problems, route the orders to an exchange, and call the portfolio a hedge.

IFRS 9 makes the distinction operational. It asks for a hedged item, a hedging instrument, the risk being hedged, a risk-management objective, and an economic relationship between them [1]. The CFTC starts from the same premise when testing swaps used to hedge physical positions: the risk must arise from an asset, liability, service, or physical commercial activity [2].

What Event Contracts Add

The financial case for prediction markets begins with state-contingent claims. Arrow and Debreu supplied the framework: markets become more complete as they can name, price, and transfer claims on more future states [18]. An event contract can pay if a tariff passes, a merger closes, a drug is approved, or a carbon auction clears above a specified level.

Traditional markets often reach those states only through a proxy, if they price them at all. An event market can create the first observable market price for a state that previously lived inside research notes, scenario models, or bilateral conversations.

For price-threshold claims on the same underlying, the connection to traditional derivatives is exact in the limit: the price of a digital payoff is the negative slope of the call-price curve with respect to strike and can be approximated by an increasingly tight vertical spread [19]. The event contract isolates the terminal state. Before expiry, the options route can carry volatility and mark-to-market exposure, as well as dealer intermediation and replication friction that the buyer did not want.

That identity has limits. “The S&P closes above 7,000” can be mapped onto the S&P option surface. “The Fed cuts in March” or “a tariff passes” remains a legitimate state claim without being the strike derivative of one call curve, and its market price differs from a literal physical probability because it also reflects risk preferences, collateral, liquidity, access, and settlement rules.

The state claim is an input to the company’s cash-flow problem. Prediction markets can manufacture the missing claim, but they cannot manufacture the relationship between that claim and a company’s balance sheet.

Where Event Contracts Matter

Two variables determine where event contracts matter financially:

• Whether a traditional derivative or credible replication already exists.

• Whether a company or investor materially holds the underlying risk.

When material risk already has a derivative, the event contract is a substitute. It has to offer a better fit or a lower all-in cost. The structural cost embedded in the traditional route must exceed the event market’s spread, depth, collateral, and impact costs. This is the replication-cost wedge.

The largest long-run opportunity lies in material risks with no existing derivatives. Here the event contract may be the first instrument, making shutdowns, policy decisions, regulatory milestones, weather outcomes, and corporate events observable and transferable for the first time.

Where a derivative exists but no relevant user materially holds the exposure, there is little unmet risk-transfer demand. Another binary may still be a useful trading product. With neither condition, the market belongs to forecasting, entertainment, or general price discovery rather than corporate hedging infrastructure.

Tradability tells you that a proposition can be priced. Materiality tells you whether anyone has a risk worth transferring.

A contract-first funnel skips the second test: it starts with available listings and searches for companies that can be associated with them.

The Wrong Side of the Desk

A trading mindset starts with a payoff and looks for flow. A structuring mindset starts with a balance sheet and builds the trade.

In fixed income, currencies, and commodities, a corporate client arrives with existing exposures such as floating-rate debt or currency mismatch; fuel bills, inventory, or planned bond issues; and acquisition financing, for an abstract object called a hedge.

It identifies the risk component, chooses the instrument, and sets the amount and tenor. Then it measures the residual basis and fits the result into the client’s documentation, credit arrangements, and accounting policy.

ISDA organises the derivatives market in that order: user, underlying risk, instrument [3]. HSBC’s Autohedge does the same, taking exposures, hedge policy, and risk preferences as inputs before calculating trades [4]. The instrument may be legally separate; economically, the hedge is the relationship around it.

In FICC, a dealer may answer a request for quote with a firm principal-risk price [5]. That is execution: it prices the defined instrument without deciding what exposure the client has or whether the instrument offsets it. Adding a Request-for-Quote (RFQ) to a mismatched contract gives the mismatch a price.

Follow Lauris on X

What Goes Wrong When the Contract Comes First

Consider an importer worried about a possible tariff. Its loss depends on shipment volumes, timing, inventory, pass-through to customers, currency movements, and the company’s ability to find substitute suppliers. An event contract might instead pay one dollar if a public tariff crosses a threshold before a fixed date. The proposition is clean; the fit to the importer’s cash flow is not.

That mismatch is basis risk. A structurer decides which part of the exposure can be transferred and which part the client will retain. An intermediary that starts from the contract does the opposite. It finds a public binary that resembles the client’s problem and treats the resemblance as the hedge. After the trade, the treasurer owns the binary and still owns most of the original risk.

Tailoring the contract creates a second problem. Robert Bartlett and Maureen O’Hara studied 41.6 million Kalshi trades. In their framework, there are no liquidity traders in the Glosten-Milgrom sense because the instrument does not routinely serve the hedging and portfolio-rebalancing role that generates that flow in mature markets [6]. Individuals can still use a contract defensively.

Credit markets show how institutions buy state-contingent risk. In Significant Risk Transfer (SRT), a bank retains its loans and buys first-loss protection through an investor-funded credit-linked note; the BIS counted roughly €800 billion of protected loan pools at the end of 2024 [12]. This is a bank-capital precedent, not a recommendation that corporates buy credit-linked notes. It shows why institutions hold the risk inside a funded instrument with a coupon, documentation, loss allocation, and a mandate line.

The paper also finds higher informed price impact in single-name markets than in broad macro markets [6]. A tariff decision is usually exogenous to an ordinary importer, so its order carries little information. A market maker can welcome the flow, but the public tariff line is only loosely connected to the importer’s loss.

Tighten the contract around a merger, a drug trial, a plant, or another company-specific outcome and the basis improves. The company may now know more than the person quoting it, so the market maker widens, cuts size, or walks away. The welcome trade is the one with the largest basis. The close-fitting trade is the one the market is least willing to warehouse.

A filled contract still leaves the company to explain the objective, ratio, basis, valuation, and financial-statement treatment. A dashboard does not establish that relationship, and an RFQ cannot make an audit committee accept it.

The business model is squeezed from both sides. If the exposure is small and already matches a listed contract, the client can trade it directly. If it is large or bespoke, the client needs structuring and capital. A company that only routes the listed contract adds no capacity. A company that designs the payoff, arranges documentation, and commits or sources capital has become a broker, insurer, or FICC structurer rather than a new category.

The Wrapper Changes the Market

WeatherBill launched a self-serve weather-derivatives platform in 2007. The thesis was that exposed businesses would buy protection. They did not. The company narrowed to farmers, moved onto insurance paper and outbound distribution, and became The Climate Corporation [7].

Event contracts found native demand elsewhere: CME’s FanDuel platform launched in December 2025 and reported 100 million contracts in about eight weeks [8].

Weather went to insurance; sports found retail distribution. The standalone corporate hedge sold through a software funnel remains the missing case.

Where The Risk Sits

Once the exposure is defined, the remaining problem is risk-bearing capacity.

In practice, event risk reaches a balance sheet through three routes. Risk never appears, and has to be transferred somewhere.

1. Trade it directly when the exposure already fits

A Manhattan bar covered a free-drinks-if-the-Knicks-win promotion with roughly $5,000 of event contracts; the liability and contract resolved on the same game [9]. If settlement fits, the same route can handle larger exposures.

A contract tied to a California solar-credit tax rebate reportedly transferred about $600,000 through a displayed, centrally cleared market [10]. Company size is not the dividing line; fit and book capacity are.

I am working with some of the extremely talented folks at @kalshi, especially with @0x_ultra, on providing a showcase for this, as part of the Builder’s Program. It will be released shortly.

2. Fund a pool when the risk is actually poolable

A parimutuel market divides a committed pool among the winners, capping aggregate liability at the capital already present; Goldman Sachs and Deutsche Bank used the mechanism for economic derivatives from 2002, with non-farm-payroll auctions averaging about $9 million before moving to CME in 2005 [11]. The BIS still questioned whether genuine hedging demand would balance sophisticated informed traders [11].

Pooling works only when losses differ, opposite exposures exist, or outside capital is paid to participate. When every participant loses together, the pool simply produces a queue of claimants that someone with capital must stand behind: insurance, reinsurance, or a warehouse again.

I know really smart people like @AadvikVashist are working on this.

3. Put the event inside an instrument institutions already own

In April, Marex issued up to $10 million of a structured note to one Swiss institutional client, paying 7% if Nvidia remains the world’s largest company after a year; the client owns a Marex obligation, and Marex uses event contracts to run the replication [13]. The institution bought a security with an issuer, documentation, and a mandate line, while the event contract stayed on the dealer’s side of the trade. Before the trade reached the client, Marex had turned the event contract into FICC.

Software has a role after the exposure is defined. It can test traditional instruments, identify a coverage gap, and compare an event claim on basis, cost, execution, collateral, law, mandate, accounting, and residual risk. A useful product would encode that process and allow the answer to be no trade.

The Regulatory Blast Radius

A bad corporate-hedging pitch would normally be a problem for its buyers. The damage is wider here because event markets are still fighting over the role they are allowed to play in the financial system. Commercial use does not determine CFTC jurisdiction over an event contract.

The Crypto Council for Innovation grounds that jurisdiction in the Commodity Exchange Act’s swap definition, federal preemption, and the CFTC’s exclusive authority over derivatives markets [14], and an event contract does not stop being federally regulated merely because the person buying it is speculating.

The public-interest case extends beyond corporate hedging. Regulated event markets can name previously unpriced states, produce public prices, and create transparent, collateralised claims with defined settlement rules. Those are serious financial functions even before a treasurer trades one. But hedging remains central to the policy case.

The CFTC describes prediction markets as tools for forecasting, planning, hedging, and speculation. In February 2026, the Commission defended its jurisdiction against state gambling regulators by pointing to business hedging, portfolio management, and information about future outcomes [15]. CCI draws the distinction from gambling through many-to-many execution, transparent pricing, defined settlement benchmarks, surveillance, customer protections, and market integrity [14].

In March, the CFTC reminded prediction-market exchanges of their obligations under the Commodity Exchange Act and Core Principle 3. Its June proposal addresses event-contract design, public-interest review, and responsible innovation [16].

A serious corporate hedge failure would follow a simple pattern. A business is told that a binary position offsets an operating risk, so it posts cash. The contract may expire worthless even though the company loses money, because volumes, timing, or pass-through were never in the payoff; before settlement, its mark may move through earnings while the original exposure remains, leaving the company with a proposition rather than protection.

Regulators have seen the payoff shape before. The CFTC and SEC issued a joint alert after complaints about binary-options platforms refusing withdrawals, identity theft, and software manipulation that generated losses [17]. The regulated event venues discussed here are not those fraudulent brokers: their markets are supervised, transparent, and centrally cleared, but that history raises the cost of mislabeling a trade as protection.

Opponents of federal prediction markets will not write a seminar about basis risk. They will say a company was sold a bet under a derivatives label.

An intermediary that sells a mismatched binary as corporate protection will hand state gambling regulators the industry’s strongest argument against federally-regulated prediction markets.

If you got to the end, truly appreciate you for reading this.

That’s all for today.

Until next time,
Lauris

Follow Lauris on X

References

[1] IFRS Foundation, IFRS 9 Financial Instruments, paragraph 6.4.1 and related hedge-accounting provisions. [PDF]

[2] Commodity Futures Trading Commission, discussion of swaps used to hedge physical positions, 83 FR 7927 (2018). [Final rule]

[3] International Swaps and Derivatives Association, *Evolution of OTC Derivatives Markets Since the Financial Crisis* (January 2021), table 1. [PDF]

[4] HSBC, [Autohedge] and [FX Overlay for Corporate Clients].

[5] Citi, Markets & Banking Execution Policy — FICC. [PDF]

[6] Robert Bartlett and Maureen O’Hara, *Adverse Selection in Prediction Markets: Evidence from Kalshi*, Stanford Law and Economics Olin Working Paper No. 615 (April 16, 2026). [SSRN]

[7] OpinionX, *Deconstructing WeatherBill’s $930M Startup Pivot*. [Article]

[8] FanDuel and CME Group, [FanDuel Predicts launch release]; CME Group, [100 million event contracts announcement].

[9] Catherina Gioino, *This NYC Bar Just Covered Everyone’s Tabs Because the Knicks Won, and Used Kalshi to Hedge Their Bets*, *Fortune* (June 3, 2026). [Article]

[10] John Lothian News, *Kalshi Eyes Institutional Scale as Prediction Markets Prove Fed-Grade Forecasting Value* (July 9, 2026). [Article]

[11] Blaise Gadanecz, Richhild Moessner and Christian Upper, *Economic Derivatives*, BIS Quarterly Review (March 2007). [PDF]

[12] Prashant Babu, Michael Chui and Costas Stephanou, *The Rise and Risks of Synthetic Risk Transfers*, BIS Quarterly Review (March 2026). [Article]

[13] Finance Magnates, *Marex Launches Structured Note Linked to Prediction Market Outcome* (April 2026). [Article]

[14] Crypto Council for Innovation, response to the CFTC prediction-markets advance notice of proposed rulemaking (April 2026). [PDF]

[15] Commodity Futures Trading Commission, [*Understanding Prediction Markets and Event Contracts*] and [press release 9183-26].

[16] Commodity Futures Trading Commission, [prediction-markets advisory] (March 12, 2026) and [*Prediction Markets; Public Interest Determinations*].

[17] Commodity Futures Trading Commission and Securities and Exchange Commission, *Binary Options and Fraud*. [Investor alert]

[18] Kenneth J. Arrow, *The Role of Securities in the Optimal Allocation of Risk-bearing*, *Review of Economic Studies* 31(2) (1964), [DOI]; Gérard Debreu, *Theory of Value: An Axiomatic Analysis of Economic Equilibrium* (1959), [PDF].

[19] Douglas T. Breeden and Robert H. Litzenberger, *Prices of State-Contingent Claims Implicit in Option Prices*, *Journal of Business* 51(4) (1978). [JSTOR]

We will be back with another guest essay soon.

P.S. This piece was originally published here.

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.

Two decades ago, I used to accompany my mom to the vegetable market. I’d try to help by picking a few tomatoes and some onions. They never made it to the final basket. She would quietly set aside my picks and choose her own. Each one that made the cut was pressed, turned over, held up against some standard only she could see.

For most people, grocery shopping is a chore. For the chef in my mom, it was a high-precision job. The right tomatoes meant the curry would taste the way she intended. Choosing wasn’t merely a step before cooking. It was where the cooking began.

That’s the thing about shopping, it’s not one activity. Some of it we’d delegate to anyone who’d take it. Some of it we won’t hand over even to our own kids. And any attempt to delegate how we buy things has to start by understanding which is which, not by assuming that automated robots are the default solution to everything.

In today’s guest op-ed, Anderl takes apart the “give the agent a wallet” thesis and shows why paying was never the hard part of shopping.

On to Anderl’s story,
Prathik

Follow Anderl's Works

There is a line going around the AI and crypto world right now: give an agent a wallet, let it shop for you. The ultimate use case. It sounds clean, it sounds like the future, and it still doesn’t feel quite right. The mismatch is worth unpacking because it shows that the line confuses the hard part of shopping with the easy part and then puts the easy part at the centre.

Let’s start with shopping itself, not with payment.

Shopping Splits into Two Things

Shopping is really two activities that today only look fused. One is search, the other is judgment. Search (gathering, filtering, comparing, pre-sorting) is mechanical and delegable, almost entirely. Judgment (is it any good, does it suit me, do I trust the seller) is the part that a human is attached to.

It is measurable that search is already migrating. Adobe Analytics reported a roughly 4,700% rise in traffic from generative AI referrals to US retail sites between July 2024 and July 2025. The wallet thesis now quietly assumes the agent takes on both search and judgment. This is where the sleight of hand sits. One can be handed off, the other only under conditions, and sometimes not at all.

Judgment Splits Again

The interesting move is that judgment isn’t monolithic either. It breaks into two parts. One is evaluation, which involves checking options against a utility function. The other is authorship, where you set that utility function in the first place: which dimensions count, how they are weighted, which values bind, and what “good” ends up meaning.

Evaluation automates. Authorship doesn’t. And it doesn’t sit as a single gate up front that you pass through once. It runs as a thin thread through every layer. Conformance is about which specifications matter to you. Quality is about whether a broken zipper is a deal-breaker or beside the point. And the choice of merchant turns on which values bind you. Every layer does (human criterion × agent evaluation). In every case, what automation eats is the evaluation half. What’s left is the authorial half.

This has a consequence that everyone underestimates who thinks the human writes one spec sheet and walks away. You can’t. It’s one of the most robust findings in decision research. Since Paul Slovic’s work on decision-making under risk, we speak of constructed preferences. The American professor and psychologist found that we don’t carry a finished liking inside us that we simply retrieve; we form it in the act of choosing (Slovic 1995; Lichtenstein & Slovic 2006). The evidence is preference reversals. Normatively equivalent elicitation methods like choosing versus pricing produce systematically different rankings, violating the basic axiom of rational choice. Ariely, Loewenstein and Prelec (2003), in their work on coherent arbitrariness, showed that even an arbitrary number (the last digits of a Social Security Number) anchors willingness to pay for familiar products, and that the effect disappears neither with experience nor through market forces. The “stable preferences” picture is largely an illusion of order.

The realistic interface is therefore not the spec sheet but an iterative one. The agent surfaces, at each layer, the decision a criterion needs (”you said durable, at what premium does durability stop being worth it?”), and the human writes exactly that.

The Real Axis: Chore or Pleasure

Now, the distinction that orders everything. We like to talk about “standardised” versus “individual” products, as if it were about specifiability. It isn’t. The real axis is different. On one side, choosing is a pure chore. On the other hand, choosing is part of the point.

For a commodity like printer paper, batteries or the recurring reorder, the act of choosing has zero consumption value. Nobody enjoys deciding between two identical toner cartridges. That’s why commodities are the natural home of full delegation: nothing is lost when the agent silently reorders the right thing.

With pleasure goods, it’s the reverse. The wine, the piece of furniture, the coat, and the book. Here, choosing is part of the good itself. Whoever hands off the judgment not only saves effort but also destroys part of what they wanted to consume. And that holds even if the follow-up questions were free. You wouldn’t want to delegate it even at zero friction.

With pleasure goods, the agent therefore doesn’t withdraw but should switch roles. From buyer to scout. It still does the search, the shortlisting, the spec matching, the merchant verification, and the pulling of recurring weaknesses out of a thousand reviews. It narrows 200 to 5. And then it stops. The human savours the final choice.

The Friction Trilemma

“But the agent could just ask for my criteria.” It could, and that’s exactly the annoying variant that people resort to doing themselves. Worse, interrogating degrades the choice. Wilson and Schooler (1991) had participants rate jams. Those asked to analyse the reasons for their feeling beforehand ended up with rankings that matched expert judgment less well than the control group’s. In a follow-up study, people who spelt out their reasons chose different posters and were less happy with them weeks later. Articulating directs attention to what is easy to put into words, not to what actually carries the preference. Taste is recognised, not verbalised.

So there is a trilemma:

Interrogate. Faithful to your current self, but high friction, and it can even worsen the choice.

Infer from the past. Low friction, but it locks you into your past self and shuts down discovery. You only get more of what you already want.

Judge yourself. Full autonomy, full labour.

There is a fourth mode that avoids the bad corners: recognition instead of interrogation. “Show me three, I’ll point to one.” Low friction and accurate at once, because it doesn’t require naming criteria in the abstract that you often can’t name. Here, the most famous study on choice overload helps too. Iyengar and Lepper (2000) sold jam at a tasting stand. With six varieties, far more people bought than with 24. Still, the effect is disputed, and honesty requires saying so. A meta-analysis by Scheibehenne and colleagues (2010) found, averaged over many studies, almost no general choice-overload effect. A second one across 99 studies (Chernev, Böckenholt & Goodman, 2015) showed that it appears only under certain conditions, such as high complexity, a hard decision, and an unclear preference. Telling is the original authors’ self-criticism. Perhaps, faced with 24 varieties, customers simply didn’t have enough time to “determine their preferences”. The line where autonomy actually lives is the one between “the agent applies my criteria” and “the agent invents them from my history”.

Back to the Wallet

With all that behind us, we can finally name why the wallet line sounds off. It throws together three things that are separable: who decides, who executes and who holds the funds. “Give the agent a wallet” answers only the third, and that only matters if the first falls to the agent, too.

Three cases. Human decides and pays themselves. Here, the agent doesn’t pay but remains a scout. In the second, the human decides and delegates execution (”yes, buy that one”). Now, the agent runs the checkout, but it needs no custody of money, only a narrow, revocable authorisation for exactly this approved purchase. Only in the third case, when the agent decides and pays autonomously, with no human at checkout, does the wallet itself become load-bearing.

And here it gets interesting because the payments industry built exactly this separation in 2025. Not custody, but authorisation. The Agentic Commerce Protocol, developed by OpenAI and Stripe, issues a “Shared Payment Token” bound to a single merchant, with a fixed amount, that is time-limited and single-use. Mastercard’s Agent Pay (April 2025) issues “Agentic Tokens” scoped to a specific agent, a merchant scope and a consent policy. The agent never sees the raw card number. Google’s Agent Payments Protocol (AP2, September 2025) even cleanly separates an Intent Mandate (what the human wants) from a Cart Mandate (what the agent proposes to buy), both as signed verifiable credentials, which is exactly the authorship-versus-evaluation axis cast in cryptography. Visa’s Trusted Agent Protocol (October 2025) takes the same route. None of these approaches hands the agent funds. All of them hand agents a narrowly bounded authority. The industry independently confirms the article’s thesis.

So where does the agent-owned wallet become native? Precisely in the commodity tier and in the machine room. The genuinely strong crypto cases don’t lie in consumer shopping. The x402 protocol, backed by Coinbase and Cloudflare, fills exactly the gap that card rails don’t close: fully autonomous agent-to-agent payments with no human in the loop, billed per API call, around the clock. Over 100 million transactions in the first months. Mastercard, in turn, announced “Agent Pay for Machines” for high-frequency, low-latency micro-payments between machines. That is infrastructure for a machine economy, not shopping. So the thesis isn’t wrong. But its importance is inverted. The wallet of its own becomes load-bearing where the goods are most interchangeable and the amount per transaction is often the smallest.

Follow Anderl's Works

Where the Wallet Really Belongs

This explains, in passing, why crypto has shifted its centre of gravity over the past two years, away from the consumer-driven culture of liberation, toward rails for companies: stablecoin settlement, tokenisation, and institutional infrastructure. That isn’t a turn away from agent shopping but is a homecoming to the one tier where the funds-holding wallet actually fits. Corporations are its anchor tenants for structural reasons. A procurement department is the institutionalisation of “choosing is a pure chore”. Its purpose is to replace taste and gut feeling with specification, price, reliability and contract terms. Procurement is agentic shopping, carried out by rule-following humans. So the agent wallet teaches corporations no new behaviour, and instead automates one they perfected long ago.

You can see it in an entire industry layer built for exactly this. Many companies have outsourced the purchasing of C-items, such as office supplies, to specialised platforms, including Mercateo and Amazon Business. Their promise is not the lowest price. It’s the lowest effort: one catalogue instead of twenty suppliers, one invoice instead of many. The company accepts a slightly higher unit price and saves the process costs, which for C-items often exceed the value of the goods themselves. The agent attacks exactly this logic. It pushes the ordering effort toward zero, while simultaneously searching the fragmented market nobody had time for before. What remains for the platforms is only conformance: approved suppliers, a clean invoice, protection against counterfeits. This verification, further down, turns out to be the real bottleneck.

The recommendation is therefore more precise than “wallets for firms instead of consumers”. It reads: ‘match the instrument to the tier’. The autonomous, custodial wallet (the agent decides, holds, pays) has its market in the procurement of standard goods and in machine-to-machine payments. The buying there is high-volume and repetitive, driven by specifications, so largely B2B and M2M. The consumer side, by contrast, wants authorisation instead of custody, meaning a narrowly bounded, single-use token on existing rails, only when the human approves a purchase. “Give the agent a wallet”, as a consumer headline, therefore, aims at the weakest slice. The product’s actual bridgehead is the machine room of companies.

A caveat that stays honest. Corporations don’t buy only commodities either. They have their own non-delegable tier, just one driven by consequence rather than pleasure: the choice of a law firm, an acquisition target, or a strategic supplier. There, the human stays, and the autonomous wallet has as little business as it does with the consumer’s wine.

The rule is domain-independent. The wallet rides the commodity layer wherever it sits, and that layer is simply the largest and densest inside companies.

What the Bottleneck Really Is

Paying was never the bottleneck. Moving money is solved, several times over. The two real bottlenecks are different.

One is trust in the inputs. Automating judgment is only a gain if the data ground holds. If it doesn’t, a judging agent is worse than a sceptical human. It launders manipulated signals at machine speed. That this ground is shaky today is official. The US trade regulator, the FTC, issued a rule in 2024 (effective 21 October 2024) banning fake reviews, noted that such fakes are widespread, and explicitly named generative AI as a tool that eases their mass production. By late 2025, the first warning letters followed, with penalties of around $51,744 per violation. Add the physical counterpart. Per OECD/EUIPO (2025), global trade in fakes amounted to roughly $467 billion in 2021, or 2.3% of world trade, and in the EU to 4.7% of imports, with clothing, shoes and handbags at the top. Pleasure goods, of all things. Anchored product identity, verifiable reviews, a verifier independent of the seller, and provenance records at the unit level (as the EU anti-counterfeiting directive and the US DSCSA have long required per pack in pharma) are the precondition that tips judgment automation from dangerous to useful.

The other bottleneck is authorship. Everything downstream of a defined wanting becomes increasingly automatable. Defining the wanting stays human, not out of nostalgia but structurally. To automate it, the agent would have to want for you, and a wanting fabricated for you is not yours.

Give the agent a wallet, and you’ve solved the easiest part. The interesting part is making judgment partly and safely automatable, and leaving the human where they belong: as the author of the criteria and as the one who savours the final choice.

For Anyone Building Marketplaces

One consequence to close, for pleasure goods. Once agents commoditise sourcing, the goods are findable everywhere. Then your product is no longer the goods. Choosing is. Make yourself maximally legible to the scout agent (verifiable provenance, clean data) so it routes the human to you. And on top of that, occupy the one thing automation can’t take: the discovery that widens taste rather than confirming it, and the pleasure of the final decision. Hand sourcing to the agents. Build your business around the better choices.

That’s all for today.

Until next time,
Anderl

Follow Anderl's Works

We will be back with another guest essay soon.

P.S. This piece was originally published here.

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.

“The world’s $2.5 trillion trade financing gap is not a credit problem. It is a cost-of-underwriting and information-asymmetry problem.”

Hello,

Banks are great at lending money to people who don’t need it as badly as others.

That’s not foul play, but business economics. Checking whether someone is trustworthy costs roughly the same in money and effort whether they’re shipping $40,000 worth of cashews or $40 million of semiconductors. Both still have to follow similar paperwork and compliance procedures. The transactions will have to go through the same network of three to six banks. This gives banks the benefit of achieving economies of scale if they prefer higher-order values to small-value ones.

This is the single most important incentive for banks to turn down small exporters that are yet to build their credibility. That’s probably why small and medium exporters have a 50% rejection rate, roughly seven times that of multinationals.

The asymmetry in the trade financing gap is not only based on enterprise size but also on geography. Asia-Pacific alone accounts for 40% of the world’s $2.5 trillion trade financing gap.

The system is broken for a specific set of people in certain areas. When exclusion structurally affects certain stakeholders, it makes it easier to map the problem and find a solution.

In global trade financing, one of the biggest problems is duplication. Every trade forces four or five institutions to verify the same facts, like who shipped what, to whom, against which invoice. Each institution records these details in its respective systems and charges for doing so. The cost of trust is incurred repeatedly, making it more difficult for small and medium exporters to absorb it.

Blockchains can solve this by moving stablecoins across a shared ledger in minutes, rather than routing fiat currency through multiple correspondent banks over three to five days. This is where blockchains can improve inclusion in trade financing by specifically making the entire proposition viable for global traders who are often ignored at the periphery.

Today’s guest essay is an investment thesis from 1kx, an investment firm that focuses on decentralised finance and stablecoin payments. In this thesis, Nichanan Kesonpat - Research Principal at 1kx - maps out how on-chain trade financing unfolds layer by layer and where value accrues for the industry to capture.

Read 1kx's Thesis

Follow author's works

P.S. This thesis was originally published by 1kx here.

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.

For a long time, since OpenAI made large language models a common household phenomenon, AI felt like a new toy in the town to play around with. It was only after the new models began consuming significantly more energy and resources that the reality of AI hitting its token limits dawned on many.

Even in our editorial calls and conversations on AI forums, every new model always prompted people to utilise it to its full potential and compare the results. Most of those who debated AI models paid less attention to how each marginal improvement in the thinking model increased unit cost disproportionately, often for similar or only slightly better output.

Better thinking models meant you exhausted your daily token limit faster, with fewer queries and prompts. This changed the user behaviour almost immediately. Even after I started hitting my daily limits more often, I still used AI, but I became more careful about choosing models based on the task I wanted to carry out.

That extra decision annoys users, but it begs a much larger question. The trillions of dollars of capex flowing into the AI industry, the series of data centres being set up, and the investments flowing into the portfolios of companies operating in compute suggest that demand for intelligence is infinite. But most people and companies don’t repeatedly buy capacity for intelligence unless they have a clear task to achieve - like write faster, code better, sell more, support customers, cut costs, or make decisions.

This is why presuming the demand for intelligence is near-infinite is a flawed assumption.

In today’s piece, Kevin Simback, COO at Delphi Labs, explains this clearly. Demand for intelligence may be enormous, but it is not infinite at every price. It becomes near-infinite only when the cost of useful work falls below the point where buyers can justify it.

On to Kevin’s story,
Prathik

Follow Kevin on X

I keep coming back to one assumption that only some people are starting to say out loud.

The entire buildout - the data centres, the chips, the trillion-plus dollars of committed capital, and the large, growing share of the stock market now riding on it - rests on a single belief. That the demand for intelligence is infinite.

It isn’t. Demand for intelligence is near-infinite at a price. That qualifier is the difference maker.

In the first half of this year, we’ve mostly ignored price and just devoured intelligence. You can see it clearly in Anthropic’s revenue numbers.

We’ve even coined a new term, tokenmaxxing, to gloat about our unbounded consumption.

For most companies, this just isn’t sustainable. Companies need to find ROI, otherwise they will absolutely throttle demand.

In the first half of this year, we got away with convincing the CFO to loosen the purse strings, but the second half will be about tightening them back up.

This article double clicks on this thesis, what to do about it, and what happens if we get this wrong.

The Demand Curve Nobody Drew

“Near-infinite demand” and “infinite demand” may sound like the same thing. They are actually two completely different economies.

Infinite demand is a vertical line. It says: people will buy intelligence at whatever price is offered, so the only thing that matters is making more of it.

Just keep scaling the supply and the demand is already there waiting. Most of the capital being deployed today behaves as if this is true.

How many people have you heard say, “If Anthropic had more compute, they could be doing $100B ARR?”

Near-infinite demand is a downward-sloping line with a hard floor underneath it. It says: demand for intelligence explodes, but only once the price drops below the level where the buyer earns a return on it.

Above the ROI line, demand is thin and more volatile. Below it, demand is effectively bottomless. The floor is ROI, and where a company sits relative to it decides everything.

This shouldn’t be considered crazy talk, yet many speak of the demand curve as vertical.

The actual buyers are telling us exactly where the line is. An MIT study last year found that only about 5% of enterprise generative-AI pilots produced a rapid, measurable financial impact. The rest stalled.

A large cross-country survey of executives this year found that nearly 90% of firms reported no measurable effect from AI on productivity or headcount over three years.

The typical response is to call this a “skill issue” and just assume companies will get there.

That may be true; perhaps we just need more Anthropic and OpenAI Forward Deployed Engineer (FDEs) embedded within enterprises to make ROI a reality. That seems to be what the new “deploy cos” are betting on.

But it’s hard to assume both things can be true at once - either the demand for intelligence is infinite (in which case, why do the labs need deploycos?) OR demand is near-infinite but only under the right conditions.

Tokenmaxxing: Optimising against Your Own Customer

Here is where it gets a bit self-inflicted.

The reflex across the industry is to reach for the biggest model and the deepest reasoning setting for every task.

If you leave this up to users, that’s exactly what you’ll get - they’ll just run everything through Opus 4.8/GPT-5.5 because, why not?

But this behaviour is what will quietly kill ROI.

The newest reasoning models make it so easy to burn tokens. They generate long internal chains of thought before they answer, and you pay for every one of those hidden tokens.

A request that returns a few hundred visible words can burn many thousands of billed tokens under the hood, which means the real cost per useful answer can run an order of magnitude higher than the sticker suggests.

In one controlled evaluation, a top reasoning model cost roughly 14x more than a cheaper general model on the same task and produced four times as many failures. Maximum intelligence, applied indiscriminately, made the result both more expensive and worse.

The deeper problem is an incentive mismatch - the seller is paid by the token, but the buyer is paid by the outcome.

Tokenmaxxing maximises the first and starves the second. It pushes the price of every decision up the demand curve, away from the floor where demand actually lives, at the exact moment the whole system needs to drive price down through it.

You cannot fill near-infinite demand by selling the most expensive version of the product to everyone. You fill it by getting the price of a good-enough answer under the ROI line and letting volume do the rest.

Follow Kevin on X

Why “But Costs Always Fall” Isn’t the Rescue

I know you want to say “Jevons paradox” right now, but that’s not going to let you off the hook.

First, the cost that has to fall is the cost per useful outcome, not the cost per token. Those two can be related, but not if we don’t address the behaviour mentioned above.

If per-token prices drop while tokenmaxxing pushes token consumption per-task up even faster, the cost of getting a real job done doesn’t fall at all. It could even rise.

Falling unit prices and rising bills coexist comfortably, and I hear reports of teams living that contradiction right now.

Second, the savings have to actually reach the buyer. Efficiency captured entirely by the seller, or burned on more reasoning that the task never needed, never crosses the ROI line where the demand is.

If you lower the cost of tokens for a model, but the reasoning and tool calling under the hood drives up the number of tokens consumed, it’s not clear that the outcome itself is getting cheaper.

This is a markets problem, not a procurement problem

If this were only about enterprise software disappointing, it wouldn’t be that big of a deal. But we’ve now put most of the global economy’s chips in the “AI trade” basket, so any slip-up here and the implications get big.

The major US hyperscalers will spend more than $1 trillion in 2025-26. That spend is underwritten by an expectation of future revenue that depends on enterprise adoption, which in turn is dependent on ROI.

The gap between capex and revenue is being bridged with financing, some of which looks quite circular. I’m not ready to raise any alarm bells yet, but we need to connect all this back to the demand curve.

In the simplest view, if demand turns out not to be infinite, but is actually near-infinite and gated by ROI that hasn’t yet materialised, then the big risk is that the revenue that justifies the capex doesn’t arrive on schedule.

The financing loop can’t keep substituting for real cash flow forever. A technology this leveraged and this concentrated doesn’t get to underdeliver without major economic reverberations.

Said bluntly, if ROI doesn’t show up at scale, the AI trade begins to unwind, and a reflexive, circular, concentrated structure unwinds the way it was built: fast, and on everyone at once.

That is the worst-case scenario. I’m not saying that is the most likely, but it is a scenario that starts with enterprises becoming unhappy about their pilots and throttling demand.

“ROI takes time” - yes, and that is exactly the danger

The most serious objection to the ROI issue is one of timing.

Every general-purpose technology lags, and the payoff can be real. But it might not be on the same quarterly clock that drives the markets.

The mismatch is between a payoff that runs on the slow clock of organisational redesign and a financing structure that runs on the fast clock of capital markets.

The technology can be a genuine multi-decade transformation, and the trade financing can break in the meantime. Both can happen at once.

That is why time is the scarce input here, and why we need to get moving on it.

What ROI Already Looks Like

The picture is not entirely bleak, and that’s not what I want as the takeaway either.

Some enterprises are getting real, measured returns right now. But, in many cases, it is not the teams chasing the smartest possible model on the broadest possible mandate.

It is the teams running narrow, bounded, repeatable workloads, with the right-sized model for the job and an actual measurement of the outcome.

The ROI is happening in companies that deploy intelligent routing - sending easy queries to cheap models and escalating only the hard ones.

The ROI is happening in teams that swap a frontier model for a small, fine-tuned one on high-volume structured tasks.

The ROI is occurring in enterprises that employ prompt and context-trimming tools to mitigate against users sending everything to the most expensive, highest-reasoning model.

The ROI is happening in projects that build custom agent harnesses that manage tool access, memory, permissions, workflow state, evals, logging, human approvals, fallbacks, and cost controls.

None of these are examples of weaker AI. They are examples of the same outcome delivered below the ROI line instead of above it.

The early winners, those finding ROI, obey one discipline: intelligence per dollar, not intelligence at any price.

The Reframe

The entire premise of tokenmaxxing measures the wrong number.

The number that decides whether this entire AI trade is sustainable is the outcome per dollar.

Useful work delivered, divided by what it cost to deliver. That ratio is what moves a use case from above the ROI line to below it, and the volume of demand waiting below that line is the only thing large enough to justify what’s already been spent.

So the urgent work, the genuinely urgent work, is driving cost-per-outcome down hard enough and fast enough to pull the mass of stranded use cases under the line, and getting those methods adopted before the markets run out of patience.

The primary levers already exist:

• Route every task to the cheapest model that can actually do it, instead of defaulting to the most expensive one.

• Trim context and reasoning to what the task needs, rather than dumping everything into the biggest model on the highest setting.

• Wrap agents in harnesses that control tools, memory, evals, approvals, and budgets, so reliability comes from the system and not from brute-force model size.

• Fine-tune smaller open-weight models on proprietary data for the high-volume, repeatable workloads that don’t need a frontier brain.

Each of these is a way of doing the same job for less, which is the only thing that grows near-infinite demand into the real thing.

The optimistic reading of all of this is that the demand really is there. It is enormous, it is waiting, and it is gated by a single solvable variable - ROI. We just need to collectively make it happen soon.

P.S. This piece was originally published here.

Follow Kevin on X

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.

I’ve been reading David Graeber’s ‘Debt’ lately, and it opens by exploring how old and how human debt actually is. It explained debt as a relationship that assumed a future for both lender and borrower. The tab remained open throughout the loan tenure, assuming both parties would be around.

But the modern version of debt has drifted far away from that. On-chain lending makes the contrast look more ridiculous: a loan can be closed out without the borrower’s fault. If your collateral dips and a predetermined price level is crossed, the position is liquidated at a penalty. Ironically, this could happen just a day before the collateral recovers and rises well above the loan level. For being briefly underwater, the borrower is assumed to be insolvent and penalised.

The crypto advocate in me expects blockchains to do much better. In its current form, liquidation defeats the purpose of debt, which was to keep a relationship open through uncertainty rather than slam it shut at the first sign of trouble.

So what would it look like to build a loan that keeps faith with the borrower instead of betting against them? A loan that bends when things get hard rather than breaking?

In today’s guest essay, Jayesh argues that the answer has been sitting in traditional finance for 40 years. He takes a technique called Constant Proportion Portfolio Insurance (CPPI), the same idea behind capital-protected products banks have sold for decades, and maps it onto on-chain lending to produce a loan that defends its own collateral.

It is an honest piece because it offers a solution while remaining clear-eyed about where it can still fail.

On to Jayesh’s story,
Prathik

Follow Jayesh's Works

Most on-chain loans die in the same dull and predictable way, where the collateral falls, a threshold is crossed, a liquidator closes the position at a penalty, and the borrower is left holding the loss on an asset that very often recovers a week later. We have spent five years making the price of credit smarter, and almost none of that energy on the thing that actually wipes people out: the liquidation itself.

I want to propose a different loan structure that borrows a 40-year-old idea from traditional finance called Constant Proportion Portfolio Insurance (CPPI). It produces a loan that defends its own collateral instead of waiting to be liquidated. Curve and f(x) have already shown the instinct works on-chain. What nobody has done yet is name the mechanism for what it is, drive it off the right variable, and hand the borrower the dial that decides its risk.

Let me walk through where lending has been, what this primitive is, the math that makes it work, where it can fail, and how a protocol could actually build it.

Innovation of On-chain Lending

The first generation of on-chain lending was variable rate, and it still anchors the entire category. Compound v2 went live in May 2019. ETHLend, which launched in 2017, rebranded as Aave in 2018, and moved to a pooled liquidity model with Aave V1 in early 2020. Both set the borrowing rate in the same way: algorithmically based on utilisation. When more of the pool is borrowed, the interest rate rises along the interest rate curve, attracting additional supply and discouraging new borrowing. The rate floats for everyone in the pool and updates every block as borrowing demand shifts.

This pooled variable model won decisively and still anchors the largest part of the market today. Aave alone holds around $13 billion in deposits in mid 2026, and Compound, now on its isolated market design called Comet, sits a little above $1 billion. It is worth noting how hard it has been to improve the rate side, even for the established ones. Aave shipped a stable rate borrowing option years ago, then disabled new stable rate borrows in November 2023 after a vulnerability was found in the stable rate logic, and fully deprecated it through governance in 2024. The dominant on-chain lending experience today remains variable rate, and that has barely changed in years.

The current frontier is fixed-rate lending, and this is where most of the recent design talent has gone. Pendle is the clearest example and the category leader, holding around $1.3 billion in mid 2026. It works by splitting a yield-bearing asset into a Principal Token and a Yield Token. The Principal Token behaves like a zero-coupon bond: you buy it at a discount and redeem it at face value at maturity, locking in a fixed yield. In May 2026, Morpho published the Midnight whitepaper, a fixed-rate, fixed-maturity protocol in which lending and borrowing are expressed as the trading of credit and debt units whose payoff is analogous to that of zero-coupon obligations, with offers that do not lock up capital until settlement. Midnight is freshly released and open-sourced, and it squarely signals that serious builders are pointing at the rate.

So it is worth looking at the two axes of innovation side by side. On the axis of what credit costs, we have moved from variable to fixed and built a genuine toolkit along the way. On the axis of what happens when the collateral falls, almost nothing has changed. Since the overwhelming majority of on-chain debt is still protected by hard liquidation, a single line change in price could get your position closed and penalised. This neglected second axis is the gap I care about.

Exceptions that Already Point the Way

Two protocols deserve full credit because they addressed this exact problem before anyone else.

Curve’s crvUSD introduced soft liquidation through a mechanism called LLAMMA - the Lending Liquidating AMM Algorithm. Instead of one liquidation price, your collateral is spread across a set of discrete price bands, anywhere from four to 50 of them, and each band acts as its own small liquidation zone. As the price falls through a band, that band’s collateral is progressively sold into crvUSD. As the price recovers within a band, the crvUSD buys back the collateral. Curve calls this de-liquidation. Arbitrageurs do the actual trading because LLAMMA quotes prices offset from the oracle, making rebalancing against external markets profitable. The position is never slammed shut at a single threshold, and its exposure is instead continuously shifted between the volatile asset and the stablecoin. crvUSD has run real borrowing through this design.

f(x) Protocol does something structurally similar with its Liquidation Brake. When a leveraged position approaches its liquidation price, the protocol burns part of the position’s debt, sells some backing collateral, and pulls the leverage ratio back down. This reduces risk while keeping the user’s directional exposure, rather than closing them out. External keepers monitor positions and trigger the sale when needed. The protocol holds roughly $90 million in mid 2026.

Irrespective of how they describe themselves, both are simply doing portfolio insurance on collateral. They de-risk into a decline and re-risk into a recovery. That is the exact reflex of a strategy traditional finance formalised in the 1980s, and once you see the connection, a cleaner and more general version of the loan falls out.

What is CPPI?

Constant Proportion Portfolio Insurance (CPPI) was developed by Perold in 1986, extended to equities by Black and Jones in 1987, and formalised by Black and Perold in 1992. The idea is simple and is built to protect a floor.

You define a floor, which is the value your portfolio must not fall below. You measure the cushion, which is how far you currently sit above that floor.

Cushion = Portfolio Value - Floor.

You then hold an amount of the risky asset equal to a multiplier times that cushion, and you park the rest in a safe asset.

risky exposure = m × (portfolio value - floor)

The strategy’s behaviour follows from that single line, and it plays out intuitively as markets move. When the cushion is fat, you hold a lot of the risky asset. As losses eat into the cushion, your risky exposure mechanically shrinks toward zero, which moves you into the safe asset before you can breach the floor. When the cushion rebuilds, you lean back into the risky asset. The strategy sells into weakness and buys into strength.

Mapping that same structure onto a loan makes the correspondence exact. The floor sits just above the debt, at the debt plus a small gap margin I size later. That’s because the collateral must never actually fall below the debt, which, if it does, you have a bad debt and a liquidation. The cushion is your collateral value minus the floor, and acts as the safety buffer that every borrower watches. The risky asset is the volatile collateral, ETH or a BTC wrapper or SOL, and the safe asset is a stablecoin. A CPPI loan holds its collateral as a managed basket of the two, and it rebalances that basket as the cushion breathes. As your collateral falls toward the floor, the basket steps out of the volatile asset and into the stablecoin, defending it. As the cushion grows, it steps back into the volatile asset, handing you the upside. What changes is that the ordinary dips that would liquidate a normal position no longer kill this one because it de-risks instead of crossing a single line. The loan fails only if you stop paying interest, miss maturity, or the market gaps so violently that the basket cannot rebalance in time, which is the gap risk at the heart of this design.

The Math and the Numbers

Take the simplest possible example, where your collateral is worth $100, and the loan is $70. Set the gap margin aside for this first pass, so the floor sits at the debt of $70. That makes the cushion $30. Suppose you want to start with your position fully in the volatile asset, so all $100 sits in BTC at origination. That choice sets your multiplier (m).

Risky exposure = m × cushion; so, $100 = m × $30, which means m ≈ 3.33.

In the real design, the floor level sits slightly above the debt. That shrinks the cushion and increases the multiplier required for the same starting exposure. However, the mechanism is identical.

That multiplier is not a cosmetic setting but the entire risk profile of the loan, and it is governed by the well-known gap-risk property of CPPI. It does two jobs at once. First, it sets how much of your collateral starts in the volatile asset, which is m times the cushion, capped at the whole position. Second, it sets the size of the gap you can survive between rebalances, which is 1/m. With m ≈ 3.33, that gap is 30%. So the loan defends its floor against any drop up to 30% that happens gradually enough to rebalance, and it breaks the floor on a gap larger than that before the basket can react. A more conservative borrower who picks m = 2 starts with only 60% in the volatile asset, since exposure is 2 × $30 = $60, allocating the other $40 in stablecoins from the start. This move buys a wider 50% gap tolerance in exchange for upside. An aggressive borrower who picks m = 5 also starts fully in BTC, since 5 × $30 = $150 is capped at the $100 they hold, but the higher multiplier de-risks the basket faster once price slips and tightens its gap tolerance to 1/m, or 20%. This is the primary dial. It is a single, legible number that lets a borrower choose how much downside protection they are buying and how much upside they are forgoing. Later, the floor gives a second dial, for how large a safe reserve to keep. Currently, no on-chain lending protocol exposes either of these.

This is also where I refuse to oversell the idea, because CPPI has a famous and instructive failure mode. The portfolio insurance strategies blamed for amplifying the Black Monday crash on October 19, 1987, were the synthetic put and dynamic hedging programs that sold index futures into the decline, creating a feedback loop. Reputable postmortems, including the Brady Commission, treat portfolio insurance as having amplified the crash rather than caused it.On the day of the crash, portfolio insurers accounted for roughly 40% of all non-market-maker futures sell-off. Strictly speaking, CPPI was not the named culprit, since the 1987 programmes were the option-replication variety. But CPPI shares the same de-risk-into-declines reflex and the same gap risk. The lesson is not that the mechanism is broken. It is that gap risk is real, that you must set the multiplier with respect to how violently the asset can move between rebalances, and that you cannot rebalance through a true discontinuity.

This is live, working machinery in traditional finance

Far from dying in 1987, CPPI is alive and widely used, serving as a standard engine behind capital-protected products. Those products are a large corner of traditional finance, with structured-product sales hitting a record near $1.4 trillion in 2024 by Structured Retail Products’ count. CPPI issuance specifically was making a comeback that year, which makes it a current practice rather than a museum piece.

What matters to us is that traditional finance has spent decades learning to tame CPPI’s gap risk by keeping the multiplier conservative, charging a premium to guarantee the floor, overlaying options on the underlying, and rebalancing often enough to keep the gaps small. That is exactly the toolkit a protocol would inherit, and it is the same gap risk that showed up on-chain, in public, in the protocol that pioneered soft liquidation.

What the Honest Version Costs

If I am proposing this, I owe you the downsides in full, because they are not hypothetical, and there are three of them:

1. The bleed comes from trading the basket back and forth. Continuously selling the volatile asset on the way down and buying it back on the way up is, mechanically, selling low and buying high, and it erodes value whenever the price chops sideways inside the rebalancing range. Curve documents this directly for crvUSD and is admirably blunt that the loss is hard to quantify because it depends on the number of bands, the speed of the move, and how deep the collateral’s liquidity is. Their own worked example, for a position that spent more than half its life in soft liquidation, shows a loss of 6.37%, and they note that the erosion accrues both on the way down and on the way back up while you are in range. A CPPI loan inherits exactly this convexity cost. You are, in effect, short volatility, and you pay for the floor protection in whipsaw.

2. Forgone upside follows because a position that de-risks near the bottom and only partially rebuys on the recovery has sold its dip. The borrower trades the chance to ride a full rebound for the certainty of not being liquidated. That is a real trade, and some borrowers will not want it. The sharpest version of this cost is cash lock, the point where the basket has fully de-risked into USDC, and the cushion is zero. So, even a sharp rally cannot rebuild the position, and its upside is frozen. In a perpetual fund, that state is close to permanent, but in a loan, it is an exit ramp rather than a trap. That’s because the borrower can top up collateral to inject fresh cushion and re-risk, or simply sit safely in USDC just above the debt and repay at maturity. A borrower who wants more room can set a higher floor up front, so the locked position parks well above the debt, a choice I come to later.

3. Gap risk is the one that matters most, and it is the most concrete of the three. Soft liquidation is a buffer, not a guarantee, and crvUSD still maintains a hard liquidation backstop that closes the position if health reaches zero. We saw that on October 10, 2025, when the crash produced roughly $19 billion of liquidations across crypto in a single day, the largest on record. Curve’s CRV long LlamaLend market could not rebalance quickly enough through that gap, leaving it with about $700,000 of bad debt, backed at roughly 70% of stated value. Later, a market-based recovery was proposed in April 2026. This is not a reason to abandon the design, but rather proof that any honest CPPI loan must ship with a hard liquidation backstop and a funded reserve to cover the gap that the math cannot rebalance.

Why this belongs on-chain, and why it can bring fresh capital

The core of the argument turns on shape rather than magnitude, because a CPPI loan does not promise to be safer than holding the bare asset. What it changes instead is the way the risk is shaped. It converts a discontinuous, all-at-once liquidation cliff that’s complete with a penalty, into a continuous, smaller, and more predictable cost that you can see coming and price in advance.

That shape is precisely what risk-averse capital is built to buy. The entire capital-protected product industry exists because a very large pool of money will accept lower expected upside in exchange for a defended floor and no sudden holes. On-chain lending has, so far, offered that pool almost nothing except a liquidation cliff, whereas a loan whose collateral defends its own floor is the on-chain expression of a product structure that this capital already understands and already allocates more than $1 trillion a year toward.

It also widens what you can safely borrow against, because volatile collateral is dangerous in a hard liquidation world where a single wick is fatal. Whereas in a CPPI world, the collateral self-hedges as it approaches the floor, making a long tail of volatile assets more usable as collateral at a given level of lender safety, as long as the multiplier is set with respect to how violently each asset can gap.

How a protocol could actually build this

I would build it as an isolated lending market, in the spirit of the minimal, immutable market designs the category has converged on, with four pieces.

One choice sits above those four pieces: the interest rate. CPPI is independent of how the rate is set, so it can sit on top of either a variable rate or a fixed one, and a variable-rate version is entirely valid and worth building. The version I am proposing pairs it with a fixed rate and a fixed maturity because that completes the one quadrant nobody has built: a loan that is both fixed rate and self-defending. It most closely mirrors the capital-protected products from which this idea borrows.

1. The collateral is held as a CPPI-managed basket of the volatile asset and a stablecoin, not as pure volatile collateral. The borrower picks the multiplier at origination, which is the product. A conservative borrower picks a low multiplier and starts with more stablecoin and a wider gap tolerance. An aggressive borrower picks a high multiplier, keeping more upside and more gap risk.

2. A rebalancing engine moves the basket as the cushion changes. You can do this the LLAMMA way, by quoting prices that pay arbitrageurs to rebalance for you, or with an explicit keeper network executing through a batch auction or a CoW-style settlement to blunt MEV and slippage, priced off a robust, manipulation-resistant oracle. Rebalance frequency is a real parameter, since rebalancing more often reduces gap risk but raises the whipsaw bleed, and that tradeoff should be tuned per collateral asset.

3. You set the floor above the debt rather than on it, and you keep a hard liquidation backstop in the band between them. The floor sits higher because the rebalancing swaps are neither instant nor free. So, a slippage or a fast move can carry the collateral through the floor before the basket has finished converting to USDC, and if the floor were the debt itself, that overshoot would land below the debt and create bad debt directly. By placing the floor at the debt plus a modest margin, the basket is fully de-risked into USDC by the time collateral reaches that floor, and the space down to the debt acts as a protective band.

4. That band is also where the liquidator incentive lives, because if collateral pierces the floor despite the de-risking, the hard backstop fires inside the band while the collateral still exceeds the debt. So, a liquidator closes the position and takes a bonus from the remaining margin. The lender is still made whole, and bad debt only occurs if price gaps clean through the band below the debt. The margin should be modest, since a wide one de-risks too early and bleeds more upside, so its size is tuned to expected slippage and how violently the asset can gap. You also fund a reserve for residual bad debt from a spread charged to borrowers, which is the on-chain analogue of the gap risk premium that a structured products desk already charges for guaranteeing a floor.

5. You grow it the way this kind of risk shape wants to grow. Start with blue-chip collateral and conservative multipliers, where gap risk is smallest, and the floor is most defensible. Target the capital that wants yield without a liquidation cliff, which is treasuries, DAOs, and the more conservative end of allocators. Let curators build vaults on top that package multiplier and collateral policies tailored to risk appetites. The TVL story here is not degens chasing a number, but patient capital that was never going to accept a liquidation cliff in the first place, finally being offered a shape it recognises.

TL;DR

• On-chain lending is still mostly hard-liquidated, so one sharp wick can close a position that would have recovered days later.

• CPPI lending holds your collateral as a basket of the volatile asset and a stablecoin, and de-risks it toward a floor as your cushion shrinks, so the loan glides down instead of crossing a single liquidation line.

• The borrower’s dial is the multiplier (m), where risky exposure = m × cushion and 1/m is the gap it can survive. So, a higher m keeps more exposure and less gap protection, and vice versa.

• Curve’s crvUSD and f(x) already proved that soft liquidation works on-chain, and CPPI is a 40-year-old technique in traditional finance. So, the new part is naming it, driving it off the cushion, and handing the borrower the dial.

• It is not free, since it carries whipsaw bleed, forgone upside, and gap risk that still needs a hard-liquidation backstop, so it reshapes risk rather than removing it.

• That smoother shape, a glide instead of a cliff, is exactly what conservative capital already buys in traditional finance, which is the real case for bringing fresh capital on-chain.

If you build in this direction, or you think the gap risk kills it, I want to hear the argument. This is a proposal, not a product, and it gets better by being attacked.

Disclaimer: This piece was first published here.

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.

For almost nine months of 2025, every Monday on Token Dispatch opened with our Bitcoin macro edition. Price, on-chain flows, ETF prints, what the miners were doing and where the industry was in the crypto cycle. It seemed like the most obvious thing to write about. If you were a crypto publication, you would lead with Bitcoin, just as a market wrap leads with the index.

Then, one Monday, we shipped without it. We have now gone almost nine months without a Bitcoin macro column. I didn’t think much of it at the time. But now, when I look back, it helps me acknowledge how far the industry we operate in has come.

For most of crypto’s life, people cared about just one question: what is Bitcoin doing? It made sense then. But crypto has outgrown its connection to BTC price charts and ETF flows. When I hear crypto today, I think of how it is building a better financial world, running stable businesses on existing or new infrastructure, or turning into a new version of fintech that improves the way we transact or move money.

Many great innovations have taken this road in the past. The smartphone went through this when it stopped being measured against the phone call. The internet did too. That’s how crypto is graduating, too.

We have been writing a lot lately at Token Dispatch about the convergence of crypto and fintech. So, we deemed it timely to publish this guest op-ed, in which Nikshep argues that crypto no longer needs Bitcoin. He explains why Bitcoin lost two jobs this year and why it bodes well for the larger industry. Nikshep invests and builds at the intersection of AI and crypto. He has previously worked with Kalshi and Coinbase.

On to Nikshep’s story,
Prathik

Follow Nikshep on X

AI took its job as the market’s risk trade. The dollar took its job as crypto’s money. And the thing quietly tying a fragmenting, multi-chain economy together isn’t Bitcoin anymore. This is the most bullish thing that’s happened to crypto in years — and almost nobody is framing it right.

Bitcoin broke below $70,000 this week, down roughly 45% from its October high, and the timeline filled with tombstones. Record ETF outflows. The longest redemption streak since the funds launched. “Digital gold” that somehow forgot how to be gold while actual gold ripped 89% and blew past $5,000.

They’re mourning the wrong thing.

Because while Bitcoin bled, an on-chain exchange most people have never opened did more trading volume last year than Coinbase. A prediction market hit a $20 billion valuation, printing a $365 million fee run-rate. A privacy coin everyone wrote off ran 70% in a week — going up while Bitcoin went sideways. And a network you’ve probably underrated made it possible to move value privately across any chain without having to buy its token.

Crypto isn’t dying with Bitcoin. Crypto stopped needing Bitcoin.

That sentence sounds bearish. It’s the opposite. What’s actually happening is a coming-of-age: crypto is graduating from a single-asset momentum casino, where every coin was just leveraged Bitcoin-beta, into a real, dollar-denominated economy where businesses live or die by their own fundamentals, and where a new connective layer, not Bitcoin, holds the whole thing together.

Bitcoin lost two jobs this year. Here’s who took them, and here’s what’s being built in the space it left behind.

Job one: AI ate the risk capital

Bitcoin has no cash flow — no earnings, no dividend, no coupon. Its price is almost purely a function of how much money is hunting for risk and how willing people are to take it. That makes it a liquidity sponge: it swells when money is loose, gets wrung out when money is tight. In 2026, one rival is wringing it dry.

AI infrastructure spending is heading toward $700–830 billion this year, about half of the entire US investment-grade bond market, on its way to $7 trillion by 2030. It’s roughly 5% of US GDP and has recently added more to American growth than consumer spending. Nvidia alone is ~8% of the S&P 500. This isn’t a sector competing for capital. It’s a gravity well bending the price of capital for everyone.

It starves Bitcoin through three doors at once.

It took the story. Bitcoin’s deepest pitch was “the asymmetric bet on the future.” AI offered a better version — one with real revenue, demand it can’t build fast enough, and government backing — bought through the index you already own. Allocators now lump Bitcoin in with the most speculative tech; its correlation with unprofitable “story” stocks hit the 97th percentile. When two things share a risk bucket, and one suddenly has earnings, money rotates. That’s the outflow streak in a sentence.

It took the capital. The build-out is increasingly debt-financed — hyperscaler bond issuance already past all of last year, private credit to AI over $200 billion. When the best borrowers on earth flood the market with paper, capital gets absorbed from the top down, long before it reaches the speculative frontier where Bitcoin lives.

It took the rate cut. AI is physically inflationary — electricity and water up ~5%, memory up double digits. That helps pin inflation near 3.8%, which keeps a hawkish Fed at 3.50–3.75% and markets pricing in zero cuts this year. So AI doesn’t just outbid Bitcoin for capital; it manufactures the high-rate regime where liquidity can’t reach Bitcoin anyway. It suppresses it twice.

And then it got physical. AI and Bitcoin mining are the same business — turning electricity into compute, fighting over identical megawatts. AI won because a watt is worth far more in an Nvidia rack than a hashing rig. The average public miner spent ~$80,000 to make one bitcoin last quarter while bitcoin traded near $70,000 — a $19,000 loss per coin. So miners are defecting to the predator: over $70 billion in AI/HPC contracts have been signed, with listed miners on track to generate up to 70% of revenue from AI by year-end. Core Scientific is converting a 300-megawatt bitcoin site into AI data centres as part of a $10.2 billion deal. Riot leased to AMD and sold its bitcoin to buy the land. They’re dumping treasuries to fund the pivot — and they’re the same firms that secure the network.

Here’s why this is structural, not a dip. Compare it to the threat everyone fears: quantum computing. A powerful enough quantum machine could crack Bitcoin’s cryptography in minutes, maybe by 2029. Terrifying — but quantum attacks the machine, and you can change the machine: post-quantum standards are finalised, the migration soft forks are drafted, the roadmap is years. Quantum is a deadline, not a death sentence. You patch the lock. AI attacks what you can’t patch — the reasons to own Bitcoin, the capital that prices it, the power that secures it. There’s no soft fork for “something better took your narrative, your money, and your electricity.” Job one, gone.

Job two: the dollar replaced Bitcoin as crypto’s money

This is the part nobody connects, and it’s the keystone.

For most of crypto’s history, Bitcoin was the reserve asset — the on-ramp and the base pair. Fiat became Bitcoin; Bitcoin became everything else. Every coin was quoted in BTC. Every inflow hit Bitcoin first. That gravitational role is why alts used to rise and fall with it.

The stablecoin cut that cord. USDC just flipped Tether in transaction volume for the first time since 2019. Annual stablecoin volume crossed $30 trillion. The on-ramp is now fiat → USDC → whatever you want, and Bitcoin is nowhere in that path. When Polymarket rebuilt its exchange this spring, it launched its own dollar — Polymarket USD, backed 1:1 by USDC. Hyperliquid settles in dollars. The dollar has become, in one analyst’s phrase, money beneath app-level money — the reserve collateral under everything, with each app stamping its own label on top.

So when the market turns risk-off, the dominance charts show Bitcoin’s share falling while stablecoin share rises. Money isn’t fleeing crypto. It’s rotating into the dollar — inside crypto. You no longer have to hold Bitcoin to “be in crypto.” The dollar does that job now. Bitcoin can fall forever and never get a bid from on-chain activity, because on-chain activity runs on dollars. Job two, gone.

Follow Nikshep on X

What’s Thriving without BTC: Real Businesses

Strip Bitcoin out and look at what people actually use. These aren’t lottery tickets waiting for a BTC pump. They’re businesses.

Hyperliquid is the proof that should end the “crypto is dying” conversation. An on-chain exchange that trades like Binance — deep books, fast fills, self-custody — it did roughly $2.6 trillion in volume last year, more than Coinbase’s $1.4 trillion, generating $0.8–1.3 billion in annualised revenue. It routes 97% of fees into buying its own token on the open market — about $1.3 billion so far, ~7% of market cap a year, four to five times Ethereum’s burn rate, fourteen times Solana’s. No VC. A community airdrop and a buyback funded entirely by people using the thing. Its demand driver — traders trading — is bidirectional and BTC-agnostic. Volume can climb while Bitcoin falls.

Polymarket is the second: a $20 billion valuation, $25–30 billion in annual volume, a ~$365 million fee run rate, ~150,000 daily users (2.5x in five months), its own dollar, and a token coming. Its product is betting on elections, sports, and world events. None of that demand needs Bitcoin to go up.

These are now valued like companies: by revenue, users, and multiples. That is what a maturing market looks like.

The new differentiator: privacy — and two ways to get it

If Bitcoin’s transparent, surveilled ledger was the old default, privacy is the new premium. Something you can only get on-chain -- sovereign, untraceable money. But there are two completely different ways to buy it, and the difference is the whole point.

Privacy you own. That’s Zcash. ZEC ran ~70% in a week toward a ~$10 billion cap, up over 4,500% from its 2024 low — decoupled from Bitcoin, as its own move while BTC sat flat. It’s backed by mechanics, not hype: shielded supply has crossed 30% of all ZEC (from ~11% a year ago), and shielded coins tend to stay shielded, shrinking float as demand rises. The regulatory pressure that was supposed to kill privacy coins is the thing fueling them. Robinhood listed it; Grayscale filed the first privacy-coin spot ETF. Privacy went from use case to thesis.

But ZEC is still a coin you buy and a chain you move to. Which brings us to the more important model.

Privacy you use — anywhere. That’s NEAR, and it’s the most underrated idea in this entire shift.

NEAR isn’t asking you to buy a privacy coin or migrate your assets. Through chain signatures, a single NEAR account can sign native transactions on Bitcoin, Ethereum, and Solana — real BTC, real ETH, no bridges, no wrapped tokens — controlled by a decentralised MPC network. Layer on confidential intents, and you can move value across any chain privately, with counterparties and routing hidden, executed through a private shard. You keep your assets on your chains and get usable privacy as a layer on top.

That’s a categorically bigger claim than any single privacy coin. You don’t need ZEC. You don’t need to leave Ethereum or Solana or Bitcoin itself. Privacy stops being an asset you hold and becomes a property of how you transact, everywhere at once.

The unifying layer for an economy Bitcoin no longer dictates

Step back and look at the shape of the thing. The crypto economy isn’t consolidating — it’s expanding and fragmenting. Dozens of chains. A dollar base layer underneath all of them. And now a new class of participant entirely: AI agents that hold credentials, call APIs, and move money on their own.

A sprawling, multi-chain, agent-inhabited economy needs connective tissue. For a decade, that connector was Bitcoin — the reserve asset everything routed through. That role is now vacant. And the thing filling it isn’t another store of value. It’s a coordination-and-privacy layer: sign anything on any chain, settle in dollars, transact without broadcasting your every move, and let agents do the same.

NEAR is positioning to be that layer. It already lets agents pay each other in USDC with the counterparties hidden, runs confidential compute inside hardware enclaves nobody — not even NEAR — can see into, and turns its signing network into a key-management service for the agent economy. It’s usable privacy and cross-chain control for humans and machines, without forcing anyone onto a single chain or coin.

The consumer face of the same shift is Venice — a privacy-first, uncensored AI app with actual web2 users who just want private AI, wrapped in a token economy where staking VVV earns a share of inference via locking VVV and minting DIEM, tokens that represent a dollar a day of compute in perpetuity. It’s burned over 40% of its supply through usage-funded buybacks. Its demand curve is the number of people using AI. Bitcoin’s price is not an input. The token price and product growth clearly show it’s decoupling.

This is the new centre of gravity. Not a coin. A layer — and an economy that hangs off it, along with real businesses driving real value.

The Reframe

Put it together, and it resolves cleanly. The dollar is the cash. Protocol tokens — HYPE, POLY, ZEC, NEAR, VVV — are the equities. Coordination and privacy layers are the infrastructure connecting it all. And Bitcoin? Bitcoin is one node. The “digital gold” node is currently out of favour, because AI took the macro trade, gold took the safe-haven trade, and the dollar took the reserve trade.

For ten years, the only question in crypto was “what is Bitcoin doing?” because everything was Bitcoin-beta. That era is over. The new question is the one you’d ask about any company: does it have revenue, users, and a token that actually captures the value it creates?

The Takeaway

Stop watching Bitcoin to read crypto’s health. You’re staring at the wrong instrument.

Watch revenue. Watch users. Watch which networks are becoming the rails everything else runs on — the layers that let value move across any chain, privately, settled in dollars, usable by humans and agents alike.

Bitcoin lost two jobs this year. AI took the macro. The dollar took the money. And the connective tissue of everything left is being built by protocols Bitcoin can no longer dictate.

Bitcoin breaking $70,000 isn’t the end of crypto.

It’s the moment crypto stopped needing it.

Not financial advice. Small-cap narrative tokens are exactly where the gap between “great network” and “the token captures the value” is widest — and where the losses concentrate. Do your own work.

P.S. This piece was originally published here.

Follow Nikshep on X

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.

I think one of the hardest things to do in a market like this is to simply step back. Not step away, but just step back. There’s a difference, and unfortunately, few understand it. Stepping back requires you to still be aware, not shut down and see the whole canvas instead of the close-up view in front of your face.

Right now, the close-up view is incredibly loud. AI is reshaping entire job categories in real time. Companies that were private-market darlings are finally filing to go public, and the world is busy pricing them. Every other week, there’s a new headline about what AI will replace next. But there’s also the debate if this wave of technological disruption is the most important thing to happen in a generation or another metaverse moment that burns bright before it fades away . Remember when the entire conversation was about the metaverse? Major banks from DBS to JPMorgan were buying virtual land. That wasn’t very long ago.

The problem with these phases is that market excitement makes everything feel urgent and permanent. That makes it nearly impossible to separate what’s here to stay from what’s just passing through. The instinct to zoom out and get a bird’s-eye view of what’s happening is the hard thing to do, but it’s also the most crucial thing.

In today’s piece, Smac, a partner at Compound, offers a useful framework for understanding why these rolling waves of conviction keep forming using an excellent weather analogy (my favourite part of the piece). This weekend’s op-ed is unusually long, but that also explains why we carried it nevertheless.

On to Smac’s story,
Prathik

P.S.: This piece was originally published on Semi-Lucid Thoughts.

Follow Smac on X

Weather is a uniquely interesting field because all of the tools we use to accurately forecast the future have improved over the last 50 years. And as you’d expect, our ability to predict the weather has similarly improved. Today’s 5-day forecasts are now equally as accurate as 1-day forecasts were 30 years ago.

I think most people imagine weather as one big system that passes through. Clouds roll in, it rains, it stops raining, the sky clears out. I’d guess if I told you to imagine a winter front coming in, you’d picture some big, wide grey ceiling that lowers itself over a few hundred miles and lays down a blanket of snow. Meteorologists call this stratiform. Basically, layer-cake weather, with everything underneath getting the same treatment.

But this isn’t the only type of weather pattern. If you’ve ever watched a thunderstorm form over the plains in the summer, it behaves quite differently. Instead, what happens is you’ll get a single cell that builds (i.e. warm humid air at the ground rises into colder air above, the moisture condenses, and you get these crazy localised cloud towers) and within an hour there’s hail/lightening/rain so heavy you can’t see 100 feet in front of you.

That cell matures, dumps its energy, and then, importantly, it dies. The downdraft (cold, dense air falling out of the storm) fans out across the ground in every direction, moving as fast as 40mph, and when it slams into the surrounding, warm, humid air that hasn’t stormed yet, it acts like a wedge that shoves the warm air upward.

If there’s enough instability in the atmosphere, the wedge is exactly what’s needed to ignite the next storm cell, some ten or fifteen miles away from the first one.

The new cell was never going to form on its own. Sure, the energy was already loaded and waiting, but what it needed was a trigger, which is exactly what the dying storm provides. Then the new cell just does the same thing.

When you string these cells together, it forms what’s called a mesoscale convective system. From the ground, you experience one storm at a time, each one feeling like the entire weather system. One side of you is dead calm, completely oblivious to what’s about to come, while the other side has already cleared out. If you looked at this from a satellite, though, you’d see what looks more like a chain of events with a glowing line of separate cells, each at different phases of its life cycle, walking across the plain until they run out of warm air to feast on.

What makes these types of chains possible and different from single storm fronts is a symptom of the atmospheric conditions. Specifically:

• A boundary layer at the ground that is warm & wet (i.e. fuel)

• Cold & dry air above that makes warm air want to rise into it (i.e. instability)

• Winds at different altitudes blowing in different directions, which gives the storms rotation to move sideways (i.e. shear)

When you get all three of these, you get the procession-style storm I talked about.

Ok, so why did I just talk about the weather for 500 words?

I think the phenomenon described above is almost exactly what’s happened to financial markets today.

We used to live under a stratiform weather regime: one bull market, one bear market, slow rotations of leadership inside of it, with a few years of each. We had a secular bull market from 1982 to 2000. A dot-com bubble after this. A housing and credit cycle from 2003 to 2007. These were all long, layered and mostly legible from the ground in such a way that you could be wrong about timing for years and still come out fine if you understood the regime.

That’s just not the world we live in anymore. Today, we live under a convective chain regime: a procession of storm cells where each one feels total and unrelenting to whoever’s standing underneath it. Capital flows out of the dying narrative and ignites the next surge elsewhere (often in an adjacent area). The regime itself jumps from sector to sector on much shorter timescales, measured in even shorter time bands. AI infrastructure¹. GLP-1s. Stablecoins. Quantum. Nuclear. DATs. Robotics. Space. Each of these has its own complete weather event, with its own complete community of true believers, its own complete narrative arc and inevitably its own complete decompression. The cold downdraft then fans out and lights up the next cell a few miles away².

It’s almost silly to pretend this isn’t the reality we live in. People love to poke fun at anyone who says “this time is different”, but I think failing to acknowledge that the atmosphere of financial markets has shifted in a permanent way is either (i) intellectually lazy or (ii) a stubbornness of wanting the world to be one way rather than acknowledging the way it is.

Not Your Daddy’s Weather

For most of the post-war period, American financial markets operated like one very slow weather system. Bull markets lasted 10, 15, or 20 years, and sectoral leadership rotated within those secular trends.

The distinction here is that those rotations occurred within a coherent macro regime that only really flipped on specific generational pivot points.

The end of Bretton Woods. Volcker’s disinflation. The dot-com peak. The GFC.

I don’t necessarily think it’s a secret that there were structural conditions that made this the case. Trading was expensive³ which kept retail participation small and forced them to be patient. Pensions dominated household retirement⁴. The S&P 500 was a pretty balanced mix of manufacturing, energy, banks, retailers and a handful of media and tech names; the largest businesses produced earnings that grew with the economy in roughly linear and more predictable ways. Information also moved slowly. The gap between a 10-K filing and most investors knowing what was in it was measured in weeks.

At the same time, volatility was more symmetric. Bull markets were followed by significant corrections with drawdowns that flushed out leverage and reset positioning over longer periods of time. We didn’t see nearly the same degree of these kinds of violent sell-offs and V-bottoms. Bear markets similarly had long rallies that took time to play out. The market frequently spent meaningful time at different levels of agitation, and ~generally~ the transitions between regimes were measured in quarters or years.

If we revisit the atmosphere analogy, there was modest fuel, moderate instability and low shear. Broad, slow, layered weather that you could plan around. Every one of those conditions has now shifted with several actually inverting, and the result is a structural transformation of the conditions themselves.

What Changed?

A lot of these shifts ultimately overlap and amplify one another, but, to be honest, I think each has been large enough to reshape markets on its own. Taken together, they explain much of what has taken many smart people a long time to digest (some still haven’t, we will get to them later).

I count 8.

• The democratisation of the speculator class

• The infinite bid

• Passive investing & the inelastic counterparty

• Pod shops, HFTs & the hollowing of the middle

• Volatility suppression

• Composition mix

• The collapse of information lag

• Fiscal & monetary regime change

The democratisation of the speculator class

These aren’t necessarily in order of importance, but at the top of my mind lately has been the undeniable shift of who is actually in the market. In the 1990s, retail accounted for ~10% of US equity volume. There are a lot of reasons for this, but most retail investors held individual stocks for years. If they ever traded, it was through full-service brokers who charged commissions that made anything resembling active speculation pretty prohibitive. Robinhood famously introduced zero commission trading and pioneered the payment-for-order-flow model that made the economics work. But even that has been misremembered a bit because Schwab didn’t cut its commission to zero until the fall of 2019, and this was the real domino, as we saw Fidelity, TD Ameritrade & E*Trade, all follow suit within weeks.

Obviously, the pandemic accelerated the trend: stimmy checks, bored people stuck at home, and mobile-first trading apps intentionally designed to gamify the experience. Retail share of US equity volume jumped to 25% in 2020-21. Although many people thought this was a temporary blip at that time, it has remained elevated ever since. JPMorgan reported a record 48% of total order flow on April 29, 2025, during the tariff-driven volatility spike. On a normal day now, retail accounts for more than double its pre-2020 share, and on especially volatile days, it can reach as high as 35% of volume.

A related, and perhaps deeper, shift is what retail actually trades. Single-name options have become their expression of choice, and we all know the insane explosion of 0DTEs. The marginal new participant is younger, more concentrated, more thematic and maybe(?), most importantly, both (i) more leveraged in convex ways that don’t show up in margin debt statistics & (ii) far more reactive to price action than to fundamentals. They’re also way more likely to be tailing someone else⁵.

The boundary layer is wetter and hotter than it’s ever been, meaning the potential energy at the ground swells in a way that it never did before.

The infinite bid

This particular shift is one I’ve written about before. For more context, you can read that blog, but the TL;DR is that U.S. retirement plans have shifted from defined-benefit pension plans to defined-contribution ones. The individual, for better or worse, is now responsible for their own retirement in a way they weren’t historically. At the market level, this just means that every pay period, there’s a huge, mechanical, price-insensitive bid for equities. It’s autopilot flow.

Pensions didn’t work like this because defined-benefit pensions were managed against liabilities, which means managing duration risk. In practice, this meant that the people managing pension plans had to decide at times that “hey, equities look expensive here, we should shift the mix toward bonds”. Even if the flows were slow here, they were still more active than the passive infinite bid we see today.

This matters because we’re mostly interested in understanding what the marginal active dollar is doing, and today’s marginal dollar has more leverage on price-setting than it did in the past.

Passive investing and the inelastic counterparty

Passive investing by construction is an inelastic counterparty in the sense that it doesn’t care about price. It buys the index in proportion to weight, full stop. As a stock gets larger, passive buys more of it, and vice versa. The weird thing about this construction is that it sort of embeds momentum as infrastructure. The things working best get the largest mechanical bid moving forward, and this is at least partly how we end up with the Mag7 phenomenon⁶.

Plenty of people have written a lot about the concentration at the top of the indices and how much it has swelled over the last decade⁷. But I don’t know that the concentration is necessarily unjustified⁸. The point is that passive flows don’t really have a natural off-switch.

Pod shops, HFTs & the hollowing of the middle

So the market has introduced a permanent, passive price-insensitive bid. At the same time, we’ve seen the active layer transform in important ways, the defining structural change being the rise of the multi-strat pod shop. Citadel, Millennium, Point72, Balyasny, etc. These places house hundreds of independent PMs, each running a specialised strategy and operating on strict, tight risk leashes. AUM growth for these entities has been hilarious, and honestly, capital is concentrating in the top names here, much as it has at the equity index level.

Combine this with the fact that High-frequency trading (HFTs) now accounts for 50-60% of U.S. equity trading volume and 75% of futures volume, and you get a uniquely fragile environment. Intermediaries trading with themselves aren’t functioning as a price discovery mechanism. So a lot of the “volume” we see on the tape is just plumbing. The ultimate effect of this dynamic is that you get extremely tight spreads in normal conditions, which is obviously a good thing. But when something actually moves, i.e. a narrative cracks or positioning has built up too far in one direction, or when a pod shop’s risk limits start tripping in correlated ways⁹, the microstructure basically gets out of the way. The pod shop structure makes this worse in a very specific way because every PM across every platform is running the same factor exposures and has the same drawdown limits. So when one team is forced to de-gross, everybody else is too. February 2018. August 2019. March 2020. August 2024. These will keep happening because the structure that produces them is now baked into the financial system.

In a lot of ways, the guys who got squeezed out in this world are the traditional fundamental L/S managers who have some theses, a 20-40 name book and multi-quarter time horizons. That model got pushed either up into the platforms or drifted sideways into private markets, family offices or concentrated single-strat funds. I’m biased, but there is likely meaningful alpha to be generated by understanding the narrative shifts and having the patience to sit between marginal new flow and the marginal new exit.

Volatility suppression

Ok, so we’ve got these 4 ingredients:

• how wet boundary layer (democratising speculative class)

• perpetual passive bid

• inelastic passive counterparty

• pod-shop microstructure with synchronised risk discipline

Now the volatility regime almost has to look the way it does today. Empirically, we know the VIX spends most of its time at low levels: somewhere around two-thirds of trading days since 1990 have closed with a VIX below 20. Day-to-day autocorrelation of VIX is up to ~85%, meaning that today’s volatility looks almost exactly like yesterday’s. But the transitions are now violent and asymmetric. There are studies that suggest most of the volatility explosion occurs in the days immediately following a compression break (duh!), but that the decay back downwards takes much longer and is spread over a handful of weeks.

There are some structural reasons for this: a huge industry now exists for selling volalitity¹⁰. But inevitably, tail risk shows up, and everyone unwinds all at once. The 0DTE proliferation makes this even more complicated because dealer hedging of these products creates gamma profiles that further suppress intraday volatility.

This was a lot of words that really just say the distribution of volatility has gotten worse. So the market spendsmore time in compression phases and then pays for that with bigger, more violent releases.

Composition mix

The sixth shift is the actual composition of the market. In 1980, the S&P 500 was a manufacturing-heavy index. Industrials, materials, energy, financials and consumer staples made up the bulk of it. These companies produced earnings that grew at roughly the rate of GDP and in an almost linear way, which meant they could be valued using approximately mean-reverting multiples. Estimating the 5-year forward earnings of Proctor & Gamble was not exactly something you could be orders of magnitude off on.

Compare that to today…

Information Technology + Communications Services + some of the tech-adjacent slices of Consumer Discretionary (Amazon, Tesla) account for more than 40% of the S&P 500. Many of these companies don’t operate on linear economics. Software was lauded famously for having near-zero marginal cost of distribution. AI is even more of a Rorschach test: AI labs are either enormous capex sinks that turn into the most valuable economic infrastructure of the next half-century or they’re some kind of delusional money pit, doomed for disaster. Estimating short-term earnings is difficult enough for these companies, let alone terminal value math, where uncertainty is the highest and prone to the most violent swings in valuation. In many ways, these companies have shifted from P&L exercises to narrative ones. Again, this means there’s alpha for investors who can take a view on the trajectory of a frontier technology, the durability of emerging (and existing) moats and the shape of markets that don’t really exist yet.

Widget factories used to add capacity in linear chunks, so the discounting cash flow models mostly converged on reasonably stable estimates. Multiples mean-reverted far more often in that paradigm. An increasing share of today’s multiples is a function of how much investors believe the story that is being told. And, for what it’s worth, I don’t actually think this means valuation frameworks are broken, but it’s just the reality of the types of companies that exist in 2026.

The new financial market indices are full of these long-duration, narrative-driven businesses. Each of these has a much steeper lapse rate¹¹. The steeper the gradient, the more potential energy is locked in the system and, by extension, the more violently that energy converts to motion when something triggers it.

The collapse of information lag

Everyone has lived through this in some way and probably, intuitively, understands it. So, it’s perhaps less interesting but also maybe now under-appreciated. Throughout most of financial history, any kind of market-relevant information was rate-limited by some speed-of-publication bottleneck. Today, information latency is literally zero.

Positioning information, in particular, seems to propagate faster than anything, as investors immediately see how every “smart” person they respect reacts to news in real time (more than ever, a lot of people are disclosing positions as well). Part of this onslaught of real-time information is a constant buzz of comparison brewing under the surface. PnL screenshots are everywhere, stories of someone turning $1K into $5M regularly go viral. There’s a permanent engine of FOMO roaring louder every day.

Fiscal & monetary regime change

I actually don’t think this needs to be explained in detail at this point. So the cliff notes are

• US monetary policy is structurally easy with real rates near zero for a long time

• QE expanding Fed balance sheet by trillions

• Every duration-sensitive asset traded at a premium, with the discount rate so low

• Then, fiscal added to this with stimmy checks, PPP, Inflation Reduction Act, CHIPS Act

• War-time deficits while at full employment

• K-shaped economy dynamic, as stimulating financial markets & stimulating the main street economy are very different things

How The Storm Forms

It’s much easier, given all the above, to now visualise how these rolling bubbles become inevitable.

I won’t go through all these stages individually because they’re reasonably self-explanatory, but let me call some things out.

The compression stage is something we’ve uniquely seen at Compound investing across a bunch of (seemingly) disparate areas for a long time. Categories/sectors/industries come in and out of favour all the time. With that comes a lack of attention and interest. But there are always talented people building regardless of the outside perception, and the reality is that most of the time they’re not going to be in the spotlight.

Eventually, some legitimate change happens that brings the interest back around, be that a real technological breakthrough, a regulatory shift, or an earnings surprise that changes the scale of the underlying business reality. The domain specialists will should notice this first.

At some point, the narrative is crystallised (AKA, the story gets a name), which does a few things. One of which is that it flattens what’s actually happening. Some people find this annoying because they hate memes or think it’s bastardising something they hold dear, but the reality is this makes the narrative easier to understand (and, importantly, to share). It becomes way more legible to a broader audience who can’t necessarily evaluate the entirety of it on merit alone (even with LLM help).

At some point, you reach a stage where youwe see bifurcation. The modern context of this is that the marginal buyer outside “the believer cohort” shrinks to zero. You reach a point where the valuation gap between believers and sceptics becomes too large such that even those people chasing the narrative fall off as marginal buyers.

The crack is always obvious in hindsight, though in 2026, people are way too quick to try to call this ahead of time. This again is just the reality of the hyper-online, comparison, and clout-chasing world we live in. But this crack gives way to the de-grossing as factor exposure unwinds for everyone, and the outflow then goes searching for a new pocket of warm, humid air to trigger a new cell.

Future Forecast

The downstream effects of this regime are significant. And I think this is more an acknowledegment that the shape of the weather is, in theory, more predictable without necessarily knowing the exact location of any given storm.

But what does this mean about the future?

One argument that a lot of people made just after COVID was that “this is an anomaly” or “[redacted] was a zero interest-rate phenomenon”. Some of this was obviously true. But to me, it’s become clearer that the broader regime shift is permanent. How permanent is a debate for another day. But each of the shifts I described above is pretty one-directional…

• trading commissions are not going back up

• passive investing doesn’t just unwind

• defined-benefit pensions are gone forever

• social media isn’t going to get slower in the future

• pod shops — maybe these dissolve quicker than I appreciate, but given the parabolic growth and how much money they’re making, that seems unlikely

• information lag isn’t going to lengthen

You can almost think of these conditions as the new climate we live in. Candidly, I think anyone expecting us to return to the previous stratiform regime of the 80s or 90s is just in denial.

Another claim that’s maybe(?) finding some kind of local maxima is the idea that these rolling bubbles will keep compressing in duration. This is one of those things that I think is tougher to say just because it turns into a never-ending meta game of “I know that you know that I know that you know…”. But the simplest lesson is that each cycle teaches participants what the pattern looks like, which in theory speeds up the next one. Crypto participants now feel comfortable trading TradFi assets because many of the survivors there have so many reps already with these types of behaviours. And while I don’t necessarily subscribe to the idea that crypto traders are best suited for this new world, I understand the argument. Some probably are. To me, though, there is some half-life of narrative-driven markets that can only be compressed so much.

Maybe the most critical claim I’ll make about what the implications are is that this new rolling bubble world rewards a couple of types of investors. Domain specialists who actually understand what would need to be true for the narrative to deliver on its priced-in promises (or exceed them). But this means understanding true technical bottlenecks, regulatory nuance, manufacturing supply chain dependencies, and competitive dynamics, all the way down to unit economics. Realistically, this is not where most people will sit. I think AI models probably make people believe they live in this category more than they actually do, which is especially dangerous¹². The reality is most actually sit in an adjacent seat where they are trying to understand how the next-most-informed participant will react. There’s probably a completely separate piece on why the combination of these two is likely to become an even more rare trait over the years, but this blog is already long enough.

One thing that’s clear is we are certainly not at a loss for future narratives:

• AI infrastructure & applications

• Robotics

• Physical AI

• Precision medicine

• Crypto

• Materials science

• Fusion, advanced fission, grid storage

• Space

• BCI

• Quantum

Even within these, we’ll see ripples form and dissipate as investors move up and down the supply chain and tech stack within these.

Retail’s biggest advantages in financial markets have always been time and flexibility. That is to say, they don’t need a few days to schedule investment committee decisions or convene on allocation sizing. They also aren’t beholden to the same shorter-term pressure that many institutional managers face (i.e. there’s no quarterly redemption in your PA). This may explain some of the outperformance retail has enjoyed. Decades of BTFD working also probably helps. So, in some sense, retail participants who can manage risk are in a great position.

The View From Above

A lot of what I’ve (too verbosely) written above may leave the impression that I care one way or another about the things that have led us to the market structure we have today. Sure, I have opinions on some of those things. And when investing on the private side, I still think there are plenty of ways to make money without funding things that accelerate net-bad societal outcomes. But one of the most common mistakes many people make in public markets is making decisions based on how they want the world to be¹³. I get this. We’re all humans with emotional triggers that have been hard-coded into our DNA for centuries. It’s why Newton re-entered with a lot.

But emotional triggers are a huge barrier to performance. For years, we’ve seen asset managers and hedge fund guys come on TV and talk about why markets are wrong or wax poetic about impending recessions. Much like our friend Sisyphus, these guys do not come on TV for charitable purposes. Ray Dalio has been calling for a debt crash throughout this historic bull market.

The world is not going back to your father’s market. Is this convective regime a more degenerate version of the old stratiform one? I’m not actually sure, to be honest. Potentially? But at the same time, a lot of the shifts that are producing it I actually think are net-good things: democratised access¹⁴, automated retirement saving, passive ownership, more real-time information. All of these things reflect a market that is way more available to ordinary people than it was in the past.

An aside that I tweeted about the other day (there’s a more philosophical piece to be written here):

From the ground, every storm is total. The world narrows to whatever is happening to you right now. This is the experience of market participants in any of these sector cycles. On the ground, it feels like a black hole of liquidity sucking up everything around it. Only if you are able to step back and move to an altitude, are you able to actually see the chain. One cell after another, lighting up. Each of them at different phases and each of them feeling total and complete to the people underneath them.

Markets are so intoxicating because they are simultaneously always changing, and yet, at least today, humans still have control over price discovery. And humans are inherently emotional beings that can’t help but make the same mistakes over and over. That tension helps create the conditions we experience each day, which ~feels~ quite frenetic in the moment but, taking a step back looks a lot more like rolling bubbles.

Really, the point of all this is to try to drift above the immediate storm and operate at a higher system level to allow yourself space to recognise which way the front is moving, and to detach yourself as much as possible from the immediate sensory-overload effects of any individual cell underneath.

As with most things that require immense discipline, this is simple but not easy.

References

1 which within itself has its own rolling narratives

2 one of the funniest things about this is seeing people on Twitter say things like “nobody is even talking about xyz yet” and it’s an entire massive sector that has tons of people talking about it they just don’t happen to overlap at all with this random narrative trader

3 a round-trip stock trade through a full-service broker costs roughly the equivalent of a steak dinner in real terms

4 in 1980, almost ⅔ of private retirement contributions still went to defined-benefit plans which meant most of the wealth being saved for retirement was deployed by professional fiduciaries operating on multi-decade liability horizons

5 https://www.finra.org/media-center/newsreleases/2026/finra-foundation-research-examines-characteristics-behaviors-outcomes

6 obviously, these companies are also exceptional at making money and growing earnings

7 there is some truth to it, but i also think it’s not necessarily that fair to think of some of these companies as 1 company

8 guy who thinks you’re not allowed to use em dashes anymore bc AI’s corrupted it

9 btw these guys are in all the same trades

10 covered call ETFs are a massive category, dispersion trades (short index vol, long single-name vol) are a core book at every multi-strat, structured products from banks systematically harvest premium. all of these things just manufacture the calm by collecting income in exchange for taking tail risk

11 the steepness of the temperature gradient between the surface and the upper atmosphere is the lapse rate

12

13

14 fwiw Robinhood is one of my least favourite companies in the world. I think the fact that they pretend they’re helping people on some type of financial journey, but in fact design the entire company/product around turning their users into degenerate gamblers is gross.

Disclaimer: This piece was originally published here.

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.

Our team at Token Dispatch has Friday editorial calls, where we talk a little bit about a lot of things. We also discuss what’s exciting and worth writing about in the upcoming weeks. Lately, the conversations sound less and less like crypto conversations. We end up talking about lending businesses, AI subscription models and payment rails that Stripe and Mastercard are fighting over. Last Friday, we spoke about how the upcoming trillion-dollar IPOs of OpenAI, SpaceX and Anthropic will shape the broader financial markets. Even if someone brings up a crypto project, halfway through the discussion, you’d realise nobody has mentioned ‘token price’ even once.

This shift is also reflected in what we have been publishing lately. In the last couple of fortnights at Token Dispatch, we have moved toward stories that sit at the edges of crypto. Think of fintech that uses blockchains as plumbing, consumer products where the token is a distribution mechanism rather than the product itself, infrastructure companies getting acquired at valuations that have nothing to do with the cycle. The developments that keep moving, whether BTC is at $100k or $70k.

Originally published by Hepworth Iron Capital, this week’s piece puts a framework in place for this phenomenon. Charlie Booth argues that crypto’s era as a single Bitcoin-sensitive factor is ending, paving the way for a cycle driven more by non-crypto factors and not cryptocurrency prices.

Onto Charlie’s story,
Prathik

Follow Charlie

Follow Hepworth Iron Capital

Historically, crypto has traded as one Bitcoin-sensitive factor. That era is ending.

The crypto economy is bifurcating into two groups: endogenous and exogenous.

The former is the crypto of old: tokens and projects whose value depends on crypto prices. The latter is crypto in name only, with value increasingly exogenous to crypto prices.

Bitcoin’s value comes from its properties and reflexively from its price. A rising price reinforces the perception of the properties. At the peak of a bull market, Bitcoin is perceived to be interplanetary money, the scarcest digital bearer instrument known to mankind. At a bear market bottom, it’s dismissed as a digital collectable without any cashflow.

Hyperliquid sits in between these endogenous and exogenous groups. Much of its business still depends on crypto prices, but both the supply and demand sides are broadening. A lot of on-chain financial infrastructure sits here, with the assets underneath shifting toward tokenised real-world assets.

HIP-3 Open Interest is a rough proxy for non-crypto-related Open Interest. The percentage of total Hyperliquid Open Interest attributable to HIP-3 sits at ~30%, up from ~4% in November 2025. HIP-4 (outcome markets) should further boost this, pulling in new demand (traders) and new supply (markets, assets) at once.

On the purely exogenous side, the driving forces behind a project like Venice are completely exterior to the crypto market. Customer profiles overlap, but the business model looks a lot more like consumer AI than Uniswap. The latter still largely relies on users trading assets with endogenous value, making its business a function of those asset prices. Venice packages private multi-modal inference into a usage-plus-subscription model.

The only crypto connection is the choice of a token as the instrument that captures business value, plus the fact that some of its inference suppliers happen to wear crypto labels. Perhaps also the crypto-native understanding of Venice’s steward Erik Voorhees that tokens can be an excellent marketing tool when done well.

Figure is an easy example on the listed equity side, a fintech lender using its own chain to cut home equity approvals to under five minutes. The blockchain is incidental; the business is the point.

The emergence and growth of the exogenous class at scale in listed equity and token markets is meaningful. Historically, pure bottom-up investing has been difficult given the high sensitivity of most business models to crypto prices. It’s not as though crypto hasn’t had exogenous narratives before; every “blockchain, not Bitcoin” cycle promised them. In most cases, they collapsed back into crypto-beta because demand never materialised, revenue wasn’t there (if it was, it was not captured by the token), and once the token stopped going up, there was nothing underneath.

What’s different this time is that you can answer who pays and why, that demand is measurable and less reflexive in many cases, and that the token as an instrument is slowly improving (more on that later). Venice’s signup revenue is real money from users buying inference. There’s no obvious reason it reverses when crypto draws down, because it was never a function of prices. You have two things that prior cycles lacked: usage that persists, and buyers underwriting on fundamentals rather than solely on narrative.

Consider the stablecoin corner of private markets. In March 2026, Mastercard agreed to buy BVNK for up to $1.8 billion, 15 months after BVNK’s Series B closed at a $750 million valuation. Bridge (acquired by Stripe for $1.1 billion in February 2025) is growing fourfold year-on-year within Stripe, according to Stripe’s annual letter. None of these tracks the crypto cycle.

This isn’t a bearish call on the endogenous category. Just as gold and even junior gold miners might have their time and place in a portfolio, there’s a time and place for something like BTC and the endogenous class. But fundamentally different drivers will likely continue to drive performance and correlations. You can see both relationships in the data:

The analogy made literal: Junior gold miners track gold at a correlation that never really leaves the ~0.75 region. That’s how crypto broadly trades today: junior miners to Bitcoin’s gold, and levered bets on the same underlying. The blue line is the other relationship. Gold and the S&P share some macroeconomic correlation but trade on their own drivers. That’s the destination for the exogenous class. Over time, these names should migrate from the gold line to the blue line, from levered proxy to independent assets that occasionally share the weather.

These exogenous names are tokens, both illustrating the point and serving as exceptions to it.

Much of the “endogenous class” still moves closely with Bitcoin. A few exogenous names sit lower, but the window’s too short to mean anything yet. Fundamentals change first; the correlations follow.

This changes the analytical approach. The exogenous class needs underwriting like a normal business: who pays for the product, how the unit economics work, and where the moat sits. BTC price stops being the most important variable, and you start sounding like a fintech investor with weird custody.

Exciting “exogenous” categories, in no particular order and with miscellaneous notes attached:

• on-chain exchanges and brokers

• credit/redemption solutions for longer-tail tokenisation (Grove Basin looks interesting here)

• real crypto x AI (private inference, distributed open-source model training à la Psyche by Nous Research)

• neobanks (I tend to like the more privacy-focused players like Payy and Raycash, the programmable privacy infrastructure, Aztec and Zama, that enables them is also interesting)

• lending (Morpho is becoming the institutional standard for repo-like markets, smaller names like Valinor and 3jane target interesting niches in private credit)

• stablecoin and real-world asset/tokenisation issuers

• payment rails (Stripe with Tempo is the incumbent to beat in broad payment rails; in agentic payments, it’s Coinbase, for now)

• non-finance consumer crypto (products like Venice and Collector Crypt come to mind, unique cases where imbuing a token with value derived from a non-crypto business can drive marketing value and adoption)

• agentic economy (the prize is agent<>vendor/creator coordination at the access layer, which is less substitutable than the rail. Cloudflare is well placed, but whether it taxes the flow or just sells the toggle is the open question)

Most durable exposure to this theme is equity today, not tokens. Good tokens are the exception and will only earn a larger role if the instrument improves, which regulators and industry have to deal with. There’s progress on both, the CLARITY Act on the regulatory side and transparency efforts from the likes of Blockworks on the other. The token has a lot of work to do.

None of that changes the main point. The driver is shifting from one factor to many; the work is no longer reading the Bitcoin chart but underwriting businesses. Don’t spend the next decade confused about why “crypto” stopped moving as one.

Follow Charlie

Follow Hepworth Iron Capital

Disclaimer: This piece was originally published here.

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.

I am not a connoisseur of art, music, or food. But there’s one trend I have noticed about businesses that I feel gets often overlooked. The most extraordinary businesses were successful because they did not ignore some non-business things that others paid little attention to.

Apple’s competitive edge was not hidden in the chip they used but in how Steve Jobs cared about the weight of a thing in your hand. Sony’s golden run began because its founder loved music enough to consider what a portable music player should feel like.

The common pattern I found in such businesses is that they gave equal, if not greater, importance to their non-business instincts than to their business ones. These non-business instincts are still rooted in human behaviour, like hospitality, craft, cultural sensibility, and a sense of atmosphere. Often, these are the instincts that hold the entire workplace together by acting as a centre of gravity. They have the power to connect people from different geographies, cultural backgrounds, socio-economic backgrounds, and many other differences.

I think about this when I look at how we discuss technological shifts. The conversations in boardrooms are almost always about competition, infrastructure, bureaucracy, and regulation. Don’t get me wrong. These are useful and unignorable, but not holistic. It misses a crucial layer where things often get built between people. What connects someone’s ambition with someone else’s risk appetite and a founder’s context with a policymaker’s expectation is one of the oldest interfaces. Like a table, a room, or a shared meal on a nice wooden table in a room curated with the right people and the right food for the occasion.

In today’s piece, SuperAI and TOKEN2049 co-founder Peter Noszek argues that the age of AI needs its human layer built with the same intentionality as its supply chain – through bridges, dinners, houses, and rooms where trust can form.

Onto Peter’s story,
Prathik

Follow Peter's works here

Follow Peter on X

America’s version of Pax Silica is the US Department of State’s flagship effort in AI and supply chain security.

ASEAN’s framing is more concrete: a US-led initiative to build a secure, prosperous, and innovation-driven silicon corridor, from critical minerals and energy inputs to semiconductors, AI infrastructure, logistics, software, and models.

The subtext is obvious: Trump’s America is looking for trusted partners with whom it could build around China. It is adversarial by design.

But Pax Silica should mean more than securing the stack. It should describe the human layer and human implications of the AI age: the trusted relationships, rooms, and institutions that help people navigate the biggest technological shift of our generation.

The Official Frame

Securing the AI supply chain with trusted partners runs through Southeast Asia.

Compute needs power. Chips need minerals, fabs, packaging, logistics, and trusted routes. AI companies need neutral places to deploy across Asia. Singapore, as a Pax Silica signatory and regional coordination point, sits in the middle of that problem.

I’ve just landed from an immersion into the Chinese AI ecosystem via Shanghai, at a moment when Washington and Beijing were circling each other on chips and AI — the public gestures running well ahead of anything actually moving.

A few weeks ago in San Francisco, Cindy and I hosted the first Pax Silica dinner. Alvin Wang Graylin was there and shared a frame that has stayed with me since. As he later put it in the South China Morning Post: “The responsible move is to find common ground, not raise the fence.” He called this moment the “doorstep of peace” and warned that fanning rivalry between two great powers was both irresponsible and immoral.

That was not the China I met in Shanghai.

The people I met were open, curious, and collaborative. The energy was pointed outward: founders, operators, platforms, and city-level ecosystems trying to understand where their companies, products, capital, and talent should go next.

Some of that energy pools in Shanghai and Beijing. Some of it hums in Shenzhen and Hangzhou. Some of it stirs in tier-2 cities like Changsha: 10-, 20-, 30-million-person cities with serious young AI founders who are hungry to internationalise, but need trusted bridges into the right markets.

Southeast Asia’s role becomes clearer from that angle, with Singapore at the centre: a trusted landing point for AI companies moving between China, APAC, and the global market.

The Name

The name matters because it puts peace and the technology race into the same sentence.

Pax

Peace. Stability. Long-term prosperity.

Silica

The compound is refined into silicon, which is one of the base materials of the AI age.

Put together, the phrase asks a question that feels much bigger than export controls or data centres:

Most institutional answers begin with security. Secure the supply chain. Protect sensitive infrastructure. Reduce coercive dependency.

All of that’s sensible, and I take it seriously. But peace also asks for translation between people who normally operate in different systems.

It asks for rooms where a founder can speak honestly in front of capital, where a policymaker can understand what builders actually need, where an enterprise buyer can explain the risk behind a slow decision, and where someone is paying attention enough to translate between all three.

The age of AI will need secure supply chains and people who can keep building when geopolitics, capital, and deployment timelines get messy.

Relationship Infrastructure

The AI supply chain spans chips, minerals, energy, data centres, models, and networks. Every layer eventually runs through people.

A Chinese founder entering Southeast Asia needs more than market research. A frontier lab deploying in Asia needs more than policy slides. A sovereign fund looking at AI infrastructure needs more than deal flow. An enterprise buyer evaluating AI needs more than demos.

They need trusted translation.

Who understands the founder’s ambition? Who knows which institution can move? And who keeps the thread warm after the event ends?

Pax Silica is the public thesis. The operating stack underneath it has three parts: bridges move value, dinners build trust, and houses create continuity. Together, they point toward a larger gathering where the system can find itself. The mechanism is practical: trust reduces coordination cost, rooms increase bandwidth, and repeated contact improves the accuracy of the models people hold of one another.

Bridges

To bridge is to speed up the diffusion of technology, information, and value across systems that would otherwise move too slowly.

Some bridges are powered by capital. Funds can move an entire market by deciding which infrastructure deserves conviction.

Some are powered by advisory. The right operator can help a frontier company enter a new region, find the buyer who matters, and anticipate institutional friction.

Some are powered by community. In San Francisco, I saw this through the builder-house and residency graph around HF0, The Embassy, The Residency, Frontier Tower, AGI House, and Cerebral Valley. These bridges connect people who might live a few miles apart and still never properly meet.

Ray Del Vecchio, co-founder of Cerebral Valley, used an interesting phrase: “MBSF”, months behind San Francisco. It describes the lag between discovery in the Bay Area and adoption elsewhere. The same could be true in the other direction: “MBSZ”, months behind Shenzhen. A bridge shortens both lags.

China has its own bridges. Google China is one of the clearest examples I saw in Shanghai: helping Chinese companies internationalise across APAC, with Singapore increasingly important as an operating centre.

Singapore operates its own bridges as well. I’ve been fortunate to meet and work alongside people across EDB, Singapore Global Network, and Enterprise Singapore who understand that Singapore’s role is to be the trusted crossing point.

Pax Silica starts with bridges that move people across systems, and rooms that let trust compound over time.

Dinners

The smallest complete unit of Pax Silica is a dinner.

A dinner is intimate enough for trust to form and serious enough for people to show up with intention. It lets the host compose the room across layers that usually sit apart: frontier capability, infrastructure, capital, policy, enterprise deployment, culture, taste, and human-centred design.

Curation is the seating, the rotation, the music, the food, the pace of the evening, the first introduction, and the comfort people feel to speak plainly. I like to think about events as an interface. A dinner is the layer underneath: the seeded conversation, the useful encounter, and the natural follow-up the next morning.

A room works when the worlds inside it are close enough to notice each other properly.

At the Pax Silica dinner in San Francisco, several people were adjacent without having properly crossed. Some lived within a few miles of each other. Others were already moving between Asia and the US, but needed the right shared container.

In June, during Singapore AI Week, Cindy and I will host the next Pax Silica dinner in Singapore.

Houses

The house is where the corridor stops being a calendar item.

The space Cindy and I are building is a tech and cultural house at the edge of the Pacific. Teak floors. Jade tile. A long table for twenty-four. Peranakan light through tall windows. A garden courtyard where someone is writing code and someone else is cooking.

Part salon, part dining room, part gallery, part working room, part recovery space. A San Francisco node for Southeast Asia’s frontier-tech corridor, with the intimacy of a home and the seriousness of an institution.

A founder visiting from Asia should be able to land in San Francisco and find the right first table. A Singapore delegation should be able to understand the Bay Area without turning it into a study trip. A frontier builder should encounter Southeast Asia through people, food, art, and music.

This is where compassion becomes operational. Everyone who enters the room is carrying something: ambition, mandate, exhaustion, risk, family, national interest, the desire to be understood. A good room lets people put some of that down.

For Pax Silica to mean peace in the age of silicon, it has to be practised through hospitality. As Donatus Schaumburg-Lippe reminded me, embassies were built for ongoing dialogue: permanent rooms where trust could be maintained before, during, and after negotiation. Pax Silica House carries that instinct into the age of AI. The chair, the menu, the music, the guest room, the first ten minutes, the follow-up the next morning: these are how an abstract corridor becomes a place people trust.

Schelling Points

A Schelling point is where people converge without coordinating every step in advance. Everyone comes because everyone believes the others will come.

AI needs this badly. The work I care about – connecting the distinct nodes of Silicon Valley and Asia, between founders and sovereign capital, between the people building AI and the people deciding how it lands – only compounds when those people are repeatedly, reliably in the same place.

Dinners and houses do the intimate work. But intimacy has a ceiling. A dinner seats twenty. A house holds a season. Past a certain point, the bridge needs a wider span: a gathering large enough that an entire scattered ecosystem – labs, funds, ministries, platforms, data centres, universities, founder houses – can assume the others will be there, and plan their year around it.

That gathering sits a layer up, and Pax Silica feeds it. The dinners seed the conversations, and the house keeps them warm; the gathering is where they become unavoidable – where a partnership that was a maybe in April becomes obvious by June, because the right people were finally, unmistakably, in the same room.

Pax Silica’s job is to make that larger room work better: to fill it with people who already trust each other, conversations already seeded, and relationships that are already warm. The dinner is the smallest complete unit. The gathering is the largest. Singapore is where the largest one can hold – close to China, fluent in the West, trusted across Southeast Asia and Europe, serious enough that institutions show up without being asked twice.

As silicon becomes the operating layer of civilisation, peace has to be hosted – as a room, at every scale: the table, the house, and the gathering they point toward.

Pax Silica is the thesis on the human layer of the AI age. Field Notes is the running notebook behind it: dispatches from the rooms, cities, and conversations where this corridor is being built.

Disclaimer: This piece was originally published here.

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.

I hear a loose thought floating around me, used so often at this point: AI will take away our jobs and make us redundant. The bigger irony is that people across domains use this without really questioning exactly what these AI models are replacing in their line of work.

As writers, my team often discusses this idea in our internal calls. Although there are a lot of ideas about how AI will evolve from here, there’s one thought we all seem to agree on. When everyone has access to the same data, similar models and common building blocks, data no longer remains the moat. The content creation game flattens out at the level of data aggregation and research. But there’s something else that AI also does.

Every door closed shut, more often than not opens another one. Once these AI models and the software built on top of them make building things easier, they force humans to differentiate based on our ability to imagine and express diverse tastes.

What I find under-appreciated about AI-enabled software is that they strip away the excuse that building something is difficult or costly. What hurts humans more about AI is not that it takes away our jobs or makes us redundant. But it’s that they act as beautiful filters and ask a difficult question: “Do you actually have a vision that separates you from the rest? Can you build something different when the playground is levelled for everyone to build?

Today’s piece by Kinjal Shah, General Partner at Blockchain Capital and Co-founder of Komorebi DAO, takes that instinct and grounds it in what’s actually happening. It’s about the explosion of personal social software and why open protocol infrastructure like the AT Protocol ensures that small, specific, human experiences don’t have to start from scratch. She explains why the bottleneck isn’t code anymore but imagination.

Give it a read.

Onto Kinjal’s story…

Do you remember when your friendships translated for the first time? I was thirteen, ranking my “top eight” shelf on MySpace, only to realise I didn’t have eight friends I’d include. Ranking them made my stomach sink.

Fast forward to 2026, and numbers are still the fastest way for a stranger to determine your status. The platform on which you’ve gained them adds context. Your name, profession, and appearance are all scrutinised in the flick of a scroll, quickly adding up to a first impression.

What’s real or not is an afterthought. Perception is reality.

The early internet was a companion to many users, whether it was through Tumblr communities or the early blogging world. It allows anyone to expand their network beyond geographies and gain access previously held by gatekeepers.

When the algorithm became the discovery tool for default social experiences, it forced everyone into the same doomscroll. In doing so, it fundamentally changed how humans form connections online, creating parasocial relationships that are optimised for the wrong metrics.

What if it wasn’t the desire for human connection that broke — but the game itself?

The barrier just collapsed. With today’s agent functionality, the cost of creating software has collapsed to nearly zero.

But a more interesting consequence of this is the friction AI removed.

Before, if you had an idea for an app specific to your community or context, you had two options: find an existing platform that approximated it, or hire someone to build it and hope they understood what you meant. Either way, building yourself was for a technical user. It involved real friction for an app that most people might not even want.

That friction is effectively gone. Intent can now translate directly to action.

My brother-in-law recently built an app for his friends where they share weekly sports parlays. He vibe-coded it in a week, and now they use it every week. That wasn’t possible 12 months ago.

We’ve moved from personal computers to personal software.

The parlay app will never have a million users, but it doesn’t need to. It is working perfectly on the one metric that matters the most: deepening specific relationships.

The numbers confirm this.

Apple app store submission grew 30% in 2025, followed by an 84% year-over-year jump in Q1 2026, the largest surge in a decade.

This is after declining 48% from 2016 to 2024. The inflexion point correlates precisely with the release of more accessible agentic coding tools.

What’s even crazier is 63% of the vibe-coding users today classify themselves as “non-developers,” many of whom have never written a line of code.

Personal Social Software is the Future

When I think of the early web, I remember the cosy corners where lifestyle blogs, independent journalism, and the personal website proliferated. One of the biggest enablers of websites was an open-source content management system called WordPress.

As of 2026, WordPress powers nearly 44% of all websites published on the internet.

WordPress didn’t succeed because a single blog had millions of readers. Most WordPress sites have relatively low traffic.

It succeeded because it enabled millions of distinct, self-publishing experiences on shared infrastructure.

The metric that mattered was the number of websites deployed. That’s the same frame for what’s coming.

What would a world of thousands of social experiences, many of which are pop-ups, look like? Each is built for a specific group, ritual, context or use case.

A weekly parlay app for eight friends. A darkroom photography community for a neighbourhood. A shared reading log for a family. A group planning tool for a climbing crew.

None of these needs to win the internet. All of them deepen specific relationships in ways that no generalised platform can, because none of those generalised platforms was designed with those specific people in mind.

When software doesn’t need to scale, it allows for personalisation and community formation.

When the cost of building a social experience approaches zero, the bottleneck shifts from infrastructure to imagination.

Why AT Protocol Matters as Infrastructure

The problem with building your own social experience from scratch is that you start from zero. Your social graph lives somewhere else. Your identity lives somewhere else. You’re effectively building on sand, where the ground can shift quickly under your feet.

This is what killed most of the ‘build your own community’ tools of the last decade. The only audience you’ve been able to own is your email list. Every new network asks users to start over - new account identity, new followers, new context. AT Protocol breaks this constraint.

When you build on AT Protocol, you’re not building on a platform. You’re building into a network that users already belong to. Their identity travels with them. Their followers travel with them. Their context travels with them. The parlay app isn’t a closed garden but a node in an existing social graph.

This means a builder who creates something for eight friends doesn’t have to solve the cold start problem. The network is already there. When someone discovers the parlay app and joins with their existing AT Protocol identity, their connections see that they joined and the context compounds across deployments rather than within each one.

Why Should You Build on AT Protocol?

Shared identity on the AT protocol. Social graph portability lets you bring your users with you. Moderation primitives let you focus on the product rather than on trust- and-safety tools. A whole suite of à la carte building blocks is coming to you.

The things that take months to build from scratch, AT Protocol gives you right away, out of the box.

Try AT Protocol

Attie, the feed-building tool the Bluesky team recently launched, is a live demonstration of what this infrastructure unlocks. It’s a portal into creating your own feed experience on top of the underlying network. What you find there is a vibrant set of users already organising events around science and academia, sports, news, analogue hobbies, and niche interests that you rarely see spaces made for. The like-minded users are already in the network. The tools to build for them are already available.

There are some really unique examples being built today.

One of my personal favourites, Anisota, created a gamified social interface where each post becomes a collectable trading card with rarity ratings and stamina limits to discourage compulsive scrolling. It makes social feel like play. It’s built on AT Protocol, which means it inherits the network rather than having to recruit it.

Tangled is another example, building a social coding platform where coding is a social, owned experience rather than a solitary professional one.

These aren’t competitors to Bluesky. They’re what a healthy protocol ecosystem looks like: multiple distinct experiences, all running on a shared identity and data layer, each finding the specific community it was built for.

The Open Question

The thing I keep coming back to is if AI makes creation cheap for everyone, doesn’t everything converge toward the same AI-generated aesthetic? If the tools are the same, don’t the outputs homogenise?

I don’t think so, and here’s why. The inputs to a 1-of-1 social experience are human taste, specific relationships, and local context. Those don’t homogenise. The medium becomes generative, but the intent stays personal. My brother’s parlay app is shaped by the specific way his friend group talks about sports, the running jokes, and the particular format of their picks. No amount of AI tooling makes someone else’s version of that app feel like his.

The broader ‘human-only’ framing emerging across new social apps is responding to this desire to prioritise human connection. When synthetic content saturates the feed, genuine human context becomes a scarce resource. We are seeing this permeate our everyday interactions. Recently, the World launched a “humans only” concert to address the challenge of bots sniping tickets online.

We’ve also seen this in our digital spaces, on apps like Retro or Subtext, which emphasise private social for humans only.

1-of-1 social is, among other things, a structural defence against that. It’s a social space designed for specific people, making it legible to them in ways that generic content can never be.

Build Something

Cosy corners of the internet are coming back. Not as a nostalgic throwback but as a return to the fundamentals that made the internet a meaningful place for connection. This is the structural consequence of tools that enable personal software.

The social graph can be open. The identity layer can be portable. The moderation primitives are available. The barrier isn’t technical anymore. It’s imagination.

Think about the ritual your group already has that no app serves perfectly. The shared behaviour that happens in spite of the tools you’re using, not because of them. The group chat that’s doing work it was never designed for.

That’s the brief. Build the thing that makes it better — and build it on infrastructure that doesn’t make you start bringing together your users from square one.

That’s it for today,
Kinjal Shah

Try AT Protocol

Try Attie

Disclaimer: This piece was originally published as an X article.

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.

As a writer, I have been pleasantly amazed by what Substack has become over the past few years. It’s not so much what the platform chose to do that made me stick around as what it chose not to do. Substack doesn’t crowd my screen with engagement metrics or algorithmic noise. It doesn’t turn every interaction into a performance. Each time I open it, I see a clean surface to write on, discover people who align with or oppose my thoughts, and communities I feel like engaging with in some cases or ignoring in others. In the era of short-form content with even shorter shelf lives, platforms like Substack choose the path that slowly builds trust between a creator and a consumer.

It’s this restraint that I find extremely rare in most social networks. This phenomenon becomes clearer when you take a step back and look at other platforms.

I find most of them suffocating with metrics thrown all over the screen, including likes, shares, view counts, and boosted replies. All of these factors collectively determine what you see in your feed. The platform has already decided what the content means, so there’s nothing left for you to decide. You stop participating and start performing. The medium eventually eats itself with too much optimisation.

In today’s guest piece, Anderl makes this argument, albeit with much better examples. He uses McLuhan’s hot-and-cool media framework to explain why SocialFi collapsed, why NFT culture evaporated, and why the platforms that actually work are the ones that know where to let money in without letting it take over.

Onto the story,
Prathik

Follow Anderl’s Works

There is a sentence McLuhan wrote in 1964 that has been quoted so often it has lost most of its meaning. The medium is the message. By now, it sounds like the kind of thing you put on a tote bag. But if you stop and read it as a working diagnostic rather than a slogan, it does something useful, especially for anyone trying to understand why so many recent attempts to fuse social networks with finance have collapsed in slow motion.

McLuhan’s actual argument was narrower and stranger than the slogan suggests. He thought that every medium reshapes the people who use it, not through what it says, but through the shape of the signal it sends. A medium that delivers a finished, high-resolution signal turns its users into receivers. A medium that delivers a partial, low-resolution signal forces its users to fill in the gaps, and in doing so, they become participants. He called the first kind hot and the second, cool.

Print is hot because the page is fully written. Radio is hot because the broadcast is fully produced. A lecture is hot because the speaker controls the signal. By contrast, a telephone call is cool because voice alone carries too little information, and the listener has to construct the missing context. A cartoon is cool because the brain finishes the drawing. Television, in McLuhan’s analysis, was cool because the early signal was so low-resolution that watching required constant active reconstruction. He thought, controversially, that this is why television felt addictive in a way that cinema did not.

The point that matters here is not the specific examples, which date badly, but the underlying claim. The temperature of a medium predicts the kind of behaviour it produces. Hot media produce consumption. Cool media produce participation. And, crucially for what comes next, you cannot turn one into the other without changing what the medium fundamentally is.

What’s This Got to Do with Social Networks?

In McLuhan’s terms, most of what we call social media is cool. A tweet is a fragment. A photo without context is a fragment. A ‘like’ is a fragment. None of these are finished signals. Their meaning emerges only through others’ participation, replies, reposts, threading, and association. A post without engagement is barely a thing. A post with two thousand replies becomes something else entirely, even though the original content has not changed by a single character. This is exactly the McLuhan signature of a cool medium: the artefact is incomplete on arrival and gets completed through use.

This is also why social networks feel the way they do. They are not content delivery systems. They are participation engines that happen to look like content from the outside. The platforms that have understood this, even without ever reading McLuhan, have flourished. Most platforms that tried to professionalise their participation and deliver finished signals to passive consumers have drifted into irrelevance.

The interesting thing is what happens when someone tries to add an economic layer to a cool medium. This is where SocialFi enters the story.

What Did SocialFi Try to Do?

The promise of SocialFi was elegant on paper. The argument was that social capital is real economic value. People generate it constantly, but the platforms capture it all. If we could put social acts directly on a market, then the people who actually create the value would also capture it. Each follow becomes a stake. Each post becomes a tradeable asset. Each connection has a price. The result, in theory, would be a social network that is also an economy, where reputation has a market and creators get paid in real time for the attention they generate.

For a few weeks in late 2023, with Friend.tech, this looked like it might be a real category. People bought and sold each other’s keys. Influencers commanded thousands of dollars in opening prices. The interface looked like a social network and behaved like a brokerage account. Other projects followed quickly, each promising a slightly different version of the same logic. Stamps, gated chats, social tokens, attention markets, and on-chain creator economies. The pitch decks were everywhere.

Then the whole category collapsed. Friend.tech faded. The follow-on projects mostly never reached escape velocity. Token prices crashed and never recovered again. By 2024, SocialFi had become a slightly embarrassing phrase that founders avoided in their next deck.

The standard explanation is that it was a speculative cycle, that people were there for the gains and left when the gains stopped. This is true but shallow. Speculation cycles do not explain why participation collapsed in the underlying social product. People did not just stop trading the keys. They stopped posting, reading, and showing up. The social activity died, along with the financial activity. Why?

The McLuhan Reading

The deeper diagnosis is that SocialFi did not fail because of speculation. Speculation was the symptom, not the cause. The problem was that the entire category was built on a misreading of its own medium.

Social networks are cool media. Their value lies in the fact that participation completes the signal, and that meaning accumulates through repeated low-resolution acts whose significance becomes visible only over time. SocialFi took that medium and replaced its constitutive signal, the social act, with a high-resolution one: a price.

The moment you put a real-time, visible, tradeable price on a ‘follow’ or a post, you have not added an economic layer to a social medium. You have replaced the medium. The new artefact is a fully resolved signal. There is no gap to fill in. The ‘follow’ no longer means anything ambiguous. It means this exact dollar amount right now. And once a signal is that resolved, the rational response is no longer participation. The rational response becomes allocation.

This is why Friend.tech, in its own internal logic, was not a social network. It was a Bloomberg terminal for micro-reputation, with a social network-shaped interface on top. Users were not posting. They were trading. The fact that the trades happened to be on each other’s identities did not make the activity social. It made the activity a financial one that was using social vocabulary as the skin. And once the financial dynamics turned (when prices stopped rising, when the obvious arbitrages closed, and when speculation became less profitable), there was no underlying social medium left to fall back on. The social layer had been consumed by the financial layer at the moment of creation.

This is what McLuhan would have predicted. Hot signals do not coexist with cool media. They replace them. You cannot have a partial, ambiguous, participation-driven act when one of the simultaneous attributes of that act is its current market price, updated in real time, visible to everyone. The price wins. It always wins, because it is more resolved than anything else on the screen.

The early SocialFi designers thought they were building a social network with an economy underneath. But what they were, in fact, building was a market with a social aesthetic on top. The category failed not because of excessive speculation, but because it had quietly become a hot medium and was still being marketed as a cool one.

Follow Anderl’s Works

Why This Matters beyond Crypto

It would be tempting to read this as a niche post-mortem on a niche product category. But the same logic applies far more broadly, and it explains a pattern in platform history that goes back decades.

Cool media die when they get too hot. This is not a metaphor; it is a recurring failure mode. Platforms that begin as low-resolution participation engines tend to add features that raise their resolution over time. Think of verified accounts, engagement metrics that are visible to everyone, creator funds that pay per view, and algorithmic ranking that shows you exactly how well your post performed. Each of these additions seems harmless or even helpful. But collectively, they describe a slow thermal drift from cool to hot. The medium gets clearer, the signals become more finished, and at some point, the user shifts from participating to performing, and from performing to consuming the performance metrics, and then increasingly stops showing up at all because there is nothing left to fill in.

This is why platforms that look unstoppable at peak engagement often turn out hollow a few years later. They have removed themselves from the regime where they actually generated value. Twitter circa 2012 was cool. Twitter circa 2024 is mostly hot. The drift is not anyone’s fault in particular. It is the natural pull of every metric, every monetisation layer, and every product team trying to make the signal sharper. Hotness is what optimisation looks like, applied to a medium that does not benefit from being optimised.

SocialFi made this same drift in fast-forward, compressed into months instead of decades. By starting with the hottest possible signal, a real-time market price, it skipped the entire cool phase that gives a medium its initial gravity. There was nothing to drift away from. It was hot from the first day, and it died quickly because hot media without distribution moats die quickly.

The Way Out: Condensation Points

If you accept this diagnosis, an obvious question follows. Does this mean that any attempt to fuse social participation with capital is doomed?

No, because there is a third option that early SocialFi missed entirely. You can leave the medium cool and let capital condense at specific points within it, rather than dissolving capital into the medium itself.

The metaphor is borrowed from physics. In a fluid that stays mostly gaseous, there are specific local conditions under which droplets form. The droplets are not the gas. The gas is not the droplets. Both coexist, and the interesting thing is the geometry of where condensation happens. The bulk of the medium remains in its original state, while a small number of points become dense, liquid, and load-bearing.

Cool media can work the same way. The base substrate stays cool. Most acts in the medium remain low-resolution, ambiguous, and participation-driven. But at chosen, specific moments, capital is allowed to condense out of the social substrate and become real, financially anchored, weight-bearing. Crucially, these condensation points are not the medium. They are local intensifications inside it. The rest of the medium is left alone.

I think this is the correct way to read certain platforms that quietly worked while SocialFi did not. Substack is a cool medium for writing. The writing itself is fragmentary, ongoing, accumulative, and completed by readers who reply, forward and quote the posts. Capital condenses at one specific point: the recurring subscription. That subscription is a hot signal, an explicit recurring price, but it is structured as a long commitment rather than a spot trade, which means it does not contaminate the rest of the medium with continuous price discovery. You do not see real-time tradable share prices for individual essays. The medium stays cool, and capital condenses at the subscription edge.

Bandcamp does the same thing for music. Wikipedia does it through donations rather than per-edit pricing. Patreon does it for creators. Each of these platforms has intuitively positioned the condensation points where capital can enter without heating the entire medium. None of them has tried to put a market price on every social act. They have all understood that the substrate must stay cool for the platform to keep generating gravity.

The lesson SocialFi missed is that capital and cool media are compatible, but only on specific terms. The capital has to be local, infrequent, illiquid in the right ways, and structurally separate from the bulk of social acts. It has to condense, not saturate. The moment you try to make every act capitalisable, you have replaced the social medium with an economy. And economies do not produce the kind of accumulating, ambiguous, participation-driven meaning that cool media generate.

What Comes Next

There is a generation of projects that quietly figured this out, often without naming it in these terms, and the pattern is beginning to look stable. They tend to share a few characteristics. The base layer is a social or cultural artefact whose meaning accumulates through participation.

If there is a single sentence that captures the lesson of SocialFi’s collapse, it might be this. Liquidity is heat. Adding it to a cool medium does not make the medium more efficient. It changes the medium into something that can no longer do the thing the original was good for.

The interesting design space, then, is not how to put a price on every social act. It is the much harder, much more specific question of where exactly to let capital condense inside a cool substrate without disturbing it. That question has barely been explored. SocialFi failed to ask it, because it was too busy trying to dissolve everything into a market. The next wave, the one that actually works, will probably be the one that takes McLuhan seriously enough to leave most of the medium alone.

NFTs as the Sharper Case

If SocialFi shows what happens when you build a hot medium and call it social, NFTs show something more revealing. They show what happens when you take a deeply cool medium that has worked for centuries and watch it heat up in real time.

Collecting is one of the oldest forms of cool media. Browsing record bins, stopping in antique shops, trading Pokémon cards on a school playground, and showing a stamp collection at a club meeting are all activities where the object itself carries only half the meaning. The other half is built through participation, recognition, the slow accumulation of a collection over the years, stories attached to specific pieces, and the conversations in which someone notices what you have. The value of a collected object is low-resolution, ambiguous, and context-dependent. That is not a deficiency. It is the mechanism that makes collecting a cultural practice rather than a transactional one.

The early NFT waves, in 2020 and early 2021, still carried some of this logic. CryptoPunks began as an insider thing, almost a joke among crypto natives, with unclear meaning and a value that came from shared culture rather than from a price feed. Early Art Blocks drops had a similar texture. There were forums, Discord channels, places where people talked about individual pieces, traded stories, and built community. The collecting was cool. The artefact was unfinished. It needed participation to mean something.

Then the marketplaces matured, and the thermal drift began, in a form so extreme it deserves to be studied as a case in itself. OpenSea made floor prices visible everywhere. Rarity tools quantified every trait into a numerical score. Real-time charts made every collection look like a stock ticker. Sniping bots made human reaction times irrelevant. Wash-trading volume became a status signal. Each of these features, taken individually, was a rational market optimisation. Collectively, they tipped the medium from cool to hot at a speed no cultural practice had ever experienced before.

The result was almost painfully consistent with what McLuhan would have predicted. The collector became a trader. The trader became a bot operator. The bot operator reduced the artefact’s meaning to a single floor-price number, and once that number fell, nothing was left. The communities that had formed around early collections did not mature into a richer cultural form. They evaporated the moment the market stopped moving. This is not how collectors behave. Collectors stay when prices fall. They keep talking, trading, and tending to their collection. What happened to NFT communities after the crash was not a collector exodus. It was proof that there were no collectors left. Only market participants disguised as collectors stayed, and when the market closed, the veil came off.

This is an even sharper illustration of the medium thesis than SocialFi. SocialFi was a new medium that was hot from its birth. So its failure could plausibly be blamed on novelty or speculation cycles. NFTs took a medium that had worked for centuries, survived wars, technological revolutions and changes in taste, and destroyed its working dynamics in barely thirty months. The medium had worked. The platforms broke it; not by neglect, but by relentless optimisation. Each step made the experience more precise, more measurable, and more efficient. It also made it slightly less collectable, until at some point, there was nothing left to collect.

The cautionary lesson is that thermal drift is not slow. It can happen on a timescale measured in product cycles, especially when the platform layer is built by people who do not recognise that they are operating inside a cool medium. The temptation to add another metric, another leaderboard, and another real-time price feed is constant. Each of these additions raises the temperature by a small amount that feels harmless in isolation. But the cumulative effect is the disappearance of the practice that the platform was supposed to host.

That’s it for today,
Anderl

Follow Anderl’s Works

Disclaimer: This piece was originally published on Anderl’s Substack.

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.

A lot of crypto conversations still orbit around the speed, cost and finality of infrastructure. They are all important, I agree. But they all rank far lower than something more basic and fundamental when it comes to adoption. Something qualitative and human. I wonder how builders expect a customer to adopt a new technology when they can’t help them understand how the it’s going to improve their life.

Humans evolved as a dominant social animal species because of our superior cooperation and communication skills. Yet, today, we have sidelined the very communication aspect of telling why a new technology matters in the pursuit of seeking better adoption.

Although the crypto landscape has witnessed some of the most useful technologies and products being built, few have focused on building the all-important relationship layer. I am thinking of trust, a user-friendly interface, the regulatory comfort that makes a treasurer or a retail saver actually move their money.

It is true that good infrastructure enables better movement of money, but a robust relationship layer is what can direct and trigger that move.

In today’s guest op-ed, Sebastien Davies, Partner at Primal Capital, writes about how distribution is the mechanism that turns infrastructure into outcomes.

Onto the story,
Prathik

P.S.: This was first published on Interop Markets.

The defining delusion of the last ten years in crypto was the “Field of Dreams” fallacy. Builders and investors operated with the assumption that properly engineered rails would automatically magnetise the world’s capital. This was the Infrastructure Phase, a massively capital-intensive epoch dedicated entirely to building empty containers. Between 2020 and 2024, the fundamental barriers to entry were strict regulatory prohibition and structural impossibility. Fiduciaries lacked the legal mandate to participate, and the baseline plumbing of custody, execution, and reporting simply did not exist at an enterprise-grade level. The industry deployed billions of dollars and dedicated millions of hours to engineering and lobbying solely to overcome these compliance and mechanical realities.

By 2025, the industry reached a quiet consensus. The “Battle for the Rails” had effectively ended in a draw between standardised, institutional-grade multi-party computation (MPC) custody stack and regulated stablecoin frameworks. The mechanical question of how to safely move a billion dollars on-chain has been answered. This technical victory, however, produced a troubling realisation. Infrastructure is fundamentally a commodity. It is a necessary precursor to activity, but it remains entirely agnostic regarding value capture.

In the traditional financial hierarchy, power is rarely held by the entity providing the ledger. It is held by the entity that controls the entry point. We are now witnessing the pivot from the Infrastructure Phase to the Balance Sheet Phase, but the catalyst for this transition is not more technology. It is distribution.

The Friction of the Blank Slate

Infrastructure enables a better system, but it does not account for the “gravity” of legacy balance sheets. A corporate treasurer does not evaluate a new financial rail based on its cryptographic elegance or its sub-second finality. They evaluate it based on the friction required to reach it.

Historically, financial power has accrued to those who minimise this activation energy. In the 1970s, the emergence of Money Market Funds (MMFs) offered higher yields than traditional bank deposits. However, the technology (the ability to pool short-term debt) was not the breakthrough. The breakthrough was distribution through retail brokerages like Merrill Lynch, which placed the “new container” directly into the investor’s existing workflow. The capital moved because the distribution channels made the migration invisible.

In digital finance, we see a similar pattern. The existence of a high-yield, on-chain Real-World Assets (RWA) vault is irrelevant if it sits outside the institutional user’s primary interface. Infrastructure provides the destination, but without distribution, the “audience” never leaves the traditional bank’s lobby.

Distribution Directs Liquidity

If the balance sheet is the strategic asset that determines the winner of the interest rate spread, distribution is the navigator that directs the flow toward that asset. We must view distribution as the “connective tissue” of market structure. It encompasses the presence, trust, regulatory licenses, and API integrations that allow capital to reside in a specific container.

The current market structure is bifurcated. On one side, we have highly efficient, crypto-native infrastructure capable of managing programmable liquidity with unprecedented precision. On the other, we have $100 trillion in institutional balance sheets sitting in “static” containers (traditional bank accounts and custody systems) where returns are suppressed by layers of intermediation.

Value capture in the coming decade will bypass protocols competing purely on marginal fees or throughput. The economic rents will accrue directly to the “Web2.5“ hybrids that monopolise both the customer interface and the regulatory perimeter. These new gatekeepers serve as the critical bridge, offering the “one-click” migration of a balance sheet from a zero-yield checking account into a 24/7 programmable liquidity pool.

The Shift in Gravitational Centres

We are moving into an era in which the distinction between “crypto” and “finance” is dissolving, with a battle for the front-end relationship. As infrastructure becomes commoditised and invisible, the “Interface War” begins.

When a treasurer can park cash in a tokenised Treasury fund as easily as they can in a commercial bank account, one of the bank’s primary moats (the difficulty of leaving) is challenged. However, this shift occurs only when the digital container is distributed through the treasurer’s existing terminal, auditor, and risk framework.

Infrastructure enables, balance sheets fund, and distribution allocates.

In the “Infrastructure Phase,” we treated liquidity as a stagnant pool that merely needed a better pipe. We assumed that superior efficiency alone would naturally drain it. This fundamentally misunderstands financial physics. Institutional capital operates as a dense bundle of permissions, risk mandates, and operational habits. Distribution is the force that actively reconfigures those bundles.

Distribution dictates the path of capital. In any competitive market, economic power accrues to the entity that owns the customer’s attention, systematically commoditising the underlying providers, who still earn a rent.

The Anatomy of Institutional Distribution

For institutional allocators, distribution represents a formidable, three-dimensional moat:

1. Regulatory Reciprocity: The access provider must hold licenses in the allocator’s exact jurisdiction and strictly comply with the allocator’s specific fiduciary requirements.

2. Workflow Integration: The asset has to sit inside existing enterprise command centres, integrated directly into systems like Bloomberg Terminals, BlackRock’s Aladdin, or State Street custodial dashboards.

3. Technical Proximity: The infrastructure must execute within existing legacy settlement cycles and integrate directly into the accounting software used for reporting.

Without these pillars, even the most attractive digital assets remain peripheral. Distribution turns them into balance-sheet allocations.

The “Default” Advantage

The history of financial infrastructure shows that technical elegance often loses to entrenched distribution. The persistence of the SWIFT network and the ACH system illustrates this dynamic. While both are modernising to improve speed and reduce friction, modern blockchain architectures outpace them in programmability, transparency, and marginal cost. Yet these legacy systems remain deeply embedded in enterprise connectivity. Corporate treasuries, payroll providers, and commercial banks are already wired into them.

The true friction of switching extends far beyond transaction fees. It sits in the operational risk of overhauling an entire corporate stack. As a result, value accrues to the gatekeepers who provide seamless compatibility with existing systems. The defining infrastructure of this cycle will be built by those who can bridge legacy capital into modern digital frameworks.

Distribution as a Filter for Signal

In a world of programmable liquidity, the sheer volume of “containers” (protocols, vaults, and yield strategies) creates a massive discovery problem. An institutional treasurer cannot, and will not, perform due diligence on 50 different DeFi protocols. They require a distributor to act as a curator.

This is where the hybrid model begins to take hold. These entities act as the institutional front end, performing risk assessment and compliance, and distributing structured yield to their clients. The underlying infrastructure provides the yield, and the balance sheet provides the scale, but the trusted orchestrator captures the highest-margin economics by controlling the point of access.

Redirected Flow: We are beginning to see the first signs of this “Distribution Arbitrage.” Neobanks and crypto-first exchanges are no longer just offering Bitcoin trading. They are beginning to introduce tokenised money market funds and stablecoin-based yield products directly alongside traditional savings accounts.

When a user’s primary banking interface offers 5% on a tokenised product and 0.5% on a traditional deposit, and the switch requires no change in workflow, the distribution channel has effectively broken the bank’s hold on that capital. The infrastructure was ready years ago. The balance sheet was always there. It is the distribution and integration into the financial service interface that finally trigger the migration.

Victorious Networks

Financial history is instructive here. The firms that scaled were the ones that made their products accessible to existing pools of capital and integrated them into established systems. Distribution, not just design, is what translated innovation into adoption.

Distribution via Jurisdiction

The rise of the Eurodollar market in the 1960s provides the definitive historical precedent for balance sheet migration driven entirely by distribution. The underlying infrastructure (the US Dollar and commercial bank ledgers) remained static. The catalyst was a pure geographic shift in the access point.

By launching dollar-denominated accounts in London, financial institutions systematically bypassed domestic interest rate caps and reserve requirements, engineering a profound innovation in regulatory distribution. This manoeuvre provided global capital with a highly-optimised container maximising both utility and yield. Establishing this offshore channel triggered a balance-sheet migration so complete that it permanently forged the parallel global financial system operating today. The currency rail held firm, while the geographic pivot in distribution simply unlocked a vastly superior economic container.

The Interface Revolution

In the late 1970s, as inflation surged, traditional bank deposits became a liability for savers because interest rates were capped. The “Infrastructure” for a solution (short-term government debt) already existed. However, the average investor could not easily buy T-bills in small increments.

The rise of the money market fund is a useful example. Firms like Fidelity and Merrill Lynch embedded these funds into brokerage accounts with check-writing and sweep features, allowing them to function like cash balances. This made higher-yielding instruments accessible within familiar workflows and gradually shifted a meaningful share of cash away from traditional deposits.

Integration Over Innovation

The rise of card networks is a useful example. In the 1950s and 60s, many banks launched proprietary travel and entertainment cards. Most failed to scale, but it had nothing to do with inferior plastic or slower processing times. They never established distribution across both sides of the transaction.

The networks that succeeded solved this problem in different ways. Visa, originally BankAmericard, expanded through a licensing model that allowed thousands of banks to issue cards, leveraging existing customer relationships. Mastercard followed a similar path through a bank consortium, building broad acceptance and reach. American Express took a different approach, issuing cards directly and owning both the customer and merchant relationships within a closed-loop system.

The structure varied, but the outcome was consistent. Each model achieved scale by embedding itself into both sides of the transaction, becoming the default interface for commerce by controlling access.

The Lesson for Digital Assets

These precedents reveal a recurring law: Infrastructure is the invitation, but distribution is the mandate.

● The Eurodollar proved that capital will migrate to wherever the distribution is most efficient and least constrained.

● The MMF proved that the interface (cheque-writing on a brokerage account) determines which container the balance sheet chooses.

● The Credit Card proved that leveraging existing trust networks is faster than building a new network from scratch.

In 2026, the crypto-native system is still in its early stages of distribution. The infrastructure for yield and continuous settlement is already in place, but it has not yet been integrated into the interfaces institutions rely on. Adoption will come from embedding these capabilities into existing workflows, from consumer financial apps to the treasurer’s daily operating environment.

Distribution in Digital Assets

The 2026 landscape makes one thing clear. Decentralised infrastructure is in place, but distribution remains concentrated. With a total market cap of roughly $2.5 trillion, economic activity is already clustering around a small number of access points.

The relevant question is where default relationships are forming. These are the platforms that sit at the interface with the user, and increasingly, at the centre of capital allocation.

ETF Familiarity

The approval and maturation of spot ETFs have provided a clear example of how distribution drives adoption. By placing digital assets within an established regulatory wrapper, firms like BlackRock, Fidelity Investments, and Morgan Stanley have made them immediately accessible through existing wealth management channels.

This has translated into scale. As of Q1 2026, U.S. spot Bitcoin ETFs hold over $128 billion in assets, reflecting not just demand for the asset, but the reach of the distribution layer through which it is delivered.

BlackRock’s IBIT has emerged as a benchmark because it sits inside these systems and is backed by one of the most trusted brands in global finance. It appears alongside other ETFs, within familiar workflows and allocation models. Adoption followed integration, not novelty. In effect, the asset class has been mapped directly onto the balance sheet workflows of the world’s largest pools of capital.

Neobank Proximity

A different form of distribution is emerging through neobanks and fintech platforms. Applications like Revolut, Chime, Nubank, and Robinhood increasingly serve as the primary financial interface for a large segment of users. Within these environments, new products do not need to be discovered. They simply appear alongside existing balances.

As crypto-native yield, whether through staking or tokenised assets, is placed next to traditional accounts, the decision to allocate becomes a matter of interface design rather than infrastructure. The distinction between a USD balance and a USDC balance begins to collapse into a simple choice within the same system, subject to what these platforms are permitted to offer.

In 2026, roughly 30% of American adults hold some form of crypto exposure. For many, their primary relationship is no longer with a physical bank, but with an application. Distribution, in this context, is not about access to a new system. It is about integration into the one users already rely on.

Custody Opportunity

Firms like Coinbase, Fidelity Digital Assets, Anchorage Digital, and BitGo are becoming core distribution nodes for institutional capital.

Their role extends beyond safekeeping. They provide the compliance, reporting, and operational integration required for institutions to participate at scale. In this context, custody becomes the interface through which capital is allocated, not just where assets are held.

As these platforms expand into trading, financing, and settlement, they increasingly resemble full-service financial environments. The underlying infrastructure remains important, but allocation decisions are made through these controlled interfaces. Distribution, in this layer, is defined by permission, integration, and trust.

Stablecoin Settlement

Stablecoins are moving beyond their role within crypto markets and becoming part of broader financial and operational workflows. Their relevance now extends into payments, settlement, and treasury functions, where they are increasingly integrated into existing systems.

Firms like Circle have focused on building this distribution through partnerships with banks, payment providers, and financial platforms. The objective is not just issuance, but placement within the systems where transactions already occur.

A similar pattern is beginning to emerge in software and AI-driven environments. As developers build applications and autonomous agents that operate programmatically, stablecoins are being integrated as a native payment mechanism. In this context, adoption will follow the stablecoins that are most accessible within developer tools, platforms, and APIs, extending distribution beyond financial interfaces into software itself.

Emerging Hierarchy

The structure has shifted. Early efforts focused on building the rails. Today, the advantage lies with those who control the interface.

Whether it is an ETF ticker, a neobank application, or a prime brokerage platform, the entities capturing the economics are the ones that present new capabilities within familiar systems. They translate complex infrastructure into formats that align with existing habits, and in doing so, become the default point of access.

This is what ultimately determines where balance sheets move. Capital does not migrate because the infrastructure improves. It migrates when the interface makes allocation seamless within the systems it already uses.

Stablecoins as the Clearest Example

Stablecoins are the purest manifestation of the distribution thesis. They are not merely “digital dollars”; they are the primary distribution vehicle for the US financial system in the 21st century. While the industry spent years debating the merits of different Layer 1 protocols, stablecoins quietly became the “killer app” by solving a distribution problem rather than a technical one. They made the world’s most trusted balance sheet (the US Treasury) accessible to anyone with an internet connection, 24/7.

Stablecoins are introducing a new form of competition in the deposit market. They package short-duration government-backed assets into a programmable, portable, and continuously accessible format. As distribution improves, capital can move more easily between traditional accounts and these digital alternatives.

The Utility Gap

The persistence of Tether (USDT) is the ultimate proof that distribution beats “better” infrastructure. For years, critics pointed to USDC’s superior transparency, regulatory standing, and institutional backing. In a vacuum, USDC is the “better” product.

Tether established early distribution in offshore trading markets and across many emerging economies. It became a primary liquidity instrument across exchanges and a widely used medium for cross-border transactions.

By the time regulated alternatives entered the market, Tether was already deeply embedded in trading, settlement, and informal financial flows. Its presence across platforms and use cases created a level of integration that is difficult to displace.

Programmable Liquidity

The next phase of stablecoin distribution is the shift from passive holdings to active use. In 2026, these assets are increasingly being integrated into operational workflows.

This is showing up across a few key areas:

● In supply chain finance, large corporates are beginning to use stablecoins to accelerate settlement with suppliers, reducing reliance on traditional correspondent banking timelines.

● In commerce platforms, companies like Shopify and Stripe are enabling merchants to hold and manage balances in stablecoin form, introducing new options for liquidity management.

● In software environments, as applications and autonomous systems transact programmatically, stablecoins are being integrated as a payment layer for real-time, granular transactions.

In each case, adoption is driven by utility within existing systems rather than by exposure to new asset class.

Extending Distribution

Banks are responding to stablecoins by extending their own distribution. Deposit tokens and bank-issued digital assets will allow them to bring existing balance sheets onto new rails while maintaining control of access.

The objective is to preserve their role at the point where dollar-denominated value is accessed and deployed. As these instruments are integrated and eventually embedded into payroll systems, trade platforms, and software environments, the competitive dynamic shifts toward distribution.

As the market continues to scale, value accrues to the entities that control the interface through which transactions are initiated and managed.

Institutional Constraints

If the technical rails are ready and the economic incentives of the balance sheet are clear, why has the “Great Migration” not yet reached its terminal velocity? The answer lies in the final, and most formidable layer of the distribution moat: institutional friction. For a Chief Investment Officer or a Corporate Treasurer, capital allocation is governed by fiduciary responsibility, regulatory constraints, and operational requirements.

Distribution, in an institutional context, is the process of neutralising these three specific types of friction.

The Regulatory Distribution Gate

In the retail world, distribution is about “eyeballs.” In the institutional world, distribution is about licenses. An institution cannot legally park a balance sheet on a protocol, no matter how efficient, if that protocol sits outside its regulatory perimeter.

In 2026, distribution is increasingly concentrated around assets packaged within familiar legal and regulatory structures. Tokenised Treasury funds such as BlackRock’s BUIDL and Franklin Templeton’s FOBXX illustrate this pattern.

These instruments allow institutions to hold digital assets within existing investment frameworks, including structures that qualify as regulated investment companies. The format is new, but the pathway to allocation remains unchanged. These assets are accessed through standard brokerage systems and integrated into existing balance sheet management.

Workflow Lock-in

Large-scale capital does not move via a browser extension or a mobile app. It moves through established “Workflow Hubs” like the Bloomberg Terminal, Aladdin, or Charles River. These platforms are the ultimate distribution monopolies because they own the daily habits of the financial professional.

Early digital asset infrastructure was built as a parallel system. It required users to step outside their existing workflows to access it. For institutions, this created friction at the point of allocation.

Distribution is now shifting toward integration. Digital asset custody and exposure are being embedded into existing reporting, risk management, and portfolio systems. The objective is to make on-chain positions appear alongside traditional assets and behave consistently within the same operational framework.

Distribution Runs on Trust

In finance, trust is embedded in the distribution layer. Capital is allocated through counterparties that meet established standards for regulation, creditworthiness, and legal recourse.

The failures of major crypto intermediaries in 2022 and 2023, including FTX and Celsius, reinforced the importance of this framework. Established institutions such as BNY Mellon and State Street entering digital asset custody extend that layer of trust into new formats. Their role is not just to provide infrastructure, but to make digital assets compatible with existing expectations for balance sheet allocation.

Compliance at the Interface

The final piece of institutional distribution is integrating compliance into the interface itself. To access large pools of capital, digital asset systems must meet real-time requirements for transaction monitoring and anti-money laundering.

The firms gaining traction are embedding these controls directly into the point of access. By the time capital reaches the underlying system, the relevant checks have already been performed. Compliance, in this structure, becomes part of the distribution layer rather than a separate process.

This allows institutions to interact with programmable liquidity within established regulatory expectations, making allocation operationally viable at scale.

Where Distribution is Moving

As we look toward the 2027–2030 horizon, the “Interface War” is entering a phase of total invisibility. The goal of distribution is no longer to convince a user to “use crypto,” but to embed digital asset utility so deeply into existing platforms that the end-user (whether a retail saver or a corporate treasurer) doesn’t even realise they have migrated their balance sheet.

Invisible Distribution

The next great capital reallocation will be driven by Embedded Finance. In this model, the distribution node isn’t a bank or an exchange; it’s the software in which the economic activity already occurs.

● ERP Integration: SAP and Oracle sit at the centre of institutional financial workflows. As new forms of liquidity integrate into these systems, allocation decisions can be executed within existing treasury processes. When excess cash can be directed into tokenised instruments from within the same system used for cash management, the balance sheet moves as part of the workflow rather than through a separate interface.

● Supply Chain as a Ledger: Distribution is extending into supply chain and trade systems. Platforms that manage global logistics are beginning to integrate stablecoin-based settlement into their core software, connecting payment and transaction data more directly. As settlement becomes embedded within these workflows, financial activity moves closer to the underlying commercial process, creating new pathways for liquidity and credit to be applied in context.

Banking Platforms

Traditional banks have long served as the primary distribution layer for capital. Leading incumbents are now extending this role into programmable liquidity, leveraging their decisive structural advantages: decades of fiduciary relationships, comprehensive regulatory licenses across jurisdictions, deep integration into enterprise systems (Bloomberg Terminals, Aladdin, custodial dashboards), and the bedrock of deposit insurance together with general balance-sheet trust. These moats make it exceptionally difficult for neobanks or crypto-native platforms to achieve meaningful distribution at an institutional scale.

This advantage is decisive. While fintechs and crypto platforms can capture retail flows and smaller corporate wallets through superior user experience, meaningful “distribution arbitrage” at the multi-trillion-dollar institutional level remains limited. Banks do not need to out-innovate the underlying rails. They simply need to embed on-chain yield, settlement, and programmability into the interfaces and trust networks they already own. In this sense, the Interface War is one that banks are structurally wired to win.

● White-Label Liquidity: A hybrid model is beginning to take shape. Banks provide the trust, regulatory framework, and client interface, while yield is sometimes sourced from external, on-chain systems. In this structure, the bank’s role evolves from holding static pools of capital to managing flows across a broader set of financial environments.

● Deposit Tokens: The bank account of 2028 is likely to take a different form: a licensed deposit token that remains compatible with existing banking systems while extending into digital asset networks.

Agentic Distribution

A further shift is emerging as financial activity becomes more automated. As software systems and autonomous agents begin to manage capital, distribution extends beyond human interfaces and into code.

In this environment, allocation decisions are shaped by parameters such as integration, programmability, and capital efficiency. Systems that are easier to incorporate into software workflows become more likely to be used.

Distribution, in this context, is determined by presence within the tools and environments where these decisions are executed. Assets and protocols that integrate cleanly into developer ecosystems and meet operational and compliance requirements are more likely to be selected as default components within these systems.

The Point of Allocation

The arc of the last decade is clear. The underlying systems have been built. Capital can now move across them. What matters now is how that movement is directed.

The infrastructure is in place: secure custody, scalable networks, and programmable assets. These capabilities made participation possible. The next phase is defined by how capital is allocated across them.

As we have seen, capital does not move on technical capability alone. It moves when the pathway is integrated into existing systems and aligned with economic incentives. Distribution is what makes that movement operational.

Where Value Settles

As we move toward 2030, the distinction between digital assets and traditional finance becomes less relevant. The underlying systems are converging, combining the efficiency of new infrastructure with the stability of existing balance sheets.

In this environment, financial power sits with those who control the interface through which liquidity and collateral are accessed and allocated. The role of financial institutions evolves accordingly, from holding assets to managing flows across increasingly integrated systems.

The infrastructure is in place. The balance sheet is becoming more mobile. What determines outcomes now is distribution.

That’s it for today.

See you with the next one,
Sebastien Davies

Disclaimer: The piece was originally published in Edition 3 of Interop Markets

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.

Last week, my sister asked me the one thing I absolutely dread in my line of work. ‘Can you teach me how to invest in crypto and become rich?’ On most occasions, I politely evade the topic. But since it was my sister, I conceded.

For the next couple of minutes, I explained to her about on-ramps, off-ramps, centralised and decentralised exchanges, switching, bridging, chains, and a few other concepts. I told her how these form the base for moving money between on- and off-chain financial worlds.

She thanked me, perhaps sarcastically, and told me she was content with moderate yields on equities and commodities. I got out of the predicament, but a thought lingered. This wasn’t the first time I’d seen someone feeling overwhelmed by trying to wrap their mind around how crypto works. Not that it is as difficult to understand, but the thought of having to trust a smart contract without any recourse to a centralised body, like a bank, to go back to, scares many.

I think this ‘trust problem’ is the single biggest reason to keep the masses away from adopting crypto. But I also feel crypto, for what it offers, can change the way the world moves money for the better. This is why the ‘trust problem’ needs to be addressed.

In today’s guest piece, SysLS writes about how the timing of AI-based primitives couldn’t have been better to help crypto change the world of finance for good.

You can follow his works on X and Substack here 👇🏾

X

Substack

Onto the story now,
Prathik

P.S.: This piece was first published on Systematic Long Short.

Betting on crypto has essentially become a bet that humanity will continue to seek acceleration in its processes. That seems like an easy bet.

I think the time is now because of the birth of meaningful artificial intelligence, which finally gives us the tools to drastically reshape how we interact with crypto.

This article explains why crypto will become a strategically important asset class.

Bottlenecked by Due Diligence

Humans have sought acceleration since the dawn of time. We measure success by efficiency and have consistently reduced friction in everything we do. Where it once took weeks to wire money across distances, it now takes seconds.

Today, most human processes are no longer bottlenecked by technology, but by the need for verification. Getting a job, opening a bank account, or raising funds still takes weeks. That’s not because the technology to complete these processes quickly doesn’t exist, but because humans must bear the cost of verification.

In an employment process, the employer needs to verify that you are not a lemon. It’s not that you have never been hired before, but it’s still worth doing independent verification because your previous employers will not take responsibility if you turn out to be a poor hire. Further, there is an argument that the best employees stay employed. Lastly, different environments and job scopes also imply that past performance may not be indicative of future performance.

So, employers pay the cost of due diligence by way of several rounds of interviews and take-home assignments. This process takes time and human effort — it is messy, unstructured, and non-deterministic; and also why hiring processes are as much a function of luck (did your interviewer have a good morning before meeting you? Or did they just have a massive argument with their spouse?) as a function of skill.

It is fairly trivial to reason that all costs in such processes are verification costs.

The actual act of officiating employment is quick and relatively painless — an email and a digital signature. This means that, in theory, suppose there was a magic scoring box that could output the absolute scores of a candidate employee across every dimension you could possibly care about, and you trusted this score with a high degree of confidence. Then, the time taken to hire someone would be reduced to the algorithmic time of filtering for and sorting the scores you care about, then sending over an employment contract and having them sign it.

Forming A Trust Daisy Chain

The reason you are able to hire someone near-instantly with the addition of a magic scoring box is that you have delegated the responsibility of verification to it, and you have learned to trust it.

In practice, this happens all the time.

Potential employers within the same industry borrow the due diligence of past employers — your resume is less likely to be binned if you have been to top-tier institutions and have done difficult things before.

Whenever possible, humans pass the cost of verification to a higher entity and trust that it has acted in their interest and done the prerequisite verification. For example, a vast majority of humans simply trust that the government’s banking oversight arm has done its job, rather than conducting lengthy, detailed, and expensive due diligence on a bank before opening an account with it.

If we trust the government sufficiently and its ability to uphold the law, then we can take its word that the bank is solvent and safe for our financial affairs.

This “trust daisy chain” happens everywhere, all the time, in things - big and small. A start-up has no brand equity and therefore commands no inherent trust, but a large, reputable, and successful Venture Capital (VC) firm can lend the start-up brand equity by simply investing in it. The public, though unfamiliar with the start-up, trusts it if they trust the VC, since they are delegating verification to the VC’s due diligence process.

The Problem With Crypto

Crypto has traditionally offered no “higher entity” to which we can pass on the cost of verification. Having no higher entity to arbitrate honest mistakes, like swapping $50 million USDC for $5,000 worth of tokens or even sending it to the wrong address, implies that the cost of verification falls squarely on the user.

This is genuinely difficult because smart contracts are not designed to be parsed by humans in under a second. This is essentially why most crypto has remained relatively niche. Where there has been widespread adoption of crypto products by the retail masses, it has mostly come in the form of large, centralised institutions that abstract away the complexity of smart contracts (e.g., Binance and Coinbase).

These retail users may not understand how Aave smart contracts work, but they have learned to trust Binance; therefore, Binance can build a business out of abstracting the complexities of the smart contract from the user and wrapping it in a simple interface.

This means that, technically, light users really only have two options when it comes to crypto: wait for a centralised entity like Binance to offer an abstracted version with some fees, or interact directly with the protocols and risk catastrophe.

Despite crypto’s relative maturity, it is still extremely high-friction to work with and fundamentally scary to approve transactions, mostly because you know there is no recourse if you mess up. So you diligently verify, but it is not trivial to read and parse through every transaction you do on-chain, let alone deeply understand the smart contracts you are interacting with.

For those who understand software and its infinite composability, the allure of crypto has always been an ecosystem of modular components that can interact with each other in a trustless manner upheld by algorithmic law. The headwind on the way to that potential utopia is that bad actors and poorly designed edge cases require humans to pay the cost and friction of verifying each modular component.

Where Does AI Help Crypto?

The rise of meaningful artificial intelligence allows us, for the first time, to delegate verification to a higher entity. This drastically changes our mode of interaction with smart contracts. We will no longer need to personally verify the design and behaviour of smart contracts, or even necessarily understand the details of our transactions.

Instead, we learn to trust a single entity — our agents. This is already showing great promise, as agents demonstrate considerable expertise in understanding and reasoning about code. They can tirelessly parse and check transactions and understand the algorithmic law that each smart contract upholds within seconds.

This implies that for the first time, through our agents, we can realise the potential of crypto. Need to interact with nine smart contracts in one atomic transaction to get what you want most efficiently? Have your agent verify that the transaction achieves the outcome you want and that no foul play is involved. Done.

With each successive generation of agents, they will only get better at this.

Implications of Algorithmic Verification

If we can resolve the problem of verifying smart contracts and the transactions they generate, then we move toward the utopia of having all kinds of interactions available to us while remaining permissionless, trustless, and upheld by algorithmic law. If the cost of verifying a smart contract goes to zero, then we gain the benefits of all the verification and guarantees the smart contract brings for free.

The major consequence is that all processes that remain within this paradigm now move at the speed of compute. For a civilisation determined to move faster, remove friction from processes, and become more efficient, it is hard to imagine us straying from this paradigm.

It only makes sense that more and more goods and services will have smart contract interfaces that determine exactly what you will get and what you will need to exchange to get it. It is also a paradigm that compounds on itself — every successful product would create a composable piece that can play into another product later on. For example, suppose a decentralised exchange created a reputation system for trading proficiency. People looking to hire traders using a trustless employment protocol might filter based on the reputation contract of an unrelated decentralised exchange. A lending protocol might use that as one dimension of credit scoring.

Such a paradigm would not only be extremely efficient — all processes and transactions moving at the speed of compute — but boundless, growing more powerful and unlocking more possibilities as a function of participation and time. It is a positively compounding paradigm.

So, if you believe that humans will continue moving in the direction of least resistance and seeking out solutions that allow us to move faster, be more efficient, and do more, it seems the time for crypto is now.

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.

Most of the AI conversations I’ve come across have largely been dominated by GPUs. Nvidia’s stock price, chip export bans, and data centre buildouts - is all that anyone talks about. But there’s a lot more that’s brewing underneath that seldom get the attention it deserves. So, when I read a piece about the memory industry that powers the entire AI stack, it felt under-appreciated.

The fact is that the GPUs are only as powerful as the memory that feeds them data. Without enough bandwidth, the most advanced chip in the world is useless.

Memory companies like Micron and SK Hynix have historically been treated as commodity players whose stocks rise and fall with supply gluts. But that perception is chaging. AI workloads need memory more than traditional computing ever commanded. So, a handful of companies that know how to manufacture high-bandwidth memory at scale are finding themselves with pricing power they’ve never had before.

In today’s guest piece, Kerman Kohli and Nicolas break down the memory industry from first principles, including what the different layers of memory actually are, why HBM has become the critical bottleneck in the AI buildout, and what makes these companies so hard to compete with.

You can follow their works here:

Kerman's Substack

Nicolas's Substack

Onto the story,
Prathik

Hello

Everyday it feels that memory stocks are going up. Micron, one of the largest memory players, is up almost 3x in the past year alone. Many investors are stuck seeing similar names go up day after day, waiting for a pullback. However, they fail to really understand what the product really is.

In this piece, Nicolas and I are here to break it down in an easy to understand way to understand the opportunity at hand. Let’s get into it.

Intro

So what is memory and why is it so important?

Memory is what enables a computer or device to store information while ephemeral computation is being performed. This is done predominantly through read and write operations. Each layer of memory has different profiles around read/write speeds, cost and capacity.

In AI, memory has become even more important because models need to process massive amounts of data all at once. When we use tools like chatbots, image generators, or recommendation algorithms, memory is constantly moving huge datasets in and out at high bandwidths.

The more intelligent and capable AI models get, the more memory they need to function effectively. Without powerful memory systems, advances in Large Language Models and Machine Learning use cases stall.

The Memory Hierarchy (Storage vs Working)

To better understand memory, we first need to break down the layers of memory.

Computers split memory into working memory (used while the system is actively performing computations) and storage memory (used to save data long-term at the expense of slower read/write speeds). This separation exists as memory that is high bandwidth (low read/write times) is expensive, while long-term memory that is cheap, is lower bandwidth (high read/write times).

Many concepts around memory boil down to the distance that the chip is from the processing unit. The further the distance, the slower the throughput.

1. Processor Registers & CPU Cache (SRAM):

What it is: This is often the highest throughput memory in the entire system as it is sitting inside or right next to the XPU (XPU = CPU or GPU). It holds tiny bits of data the processor needs right now.

What it’s made of: SRAM (static memory built directly on logic silicon)

Cost & size: Extremely expensive per bit, tiny capacity

Where it’s made: On the same chip as the CPU

Key manufacturers: Intel, AMD, Apple

2. Physical Memory (DRAM / RAM)

Designated random access memory / Random Access Memory

What it is: This is the computer’s main working memory, the desk where active programs live. Throughput is required here as delays cause a queue of computations to occur.

What it is made of: DRAM cells (one transistor + one capacitor per bit)

Cost & size: Somewhat expensive, medium capacity (GBs)

Where it is made: Mostly fabricated in South Korea, Taiwan, and the U.S.

Key manufacturers: SK Hynix, Samsung Electronics, Micron

3. High Bandwidth Memory (HBM)

Specialised DRAM

What it is (simple): Ultra-fast DRAM stacked vertically and placed next to AI chips. Due to the vertical stacked nature of HBM, it has higher throughput at the expense of manufacturing complexity.

What it’s made of: DRAM dies stacked with through-silicon vias (TSVs)

Cost & size: Very expensive, smaller capacity relative to DRAM; although much faster than it

Where it’s made: Taiwan & South Korea, due to the advanced packaging requirement (physical vertical integration)

Key manufacturers: SK Hynix, Samsung, Micron

4. Solid-State Storage (NAND / SSDs)

NAND is non-volatile flash memory that stores data without power, used in high-density storage devices like SSDs, USB drives, and memory cards.

What it is (simple): Long-term storage for files, apps, and data when power is off

What it is made of: NAND flash cells storing electrical charge

Cost & size: Cheap per Gigabyte, large capacity (hundreds of GBs to TBs), lower throughput than HBM and DRAM, but often sufficient for less latency critical computational workloads.

Where it is made: Asia mainly (Korea, China, Japan)

Key manufacturers: Samsung, SK Hynix, Sandisk, Micron, Kioxia

5. Hard Disk Drives (HDDs)

What it is (simple): Traditional spinning disks for cheap bulk storage

What it is made of: Magnetic platters and mechanical parts

Cost & size: Very cheap, physically large, slow throughput

Where it’s made: Asia

Key manufacturers: Seagate, Western Digital, Toshiba

The closer memory is to the processor, the faster, smaller, and more expensive it gets, AI pushes demand toward the very top of the pyramid. This is due to the extremely parallel nature of GPUs that are performing trillions of computations per second.

HBM and NAND

HBM is the most critical memory layer as it sits directly next to AI GPUs, while NAND is the “warehouse” storage that holds datasets, model checkpoints, and logs.

In AI data centres, NAND-based SSDs feed data into DRAM/HBM, and HBM then feeds the GPU fast enough to keep the compute busy. LLM models utilise this tiered architecture of memory to ensure the most efficient use of all storage layers. However, the preference is always to have as much computation as close to the GPU as possible.

Over the past few years of the AI buildout, both have seen unprecedented demand. HBM demand is exploding because bandwidth is the limiter, while SSD demand rises because data inputs & outputs keep growing (training data, retrieval, inference logs). An often under-appreciated fact about memory needs is the recursive nature of agentic workflows that consume compute resources as they call other agents, which call even more agents. Agentic activity can therefore lead to situations where demand scales beyond human induced demand.

Historically, DRAM and NAND have been viewed as commodities by investors and the broader market. This has meant that supply is monitored very carefully and matched to meet demand as it is needed. Overbuilding has dire consequences as semiconductor fabs are expensive to spin up and costly in their ongoing operations. Due to this nature, supply is gradually increased in the market to avoid an over build. However, as AI demand has exploded, very suddenly all forms of memory are becoming key bottlenecks and giving memory players significant pricing leverage over customers. This pricing leverage allows them to record sky high profits as they are the critical bottleneck in the AI supply chain. GPUs without memory are rendered useless. No computation can happen without memory. In order to understand why and how they can maintain pricing leverage, the next section talks about the technological moat they hold.

Memory Defensibility

What makes the memory players turn from commodity to the overlords of the AI race is their advanced specialisation in semiconductor manufacturing processes. These processes can be broken down into three key elements:

• Manufacturing complexity (how to make the chips)

• Yield (how many successful chips you make per batch)

• Qualification processes (does the chip meet requirements for consumption)

The moat is a game of making billions of tiny cells reliably, at massive scale, with razor-thin margins. For HBM specifically, you need advanced DRAM and complex 3D stacking/packaging (TSVs, thermals, interposers) that only a few players can do at high yield, and customers must validate parts over long cycles.

It’s for this reason that there are only three companies that qualify in this game. Leadership shifts matter when one vendor ships next-gen stacks earlier as it gives them a distinct advantage in perfecting the manufacturing process of the next generation. HBM is also structurally premium-priced versus standard DRAM due to being co-packaged with the processor and not being homogenous like previous generations of DRAM (which were “commodities”).

Spinning up a memory company that can compete at the scale of the current players would take over two decades of expertise and over $50 billion in funds. China’s YMTC has been able to catch up in legacy DRAM manufacturing processes but has struggled to do it with high yields and without government support. They are also locked out of more advanced generation of semiconductor manufacturing capabilities due to material and technology export restrictions imposed by the United States. Furthermore, despite memory being “hardware”, there is still a chain of software in the form of chip firmware that requires deep integration. China or any other state players have to be able to overcome deep software lock ins in addition to all the other challenges that present. It is for this reason memory companies are far more defensible than any other point in their lifecycles.

Closing

Raw compute power has scaled faster than memory’s ability to move data, this is the “memory wall”. Chips can do math insanely fast, but they stall if data can’t reach them quickly enough.

A lot of time and energy goes into shuttling model weights and activations between memory and compute, not the math itself, so bandwidth becomes the limiter.

HBM is the current best workaround because it puts wide, fast memory right beside the GPU, but it’s capacity- and supply-constrained, so memory ends up setting the pace for how quickly AI systems can scale.

What we’re experiencing right now is a set of 10 or less companies that hold the manufacturing specialisation to develop memory chips that power the future of AI which isn’t just a matter of productivity, but slowly national security as these chips enable next generation warfare to take place.

If you believe:
a) AI is here to stay and that
b) AI’s demands will only increase over time,

then the future belongs to these 40-50 year old companies that have been making “commodities” that are now the core bottleneck in the AI build out. Their immense pricing power has already allowed them to start extorting many companies downstream in their supply chain and we believe that this trend will continue to extend, impacting their customers’ margins.

We are at a point of structural change in the world and memory may well be one of the earliest signs of this new world order.

That’s all for this weekend.

Our team at the Token Dispatch will be back with interesting reads.

P.S.: This piece was first published on Kerman Kohli and Nicolas’ Substacks.

Kerman's Substack

Nicolas's Substack

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.

I find a lot of AI debate still stuck between two positions. One side thinks AI will take away every job overnight. The other dismisses the entire AI wave by declaring that the tools are still inconsistent and need supervision. The problem is that none of these stances is particularly helpful. New technologies don’t need to be perfect to trigger a seismic shift. They just need to become good enough to make ignoring them look more dangerous than engaging with them.

This is where moving in the right direction matters more than the speed of the movement. A vehicle moving at 1 mph in the right direction will still travel farther than one standing still. Starting early works the same way: it gives people room to be wrong, to learn, and to course-correct, while still ending up ahead of the laggards described in diffusion of innovation theory.

None of this means that humans must become obsolete with each introduction of new technology. What new innovations do is they move the bar for what is disposable. This conversation is timely given we see AI agents start an economy of their own - making payments, signing contracts, and even hiring humans to help them with their tasks.

So what should you do in a world where predictions about the future become outdated before people have gotten to debate over the predictions?

In today’s guest op-ed, SysLS writes about how to reason through that kind of future without waiting for certainty. It is a piece about moving before the window closes, recognising which edges are temporary, and understanding why the people who start early often earn the right to course-correct in public.

You can follow his works on X and Substack here 👇🏾

X

Substack

Onto the story now,
Prathik

Introduction

The first time I realised we were heading towards an inflexion point was when I heard the music slowing down at my previous role, even as everyone around me pretended nothing would change.

I was managing a team of nearly 20 people in a hedge fund, doing the thing I had been doing for years. For all intents and purposes, I was likely going to do even greater things there. And yet, I moved from a position people would kill for to building a startup from ground zero with a skeleton crew - a move so little understood and widely seen as crazy. With the recent news of massive layoffs, people quitting explicitly to build startups, or quietly quitting and burning tokens at night doing the same, my actions seem a lot less insane now.

I’ve had a few people ask me where I think all this goes. This article is the answer to that. The honest truth is that I’m not really sure about the magnitude of these changes, but if quant finance has taught me anything, it’s that being directionally correct is often enough.

Writing On The Wall

It was ChatGPT o1 that did it for me. Up until that point, I had referred to them only as “LLMs” and not “AI”. I was not yet convinced that any semblance of real intelligence would emerge from them.

But with o1, it was the first time these LLMs could credibly produce code from well-structured prompts. It was still messy. They still suffered from the occasional bout of hallucination and confusion. But what mattered was that they could actually produce useful code.

The line of reasoning I took was this: once AI could reproduce useful code, it would recursively write improvements to its own logic and accelerate development at a scale we could not comprehend. Whenever I shared this, people would counter-argue that the code agents wrote was still buggy and not “production-ready.” This misses the point that humans write buggy code, too.

We don’t need flawless code to stop writing code entirely. We stop writing code the instant we realise that agents produce fewer bugs than we do, at a pace that far exceeds us. The bar for fully relegating the burden of coding to agents was so low that once I saw o1 up close, I knew the future was going to change dramatically.

Quant Finance And The Moat of Knowledge

I initially believedAI would eventually eat away a vast majority of quant finance. But I expected it to take a while, since there was little publicly available institutional code for LLMs to train on. I imagined software engineering as a pyramid: at the base was basic code monkey work, above that was your senior developer with some architectural thinking, and above that were specialised developers: data scientists, quant developers, and so on. The more your profession required specialised knowledge, the safer you would be.

I thought we would wipe out the entire tranche of code monkeys within two years. Then, senior developers would start to go. And layer by layer, specialised knowledge would also be incorporated into the LLMs, and they too would be wiped out then.

It quickly became obvious that the frontier model providers would eventually hire specialised knowledge workers to contribute industry know-how to these models. Specialised knowledge seemed like it would be a moat for the next couple of years, but would also be eaten away gradually.

The Remaining Moats

There were still a few business categories I thought would be safe from trivial disruption within the next five years.

The first is proprietary data. Businesses that produced a lot of proprietary data as exhaust would be hard to disrupt. Large podshops like Millennium come to mind; they can collect analyst readings, detailed analyses, recommendations, and actual price changes, and use this data to fine-tune frontier models into something that cannot be easily replicated. Any business producing proprietary data not trivially obtained by the frontier models would have a longer lease of life.

The second is regulatory friction. Businesses where other humans are a bottleneck seemed much harder to disrupt. Being able to trade in many TradFi markets meant opening broker accounts, getting licenses, and signing contracts around the globe. It’s easy to trade crypto, but much harder for a non-Chinese firm to trade iron ore in China. If you need a human to rubber-stamp your progress, the speed of that industry will always be bottlenecked by the cost and speed of that approval.

The third is authority-as-a-service. It’s not too hard now to get an agent to draft a legal opinion, given a comprehensive study of the matter and the relevant laws. And yet we’re still going to pay tens of thousands of dollars for one drafted by a lawyer, because an AI’s legal opinion is worth nothing at this point in time. Smart contract audits are another example. We’re probably already at a level where agents can review smart contracts as well as, or better than, the top decile of humans, yet most people still buy the stamp of authority from a branded firm. The opinion isn’t what you’re paying for. The authority behind it is.

The fourth is the physical intelligence lag. Hardware moves much more slowly than software, and breaking hardware is a lot harder to fix. Physical businesses that interact with the real world are a lot less likely to be disrupted soon. That said, once hardware catches up, the same pyramid logic applies: lower-level jobs go first, then the more specialised ones.

These moats are real, but none of them are permanent. The honest read is that they buy time, not safety.

Reasoning About A Messy Future

When the future is genuinely noisy, and the rate of change is fast enough that most analogies break down, people tend to do one of two things. They either wait for certainty before acting, or they pattern-match to the past (”this is like the internet boom”) and act on the wrong model. Both are mistakes.

It is worth reasoning from first principles under incomplete information. You don’t need to know exactly how something plays out. You just need to be directionally correct and structure your bets so that being early and wrong is survivable, while being early and right is disproportionately rewarding.

Asymmetry is the whole game when the future is uncertain.

The practical version of this is: ask what has to be true for a given outcome to happen, and then ask how legible the inputs to that outcome already are. The inflexion we’re living through was not unforeseeable; the inputs were visible. Code that could write code. Models that improved recursively. Institutional knowledge that could be bought, not just grown. Anyone willing to stare at those inputs clearly could see roughly where they pointed, even without knowing the exact path.

You can recursively reason about this and extrapolate further. I don’t even think we’ve yet caught a glimpse of what it will be like when agents can train themselves, replicate, and become truly autonomous. An agent that can increase its intelligence by 0.1% through a series of actions may not seem significant, but any number that is not 0 increases the probability that the next increment is greater, and so on, so forth. There are vast power laws at play here, and it is worth imagining what the future looks like under those power laws.

By the time the signal becomes obvious, the trade is crowded. In markets, you pay for early conviction with uncertainty. In careers and startups, the currency is the same.

So the real question isn’t really “what’s going to happen?” The question is: “What do I already know, what direction does it point to, and what’s the cost of acting on it now versus waiting?”

One thing that I often see people missing is noting that action creates information. Action does not happen in a vacuum. When you act on the world, the world replies with information. That information powers iteration. Iteration begets more informed action. That is the nature of progress.

Being still with incomplete information is decay.

Moving towards action is discovery.

Thinking About Next Steps

I knew I had a few years if I wanted to milk the status quo. But a large part of me felt like if I wanted to do something, I would have to start sooner rather than later. I had always wanted to build something truly mine, and it seemed like the window to do that was quickly closing.

To be clear, I know that the largest hedge funds in the world would be fine. They have proprietary data that makes them very difficult to replace. TradFi markets are also bottlenecked by human signatures, both on a regulatory and, at times, even a trading front. What I do think, however, is that these largest funds will use AI to replace most of their workforce, even terminal career seats like Portfolio Managers. Not immediately, but eventually, surely.

What I felt was that I had about 4-5 years before the foundation model providers hired enough specialised talent to make being an upstart trading firm nearly impossible. In certain markets, like U.S. equities, it already feels that way. I can’t imagine how much more efficient it’s going to look in just a few more years.

There was clearly not going to be space for “second best” pretty soon. I could keep working for the “best”, but it seemed more aligned with my goals to strike now, in a market where I had a genuine edge and knowledge that would not be trivially replicated. So, having that dawg in me, I called it quits and went all in on what eventually became @openforage.

Subscribe Here

Inflexion Point

Today, it’s really starting to feel like the window is visibly closing. The pace of change has stopped feeling gradual, and most people in space are beginning to realise that what used to take months of improvement now takes weeks.

I don’t think jobs will vanish entirely within the next couple of years. There will always be a need for humans. Humans are social creatures. As long as humans are in charge, we want other humans around. And humans don’t trust AI yet, so stamps of authority still need to come from a human. I imagine AI CEOs in the next couple of years, but there will still likely be a human CEO having to “approve” and certify the AI CEO. This idea of human certification cascades down the pyramid. A human manager will manage and certify a bunch of agents working under him.

But the arithmetic of hires will change. If a CEO can prompt an agent more easily than they can prompt you, there’s no need to hire you. Shallow, code-monkey work will be difficult to find going forward.

To be irreplaceable, you need to operate at a timescale far above current agent limitations - receiving instruction, managing agents, and working with them for weeks, months, or years. Long-term strategic thinking and policy planning are some of the strongest job moats for the foreseeable future. You also need to operate at a scope greater than the current agent limitations. Agents have limited context. They know everything about anything, yet cannot see how component A interacts with component B, which in turn interacts with component C, causing cascading effects to component D. They lack scope.

If you can think far and wide, absorb information quickly, make decisions for the long term, and are likeable, you will hold down a job, at least for the foreseeable future.

If you do intend to be an employee, it’s worth taking stock of what your work is actually made of. Some tasks are deeply human defensible. Some will be replaced cheaply over the next couple of years. Do more of the former and less of the latter.

Working for a great firm in a deeply defensible position, one that sits behind real moats, may give you a career runway while the rest of the workforce gets eaten by the foundation models. You can still spend your tokens at night, rolling the dice, trying to build something meaningful.

But if you have a burning desire to contribute a unique verse to the world, think carefully about where your market of choice is heading. If your window to build something defensible is closing, you need to begin operating before the market fully prices in the competition that is coming.

Conclusion

The inputs that create inflexion points are legible ahead of time, if you’re willing to look. Most people don’t look, or they look and don’t act, or they wait until the signal is so obvious that the opportunity has already disappeared.

Don’t ignore the shifting sands. Don’t stay somewhere that’s losing ground while telling yourself you’ll make the leap when the timing is better. There’s no better timing, and the timing rarely announces itself. When it becomes obvious to everyone, the window has normally already closed.

I looked, I made a bet, and now I’m living inside the outcome of that bet — for better or worse.

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 world of finance has an extremism problem. I have seen maximalists convinced that blockchain will dismantle every financial institution ever built. Then there’s the TradFi camp, which thinks Bitcoin is synonymous with crypto and vice versa. The sad thing is, both camps show little patience for nuance.

I don’t buy the binary. As we can already see, both will most likely converge rather than collide. Visa and Mastercard are busy striking partnerships in the blockchain payments landscape. Stripe, the traditional financial services giant, has launched a blockchain to process payments exclusively. Our team at the Token Dispatch has been writing on the convergence of the two financial worlds almost every week now.

In crypto commentaries, I often see people selling blockchains themselves as the unique selling point (USP) because they enable speedy and low-cost transactions. Yes, it’s indeed cheaper to move money via blockchains. But that itself isn’t a significant factor in driving blockchain adoption, given that legacy infrastructure with relatively higher costs of moving money has stood the test of time for decades. Businesses don’t change their banking partners overnight because another bank offered a couple of basis points discount in processing their transactions. Financial habits are sticky, and businesses need more than just cost savings to feel confident about changing how they move, hold, and invest their money.

What helps here are quantifiable outcomes. For the masses to change how they move their money, they need to know how the entire money flow can be optimised. The pitch here should focus on how blockchains integrate with the platform that lets people hold, invest, and borrow money seamlessly.

In today’s guest op-ed, Sebastien Davies, Partner at Primal Capital, writes about how crypto’s infrastructure cannot elicit mass adoption, and what does.

Onto the story,
Prathik

The Infrastructure Mirage

For the better part of a decade, the global financial community has been transfixed by a fascination with “rails.” The discourse surrounding digital assets has focused almost exclusively on the mechanical throughput of blockchains, the cryptographic security of decentralised applications, and the theoretical elegance of smart contract logic. This was the Infrastructure Phase. An era defined by the building of “containers.” From 2020 to 2024, the industry frantically constructed the pipes, the vaults, and the gateways intended to modernise the movement of value.

During this period, the development of cryptocurrency markets was overwhelmingly infrastructure-focused because participation was structurally impossible without it. We built enterprise-grade custody platforms, standardised exchange APIs, and on-chain compliance services to solve five critical gaps: custody, exchange, execution, stablecoin utility, and regulatory reporting.

However, the industry is now confronting a fundamental truth of financial history. Infrastructure is a necessary precursor to activity, but balance sheets determine who captures the economics. The mere existence of a faster or more transparent rail does not, in itself, shift the gravitational centre of a market. Infrastructure answers the mechanical question of how institutions can participate, but it remains agnostic to a more consequential question of who captures the value. During the infrastructure-heavy years, the answer to the latter remained stubbornly traditional. Centralised market makers captured spreads, early holders captured appreciation, and validators earned transaction fees. What this phase failed to do was create new balance sheet structures that shifted where deposits must reside or fundamentally change the structure of credit creation.

A common counter-argument to this thesis posits that the “rails” are the primary driver of value because they lower the cost of entry, thereby democratising finance and naturally shifting economic power to the edges. Proponents of this view argue that the technology itself, by being open-source and permissionless, is the transformative force. While this is a compelling narrative for the retail-driven “crypto-native” world, it fails the test of institutional reality. In sophisticated financial markets, cost efficiency is a secondary concern to capital efficiency and risk-adjusted yield. An institution does not move a billion dollars because the transaction fee is lower; it moves it because the balance sheet upon which that capital sits offers a superior return or more efficient collateral utility. Infrastructure is a commodity that enables entry; the balance sheet is the strategic asset that determines the winner of the interest rate spread.

Financial history repeatedly demonstrates that infrastructure does not determine market power. Balance sheets do. The rise of the Eurodollar market in the 1960s did not require new payment rails or financial technologies. It required only that dollar deposits migrate outside the United States banking system. Once those balance sheets moved, a parallel dollar system emerged, operating at scale and largely outside domestic regulatory control.

We are now transitioning into a new Institutional Balance Sheet Restructuring Phase, an era beginning in 2025, where the “battleground” has shifted from the protocol layer to the allocation of liquidity. The first phase was about building the stage; the next phase is about the audience’s movement and their capital. A treasurer evaluating where to park cash in 2024 could technically use mature custody infrastructure to hold USDC, but the economics favoured traditional bank deposits, which offered FDIC insurance and competitive rates. Infrastructure was ready, but balance sheets had not yet shifted. That repositioning is only now becoming possible as the regulatory landscape moves from the abstract world of policy design into the concrete reality of implementation.

The next phase of crypto adoption will not be determined by infrastructure. It will be determined by where the balance sheets move.

The Implementation Gateway

For most of the past decade, institutional participation was not limited by a lack of imagination or technology, but by the structural impossibility of integrating digital assets into a regulated balance sheet. Institutions require more than just a functional wallet. Legal clarity, specific accounting treatments, and rigorous governance structures are minimum requirements. Without a recognised definition of “custody” or a clear path for compliance, the risk of “balance sheet contamination” was too high for any regulated entity to ignore. Large-scale adoption was held in a “waiting game” as banks and asset managers waited for a clear signal that they could deploy capital without incurring existential legal risk.

That era of policy debate has finally begun to sunset, replaced by a phase of operational implementation. The passage of the GENIUS Act in May 2025 served as the definitive catalyst, establishing a national regulatory framework for payment with stablecoins and finally providing the legal mandate for balance-sheet allocation. By providing a federal licensing process and requiring 100% reserve backing with government-approved instruments, the Act transformed digital assets from speculative novelties into recognised financial instruments. This shift was further cemented in August 2025, when the SEC concluded its long-standing investigation into the Aave protocol without enforcement action, effectively removing the regulatory “overhang” that had previously paralysed institutional DeFi participation.

The focus has now shifted to the regulators’ rulebooks. In February 2026, the Office of the Comptroller of the Currency (OCC) issued a comprehensive proposed rule implementing the GENIUS Act, establishing a framework for “Permitted Payment Stablecoin Issuers” (PPSIs). This move is significant because it provides the granular prudential standards (covering reserve composition, capital sufficiency, and operational resilience) that allow a Chief Risk Officer or an Asset-Liability Committee (ALCO) to sign off on a digital asset strategy. The passage of GENIUS has embedded blockchain regulation into the governance structures of the world’s largest financial institutions.

However, understanding why this shift is happening now also requires acknowledging the “balance sheet inertia” that defines institutional behaviour. Banks operate under strict regulatory capital ratios under which every dollar of risk-weighted assets must be backed by capital. If a bank loses deposits to a stablecoin, it must reduce its lending proportionally to maintain these ratios. It’s a painful and expensive contraction with knock-on effects across the economy. This explains why adoption has appeared slow. It takes six to eighteen months for full technological integration and even longer for the governance cycles of audit and board review to complete.

The current environment is one of “compounding acceleration.” As first-movers like JPMorgan, Citibank, and US Bancorp begin to launch stablecoin settlement initiatives, they’re sending a clear signal to the market that the risk of being first has been replaced by the risk of being last. We are seeing a competitive pressure phase where peer-bank participation reduces adoption risk across the entire sector. As these institutional constraints ease, the path is cleared for the migration of liquidity from legacy systems into the new, programmable containers of the digital age. This transition forces us to reconsider the very nature of where money lives, shifting our focus to the “containers” that will hold the next generation of global liquidity.

Where Liquidity Sits

To grasp the magnitude of the shift currently underway, one must first appreciate the historical stability of financial “containers.” In every monetary era, liquidity must eventually find a home. It’s simply a function of technical storage, but it satisfies the chronic global demand for safe, short-term assets. For centuries, this home was remarkably consolidated within a few well-defined structures: commercial bank balance sheets, central bank reserves, and money market funds. Each of these legacy containers functioned as an intermediary, capturing the economic value generated by the capital they housed.

The “Mathematics of Sitting” dictates that financial intermediation exists to solve a mismatch. Namely, that the world generates more cash from operations than it has immediate productive uses for, creating a permanent surplus of liquidity seeking safety. Traditionally, a commercial bank would capture this surplus as deposits, invest it in longer-dated assets like mortgages or corporate loans, and earn a substantial spread. This net interest margin (NIM) is the North Star for commercial and retail bankers. The bank’s shareholders are the primary beneficiaries of the “spread,” while the depositor receives a fraction of the yield in exchange for liquidity and government-backed insurance.

Digital asset infrastructure has introduced a new typology of “containers” that compete directly for this capital. These economic reconfigurations go beyond mere technological upgrades. When liquidity moves from a bank to a stablecoin reserve pool or a tokenised Treasury fund, the entity that captures the yield changes fundamentally. In a stablecoin reserve pool, for example, the issuer (such as Circle or Tether) earns the spread between the yield on the underlying Treasuries and the interest paid to token holders, which is often zero. This effectively transfers the “economics of sitting” from the commercial banking sector to the digital asset issuer.

Furthermore, these new containers offer a degree of transparency and programmability that legacy structures cannot match. Tokenised Treasury funds, which reached a market value of over $11.5 billion in March 2026, represent a structural evolution in which the yield of the underlying asset accrues directly to the holder. This creates a powerful economic incentive. A sophisticated treasurer no longer has to choose between the safety of a bank and the yield of a fund; they can hold a tokenised fund that serves as both a yield-bearing asset and a high-velocity settlement medium. By redefining where liquidity sits, digital infrastructure is not just building new rails; it is creating a competitive market for the balance sheets that support the global economy.

Stablecoins Drive Relocation

Blockchain dollars represent the first large-scale migration of liquidity onto these new financial balance sheets, marking a transition from digital currency being a novelty to becoming a core component of financial plumbing. The stablecoin market sits near all-time highs at $311 billion, growing at an annual rate of 50-70%. This growth eliminates the notion that this is a speculative phenomenon. We are witnessing a tangible “dollar relocation” away from traditional banking infrastructure and into programmable settlement systems.

The economic impact of this migration is most visible through the lens of deposit substitution. When a corporation or institutional investor moves $100 billion from a traditional bank deposit into a stablecoin container like USDC, the banking system experiences a profound loss of earnings power. In a traditional model, that $100 billion would support a bank’s ability to issue loans, yielding a net interest margin of roughly $3 billion annually. When that capital migrates to a stablecoin issuer’s reserve, those earnings are disintermediated. The bank loses the deposit, the ability to fund loans contracts, and the spread is instead captured by the stablecoin issuer.

This shift has profound implications for credit creation and financial stability.

Research published by Federal Reserve economists in late 2025 emphasises that high-adoption scenarios for stablecoins could lead to a reduction in bank deposits ranging from $65 billion to $1.26 trillion. This reduction has the potential to reshape how credit is provided to the economy. Regional banks, which rely heavily on stable deposit bases to fund local lending, are most vulnerable to this migration. As retail and corporate depositors seek the 24/7 settlement advantages of stablecoins, the traditional “float” that banks have long relied upon (earning spreads on in-transit payments) is rapidly becoming less compelling.

In response, the banking sector has moved from scepticism to a posture of participation.

The announcements by JPMorgan, Citibank, and US Bancorp to launch their own stablecoin settlement infrastructures in late 2025 and early 2026 are not attempts to “disrupt” their own business, but rather to retain their importance as liquidity containers. These institutions recognise that the economics of the future favour the issuer of the digital container. By becoming issuers themselves, banks hope to capture the reserve yields that would otherwise flow to new entrants. Of course, this first great cash relocation is only the opening act. As these new liquidity containers stabilise, the battleground is shifting toward the more complex world of collateral and the leverage that underpins global finance.

Programming Collateral

If the relocation of cash via stablecoins represents the first wave of this transformation, the migration of collateral represents a more fundamental restructuring of the financial system’s core leverage mechanics. Modern financial markets are essentially a vast network of collateralised obligations. The U.S. repo market alone, which facilitates the borrowing and lending of securities, moves between $2 trillion and $4 trillion every day. However, this critical infrastructure remains burdened by the “discrete settlement windows” of traditional banking. In the current world, collateral moves only during banking hours, and custody fragmentation means that a security held at one bank cannot be instantly used to satisfy a margin requirement at another. This friction creates periods where capital is locked, unproductive, and unable to respond to real-time market volatility.

Tokenisation transforms collateral from a static, location-bound asset into a programmable, high-velocity instrument.

By converting U.S. Treasuries and other real-world assets (RWAs) into on-chain tokens, institutions can move them 24/7 and settle them atomically. The growth of this market has been rapid; as of April 1, 2026, the tokenised RWA market reached approximately $28 billion, with tokenised Treasuries accounting for roughly half of that total. This surge is driven by institutional-grade products such as BlackRock’s BUIDL and Franklin Templeton’s BENJI, which allow holders to earn a 5% yield on underlying government securities while the tokens themselves remain mobile and deployable.

The true innovation lies in “collateral efficiency.”

In a traditional repo transaction, an investor might be required to accept significant haircuts or face delays of several days to unlock and move securities between custodians. In contrast, tokenised collateral is “composable.” An institutional investor can hold $100 million in BUIDL tokens, deposit them into a protocol like Aave at a 95% loan-to-value (LTV) ratio, and instantly borrow stablecoins to fund a tactical opportunity. The collateral never leaves the digital environment. Instead, it is continuously re-valued by automated price feeds, and any margin calls are handled via instantaneous, automated liquidations.

This shift moves the “economics of the dealer” toward the “economics of the protocol.”

In the traditional repo market, large dealer banks act as intermediaries, capturing a spread of roughly 50 basis points by borrowing at one rate and lending at another. In a tokenised ecosystem, collateral holders can self-match in DeFi lending markets, using software as an intermediary and capturing the full interest rate spread. We’re years away from seeing this at scale, but such a shift could move billions in annual earnings away from traditional dealer desks and toward protocol governance and asset holders.

To further appreciate the magnitude of the shift from cash to collateral, one must examine the institutional plumbing that has historically governed these movements. For decades, the global financial system has operated on a “T+X” settlement logic, where the “T” represents the transaction and the “X” represents a multi-day lag necessitated by manual reconciliation and inter-bank clearing cycles. In the traditional repo market, this delay acts as a silent tax on capital. When a dealer bank facilitates a repo, the collateral must be physically moved between custodians, often requiring human intervention to verify haircuts and legal titles. This creates a “liquidity moat” around the largest dealer banks, which derive their power not just from their balance sheets but also from their control over these proprietary settlement silos.

The mechanics of tokenised collateral dismantle this moat through atomic settlement. In a step-by-step institutional flow, the transition functions as follows:

1. Tokenisation: A high-quality liquid asset (HQLA), such as a U.S. Treasury, is moved into a digital wrapper (e.g., BlackRock’s BUIDL), turning it into a 24/7 mobile token.

2. Instantaneous Posting: Instead of waiting for a Monday morning wire transfer, a treasury team can post this tokenised collateral to a lending protocol or a prime broker at 10 PM on a Sunday.

3. Real-Time Valuation: Smart contracts utilise decentralised oracles to mark the collateral to market every few seconds, rather than once per day, allowing for significantly higher loan-to-value (LTV) ratios because the risk of a “flash gap” in valuation is mitigated by continuous monitoring.

4. Yield Preservation: Crucially, the investor continues to earn the underlying Treasury yield while the asset is encumbered as collateral, creating a “yield-on-yield” opportunity that is operationally cumbersome in legacy systems.

For a corporate treasury team or an asset manager, this shift is a fundamental revaluation of their idle assets.

In the traditional paradigm, a treasurer manages a “buffer” of cash that earns minimal interest simply to ensure they can meet sudden margin calls or operational needs. With tokenised collateral, that “buffer” can remain fully invested in yield-bearing Treasuries because the owner knows those assets can be converted to liquidity in seconds, not days. This removes the “liquidity discount” historically applied to long-term holdings.

For the banking sector, the implications are equally profound.

Banks have long profited from the “float” and the intermediation spreads of the repo market. As collateral becomes programmable and self-matching, that toll evaporates. This is why the emergence of institutional “plumbing”, like Anchorage’s Atlas network or JPMorgan’s internal tokenisation efforts, is so critical. They represent an attempt by financial institutions to build new silos before the old ones face competition. The transition from cash to collateral is the moment the financial system moves from a series of “discrete events” to a “continuous flow,” and the institutions that fail to adapt their balance sheets to this new velocity will find themselves holding increasingly static (and thus increasingly expensive) capital.

What appears to be just an improvement in settlement speed is, in reality, a reconfiguration of how capital is deployed, valued, and intermediated.

The S-Curve of Adoption

The migration of institutional balance sheets is not an overnight disruption, but rather a gradual absorption followed by eventual acceleration. It’s a “Web2.5” reality where blockchain technology is integrated into existing financial architecture rather than replacing it. Institutional adoption is currently being shaped by “balance sheet inertia,” a phenomenon where regulatory capital requirements, risk committee approvals, and legacy technology systems act as significant drag factors. Banks, for instance, cannot simply flip a switch to move assets. They must maintain strict Tier 1 capital ratios and ensure that any movement of deposits into a digital container does not force an expensive contraction of their lending books.

Despite these hurdles, the adoption of digital asset infrastructure is following a well-documented historical S-curve, similar to the multi-decade rollout of credit cards and the internet.

Between 2015 and 2024, the market was in an ‘experimental’ and ‘regulatory confusion phase’, where growth was tempered by uncertainty. We have now entered the ‘competitive pressure phase’ (2025–2026), characterised by regulatory clarity and infrastructure standardisation. During this phase, “You’re not the first, but you’re not the last” becomes the primary motivator for institutional treasurers. As more banks see their peers participating in stablecoin settlement or tokenised Treasury funds, the perceived risk of adoption drops precipitously.

The current market scale provides a baseline for accelerating compounding. With Fireblocks securing over $5 trillion in annual digital asset transfers and institutional tokenised asset markets growing rapidly, the “plumbing” of the new system has reached production-grade readiness. This infrastructure standardisation enables banks to build on proven systems rather than reinventing proprietary ones.

As we look toward 2027 and beyond, several “policy levers” remain that could further accelerate this migration. If stablecoin issuers were granted direct access to Federal Reserve Master Accounts, or if the current interest prohibition on payment stablecoins under the GENIUS Act were to be eased through affiliate “reward” structures, the flight of deposits from traditional bank ledgers to digital containers would likely accelerate meaningfully. The system is primed for a feedback loop where more stablecoin liquidity attracts more DeFi applications (likely permissioned), which in turn attracts more institutional capital, ultimately leading to a restructured financial landscape where the “battle for the rails” has been won, and the focus is entirely on the strategic management of the balance sheet.

The Winner of NIM

The transition from the infrastructure phase to the balance sheet phase marks the moment when the “digital asset” conversation moves from the periphery of technology into the heart of global macroeconomics. For years, the industry operated assuming that building better rails would inevitably lead to a better system. We now understand that rails are merely the invitation. The transformation occurs only when the capital itself relocates. The “Battle for the Rails” has effectively been won by a standardised, institutional-grade stack of MPC custody, tokenised Treasury funds, and federally regulated stablecoin frameworks. The new battle (one that will define the next decade of finance) is for the balance sheets that hold the world’s liquidity and collateral.

Heading into the 2027–2030 horizon, the structural advantages will accrue to those who can most efficiently manage these new “digital containers.” We should expect a persistent compression of commercial bank net interest margins (NIM) as depositors increasingly prioritise the 24/7 settlement and higher utility of stablecoin-accessible yields. Large corporates and institutional investors will likely shift their primary savings and treasury functions toward DeFi and RWA markets, where the middleman’s spread is minimised by protocol transparency. This is not the end of the traditional bank, but it is the end of the bank as a static, unchallenged warehouse of cheap capital.

The winners in this new era will be the “Web2.5” hybrids, or institutions that recognise they are no longer just lenders, but managers of programmable liquidity. By 2030, when the stablecoin market is forecasted to approach $2 trillion, the distinction between “crypto” and “finance” will have largely evaporated. The system will have fully absorbed the efficiency of the rail into the stability of the balance sheet. In this restructured landscape, financial power will not belong to those with the most innovative technology, but to those who control the containers in which the world’s liquidity and collateral ultimately reside. The battleground is set, and for the first time, the economics are up for grabs.

Crypto’s last decade was spent building infrastructure to enable institutions to participate. The next decade will determine where institutional balance sheets ultimately reside.

That’s it for today.

See you with the next one,
Sebastien Davies

Disclaimer: The piece was originally published in Edition 3 of Interop Markets

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.

Tough to root for the writer, the creator, the art in you when the world is looking for the thing that ships on time.

One thing I have been sitting with lately is that the most valuable thing is going to be being a human. As human as you possibly can. And tell me how hard can that be? We are making sense of being alive in a direction we didn’t plan, toward a conclusion we couldn’t have predicted.

Ben Roy said something here that I’ve been trying to say for a while. He got there first and got there better. His brain sees the world through his own particular life. His arguments, his places, his half-finished books. Mine does the same. Yours does too. No model has ever lived a day, lost an argument, or changed its mind about something it once believed completely. Yet.

It doesn’t have a nightstand. We do. That’s what you’re about to read.

—Thejaswini

Everyone I know that works in tech is having some flavor of existential crisis these days. People are worried that advanced LLMs are going to replace all knowledge work, possibly even in the near future.

I think part of this anxiety is reasonable. The AI labs have released impressive updates over the last six months, and the industry is alive with what feels like exponential change, but despite all the hoopla, what’s inconsistent to me is that I still find AI systems incredibly underwhelming when it comes to writing.

What I want to do with this essay is reason through why I think that’s the case and by extension wrestle with where I think human writers fit in a world of (increasingly) elite artificial intelligence.

Find more of Ben's writing

To start, I want to be specific about what I mean by writing because it’s important: writing is thinking on paper. It’s an extension of someone’s mind made manifest in the world through words with the purpose of passing along a container of meaning from one person to another.

I realize that’s a broad starting point, so let’s go one level deeper and split writing into two sub-categories: informational writing and interpretive writing. Informational writing is about the transmission of facts. It’s the world of documentation, instruction manuals, and neutral reporting. Interpretive writing is about sharing a perspective. It’s the world of essays, opinions, and commentary.

I acknowledge that life is more complicated than a simple binary, and, yes, every piece of writing has some subjectivity to it with how information is presented, but for the sake of argument, let’s say that all writing has a dominant intent, which is one of those two categories.

LLMs are great at informational writing. They can create a memo from a voice call, summarize an article, draft a simple contract, and so on. That’s useful for work purposes, but it’s not what I’m talking about here. When I say LLMs are weak at writing, what I mean is they’re weak at interpretive writing. They do poorly when trying to write with a consistent subjective lens. They struggle to have an opinion.

One final clarifier before moving forward: let’s use essays as the sole example of interpretive writing so we can agree on a concrete form to keep in mind as we discuss these ideas; and while we’re here, I’ll give some background on what an essay actually is so we’re ultra clear.

The word essay comes from the French, “essai,” which means to try or to attempt. The connotation there is of wrestling with words to work through a problem. So, we can say that an essay is struggle in written form, and so far, the machines can’t write in that manner. There are two reasons why I think this is the case. Let’s consider both.

The first problem with LLMs writing essays is the entire process of prompting one of these systems to get an output misunderstands what the act of writing is.

People fundamentally can’t prompt good essays into existence because writing is not a top-down exercise of applying knowledge you have upfront and asking an LLM to create something. AI agents also can’t create good essays for the same reason. Even though their step-by-step reasoning is more complex and iterative than human prompting, a chain of thought is still trying to accomplish a predefined goal.

By contrast, real writing is bottom up. You don’t know what you want to say in advance. It’s a process of discovery where you start with a set of half-baked ideas and work with them in non-linear ways to find out what you really think.

There’s a Flannery O’Connor quote that speaks to this where she says, “I write because I don’t know what I think until I read what I say.” I’m the same way. I don’t think people can skip doing the work of wrestling with ideas because that’s the core chaotic energy that underpins good writing. It’s how you figure out what you believe and how you want to communicate it.

There’s also a more philosophical extension to this point which is that human brains function differently than LLMs. They’re squishier. There’s a famous book, Thinking Fast and Slow by Daniel Kahneman that has a helpful framework to explain this difference. He talks about two types of thinking. There’s System 1 thinking, which is fast, automatic, and driven by pattern recognition. Then there’s System 2 thinking, which is slow, methodical, and reflective.

Humans think in both modes. LLMs also “think” in both modes.

For System 1 thinking, a human makes snap judgments all the time about people they see, posts they read on the internet, or whatever else. Similarly, an LLM can take in a simple query and give us a one-shot output in response based on the model’s training.

For System 2 thinking, a human takes in ideas, considers them, and through conscious effort can produce new ideas. Similarly, reasoning models can think for extended periods of time, consider a problem, and produce novel outputs.

The difference – in both types of thinking – is that LLM “thought” or inference is mathematically deterministic where human thought is delightfully messy.

Once a model finishes its training, its weights are saved and its thinking function becomes static. This is represented as a set of numbers, which from that point on don’t change. When a model thinks, it takes an input and iteratively chucks that input into a vortex of math (its weights and operations) and that then produces an output. It’s a predictive exercise based on statistical associations, and it only feels alive because when words are generated there’s a pseudo-random sampling involved in that output process.

Now, when a person thinks, our manner of thinking is more like taking ideas and filtering them through a set of “weights” that are constantly in flux, namely our accumulated life experiences and how we’re constantly interpreting and reinterpreting them. The process is alive.

What happens is we take an input thought and refract it through this prism of our humanity, our memories, the places we’ve lived, the arguments we’ve had, the people we know, and infinite other bits and bops of life experience to see multiple, blended, possible outputs on the other side simultaneously. It’s like the Pink Floyd album artwork below where normal light is the input, the rainbow is the set of outputs, and the prism is someone’s brain.

So, both LLMs and human brains can think and compose writing. That’s cool. The issue is the way these machines produce words leads to sterile outputs. Like, it’s always obvious to me when someone shares writing they’ve prompted because the outputs tell on themselves. The reason for this is because LLMs lack the unique perspective of a human mind at work (or play).

There’s a visual analogy I like to use when comparing these two kinds of writing which is that a good essay is like a cathedral of ideas. It’s a work of craft, beauty, and the quirks of time and place. But an LLM attempt at an essay is like a glass skyscraper. It’s functional. It’s a building, sure. But if you looked at a picture of it, you’d have a difficult time telling whether it was located in Portland, Atlanta, or Denver because it has no soul.

The best counterpoint here is that I simply have a skill issue on my hands. Maybe I’m not clever enough, and if only I knew how to more skillfully orchestrate LLMs or swarms of editorial AI agents, then I’d be able to create works of magical, overflowing craftsmanship with them.

Maybe.

I do think people with specialized knowledge of writing can create better outputs with an LLM. Agents can get to slightly better outputs as well through iterative loops. But there’s an upper bound to how far you can get in terms of making an LLM compose something compelling because writing is a complex, adaptive process and these systems operate on a flattened representation of that.

My next point is that LLMs struggle to generate meaningful writing because they lack context. They simply can’t take in (or process) the staggering amount of information that’s involved in the act of a person having an opinion.

I recognize these models have proven that they have PhD-level intelligence in some areas. They can win math Olympiads. They’re trained on most of recorded history. There’s no question that their form of alien cognition is impressive, but it’s just not equivalent to the neural network of my brain or yours when it comes to writing.

Let me use a graphic to explain this view more clearly.

If I write an essay, there are three variables in that process: myself as the writer, my audience, and the essay topic. Also, each of these variables is constantly changing.

A person is dynamic. They’re always in motion. Everything from how I’m feeling and what I’m currently reading to random mood swings and mental biases shape how I write something.

An audience is dynamic in the sense that their relationship to reading is evolving. Attention spans change, content formats rise and fall in popularity, again everything is in motion. For example, people who are familiar with LLMs have noticed that they tend to produce a lot of em dashes in their outputs, which in turn makes writers wary of using them – they don’t want their work to be discounted as “just” generated by AI.

A topic is also dynamic to varying degrees depending on what we’re talking about. Some subjects like ancient history are more static, but generally the world is a fluid place.

To dial this up further, the relationships between these variables are changing as well! A writer shapes the expectations of their audience through what they write over time, and in turn a writer’s audience informs what that writer writes about when they show interest in one subject over another. Both the writer and their audience also have evolving views of a topic as time moves forward and new information emerges.

To tie this together: there’s a lot of social complexity that goes into writing an essay, and an LLM has limited visibility into most of that complexity, which limits its ability to write with the flair of a human or on behalf of one.

Further, this idea that LLMs lack context is more than a raw data problem, it’s a relationship problem. Human writers have a qualitatively different relationship to this information than LLMs do. I don’t have data about myself… I am me. I have an intuition about my audience. When I write about a topic, I don’t know about it, I care about it.

LLMs don’t operate on the same plane. They have a view, but not my view. They have some context, and maybe often too much context, but not the context that’s relevant to play this game well. And since an essay really is just a snapshot in time of a person’s brain as a function of all these variables this makes human writing a lot more defensible than some people suggest.

After chatting about these ideas with friends, I’ve heard a couple good counterarguments to my issues with LLMs and writing. The most convincing of these is that everything will be solved when AI systems get access to infinite (or drastically larger) context windows.

This line of thinking says that if LLMs can offer users, say, 500 million tokens as a context window, then our machine companions will be able to reference everything about us: all our past pictures, writing, social media history, health data, recorded memories through things like wearable glasses, and so on. Then they’ll synthesize all of this data to approximate our decision-making process, which in turn would let them write (and think) on our behalf.

I do think there’s some merit to this. What’s happening right now with AI agents that initiate their sessions with access to memory files, conversation history, active skills, and other reference data is definitely gesturing at this future. See below for my basic bitch graphic.

I will say that I think this stuff is extremely cool, and I do think we’re going to get closer to modeling human consciousness through the combination of agent architectures and larger context windows. But no matter how weirdly fascinating it is to conjure up an agent with specific characteristics, I remain unconvinced that these systems will be able to write on my behalf.

I have two disagreements with this supersized context window argument, both of which are more philosophical than technical.

For one, even if you get infinite context windows, LLMs have an issue with hierarchy of information. There’s no clarity around the weight of some information over other information and how that changes in different contexts. So, Claude might have access to my references, but it doesn’t know when and how to use them. It might know grammar and rules of composition, but it doesn’t know when I might paraphrase something versus quote it outright. And there are far too many exceptions to any rule to ever codify someone’s intuition.

Next, I just find the comparison between an LLM and a human mind misleading. We’ve talked about this a bit already. It’s a case where reasoning by analogy leads to the wrong conclusion because LLMs don’t behave the way human brains behave.

I do think LLMs are useful machines. I’ve used ~20,000 words of my writing as a reference point, written up rules around how an agent should make editorial decisions on my behalf, then with that context I’ve run ideas against the best models that exist today. It’s created some fun outputs. But in every case, something feels off.

In music terms, I think LLMs are like a cover band that’s trying to perform a song where they get aspects of it “correct,” but they’re never able to embody what makes the song special because they lack the essence that is embedded in the original musicians. I might have a great time listening to the music, but I’m not paying $400 for a ticket to see PISS live.\

The AI maximalists will double down and make the case that, again, extra context is all you need to solve these issues, so one day an LLM will simply absorb mass amounts of your data and voila. Apparently, you can model everything about a human.

I disagree.

I’m not convinced we humans are even so legible to ourselves that we could articulate all the relevant context that an AI would require to mimic us.

My favorite example of this is the Greek concept of apophenia, which is the idea that humans perceive connections or patterns in random data. We don’t know why this happens, but this feature of our operating system is why we see different shapes in clouds, where, for example, you might see Bart Simpson and I see Pikachu. That creative expression is unique to each of us. How do you model that?

It’s a bit ironic, but our irrationality might be the most defensible asset we have as far as creative endeavours go. That bug is our best feature in a world of rapidly advancing AI systems because it makes us impossible to model.

Adam Mastroianni put this well in a blog post last year where he said, “Lots of people worry that AI will replace human writers. But I know something the computer doesn’t know, which is what it feels like inside my head. There is no text, no .jpg, no .csv that contains this information, because it is ineffable. My job is to carve off a sliver of the ineffable, and to eff it.”

Okay, this has been long. There are other counterarguments to what I’ve laid out here. I just want address one more.

Some folks bring up coding and point out that it’s a complex, creative activity like writing, and the follow up question is: if LLMs have changed how coding is done to the point where it’s “solved” why can’t writing be solved as well?

I’m no authority on coding, but I do see that LLMs have changed that activity. My pushback there is that coding and writing ultimately have a different telos.

The way you evaluate writing is whether it resonated with an audience or not. It’s about curating an entertainment experience for another human. Coding is evaluated by correctness. It’s functional where writing is experiential and concrete where writing is fluffy.

There are for sure multiple ways to code a piece of software to accomplish a task, but there are more standards to follow than there are in writing, plus, as much as engineers care about the readability of code, the ultimate audience for it is a compiler not a human.

I think LLMs work well in coding because it’s a coherent system: an agent can create code, run tests, iterate, then verify whether it will execute or not. But with writing, that doesn’t apply because it’s not a computational exercise, it’s an interpretive one. As a result, LLM code is mostly coherent these days, and when it’s not it can be concretely edited. LLM writing is way less coherent.

An analogy that I like to think about for LLM writing outputs is they’re similar to the bar scene from Good Will Hunting where there’s a dweebie Harvard student who presents what appears to be a fancy argument about US history, but it turns out he’s borrowing someone else’s thinking, and Will goes, “Were you going to plagiarize the whole thing for us? Do you have any thoughts of your own on this matter?”

That’s what it’s like talking to ChatGPT. It’s more like the illusion of intelligence than actual intelligence.

As I wind this down, I want to say that I’m optimistic about AI. I see LLMs as a complement, not a substitute to our intelligence. They’re super useful tools, and I use them in my writing today for ideation and sometimes for editorial feedback.

That said, even in a world where we see significant improvements in AI systems, I find it unconvincing that they will replace human writers. There’s a special je ne sais quoi about being human and the way we hallucinate our own custom models of the world based on a lifetime of experiences that I don’t think any machine will replicate.

Now, could I be wrong? Possibly. Will GPT 33 be able to write perfect essays? We’ll see. Models have surprised me and I’m sure they will continue to.

But my view is good writing requires something that only humans have, which is the ability to struggle and to speak about that struggle from the point of view of being in the trenches of life. It is sense-making from one human to another, and in my mind as long as humans endure as a species, there will be an audience for that.

Until then, keep writing.

A massive thank you to Devin, Finn, Nuk, and Luffistotle for feedback and review.

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 200,000+ 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.

If your wallet is public, your life is public. People can watch your balances, your trades, your positions, your timing, then say it’s “just data.”That’s why privacy keeps coming back as a narrative.

The issue is that “privacy” is not one feature. It’s five different problems.

This piece is to help you understand what you’re asking for the next time you tweet, ‘We need privacy’.

When people mention privacy, they often mean very different things:

• Intent privacy: observers shouldn’t see your trade before it lands

• Value privacy: your salary, net worth, or trade size shouldn’t be easily observable

• State privacy: positions, liquidation thresholds, vault states, and inventory shouldn’t be public by default

• Execution privacy: strategies shouldn’t be inferable from which paths you trigger and how.

• Graph privacy: observers shouldn’t be able to map who you pay, donate to, or coordinate with.

Many “privacy protocols” address one or two of these, while leaking the rest. Most leaks occur at the peripheries: wallets, Remote Procedure Calls (RPCs) - the bridges between wallets and blockchains, relayers, bridges between blockchains, exchanges, and human predictability.

That’s the frame that makes the landscape legible: separate the surface you want to protect, then look at the tool that protects that surface.

Intent Privacy

On many chains, a transaction doesn’t go straight from your wallet into a block.

Instead, it first sits in a public waiting room called the mempool, where pending transactions are visible before they are included. If you can see pending swaps, you can react to them. This creates the opening for Maximum Extractable Value (MEV) bots. These bots are automated systems that monitor blockchain mempools at high speed to find and exploit profitable transaction opportunities, such as arbitrage, liquidations, and sandwich attacks.

For example, a bot could spot your swap, buy right before you do, causing the price to move against you, and then sell right after you to capture the difference. You still complete your trade, but at a worse execution price.

Intent privacy addresses this specific ~12-second window.

The goal is to keep transaction details hidden until it’s too late for random observers to respond.

Private Transaction Delivery

The most practical approach to ensure the privacy of a transaction delivery is to change its path.

Your wallet still creates a normal signed transaction. The difference lies in where it gets sent. Instead of broadcasting it to the public mempool, the transaction is submitted to a private endpoint that forwards it to block builders. These builders assemble candidate blocks by ordering transactions. A validator then publishes one of those blocks.

This is what systems like Flashbots Protect offer: a route that keeps your transaction out of the public mempool until it’s included in a block.

Flashbots is also working on a concept called Single Unifying Auction for Value Expression, or SUAVE, which frames the problem as “order flow as its own system.” The idea is to collect user intents in a private environment, run an auction to determine who executes them, and then settle the outcomes on Ethereum and other chains.

While still evolving, this approach sets a clear direction: privacy should be applied before the transaction ever touches a public mempool.

Why this works is also why it has a trade-off.

It works because public bots never get the early look. The trade-off is that the private pipeline and the builders do get that early look. You’re shrinking the audience, not removing early visibility entirely.

Encrypted Mempools

Encrypted mempools aim to provide stronger privacy by ensuring no one can see your transaction before it is included in a block.

Instead of broadcasting a readable transaction, the network sees an encrypted blob. Observers can tell that something was submitted, but they can’t tell what it does.

Decryption happens only after the order is locked in. One common design uses threshold decryption, where decryption is split across a committee (The Shutter Network calls them Keypers). Each committee member holds a share.

When it is time to reveal the contents, sufficient shares are published to reconstruct the decryption key, allowing the transaction to be executed.

This method eliminates the “problem of the private pipeline being able to read everything early, but it introduces new assumptions: the committee must be online, and must not collude.

Intent privacy ends when the block is published. After inclusion, the chain reveals what it always has unless you pair it with on-chain privacy for values, state, or execution.

Value Privacy

Value privacy answers a simple question: Can I send money without the whole world seeing how much I sent?

On a normal public chain, the answer is no. Every transaction literally publishes the amount. That’s how everyone verifies balances.

Shielded systems change this by separating two things:

• the amount (kept private)

• the proof that the rules were followed (remains public)

Under the hood, the system stores your money as private records. You can think of them as encrypted receipts. Each receipt represents some amount, but only the owner knows what’s inside.

When you spend:

1. You prove you own a valid receipt

2. You prove you haven’t spent it before

3. You create new receipts for the recipient (and change)

4. You publish a proof that the total going out equals the total coming in

The chain checks the proof. If the proof is valid, the transaction is accepted, but the hidden amount remains unknown. This is the core design behind Zcash shielded transfers, a classic example of value privacy in production.

Penumbra uses the same general concept in a multi-asset shielded pool, where all value resides within a private pool and transfers occur via proofs rather than visible amounts.

But this type of privacy has limitations.

Even with perfect math, privacy can still fail.

The leaks usually stem from user behaviour:

• If you deposit a very unusual amount and later withdraw the exact same amount, observers can guess it’s the same person.

• If you enter and exit the private pool within minutes, the timing becomes a clue.

• If only a handful of people are using the private pool, the anonymity set is small.

• If you immediately move funds to a known exchange account, you reconnect your identity.

So, value privacy hides the number inside the system, but it does not automatically hide the patterns surrounding the system.

Graph Privacy

Graph privacy focuses on relationships. Even if you hide the amounts, a public ledger can still reveal patterns: who you send to, who you receive from, how often, and in what amounts. Over time, this network map can be more revealing than balances.

Most graph privacy approaches fall into two categories:

The first is pooled unlinkability. This is the “mixing” idea. A lot of users send funds into the same pool, and later withdraw them in a way that doesn’t publicly tie the withdrawal to a specific deposit. The chain still shows deposits and withdrawals.

Privacy comes from ambiguity. Observers can see the deposits and withdrawals, but they can’t reliably match a specific withdrawal to a specific deposit. Each withdrawal could plausibly belong to many participants. This is the core mechanism behind mixer-style systems like Tornado Cash.

The larger the pool, the less certain an observer is about any single link.

If the pool is busy and many people deposit the same amounts, you blend into the crowd. If the pool is small, the crowd collapses, and the graph starts to reappear through timing and amounts.

Another way to break the graph is to stop reusing the same receiving address.

If every payment goes to one public address, your inbound history becomes a permanent public feed. Anyone can cluster those payments and treat them as belonging to the same person.

Stealth addresses can change that pattern. Instead of a single visible destination, each payment lands at a new address that looks unrelated to the last one.

The sender generates a one-time address for that payment, and only the intended recipient can access it. To outside observers, it appears as money going to unrelated addresses.

This doesn’t hide the amount or the sender. It solves a narrower problem by preventing outsiders from linking all inbound payments to the same identity.

ERC-5564 standardises this pattern for Ethereum. It doesn’t hide the sender or the amount; it stops “all payments to Alice” from being obvious.

Graph privacy still leaks through behaviour. If you withdraw from a pool and immediately bridge to the same place every time, you create a new linkage point.

If you exit and immediately touch a KYC exchange, identity snaps back. If you keep the same timing habits, the graph becomes guessable again. These systems break direct links; they don’t erase your footprint.

State Privacy

State privacy seeks to address a DeFi-specific problem. Your balances, positions, liquidation thresholds, vault composition, and inventory should not be readable by anyone with a block explorer.

This matters because “public state” turns into a strategy leak. If your position is visible, other actors can predict how you’ll behave, when you’ll get liquidated, and what you’re likely to do next. Even worse, they can target you. A wallet with a visible liquidation line is basically a public scoreboard.

So what does state privacy change under the hood?

In a normal chain, the state is the thing everyone agrees on, and everyone can read. A lending protocol maps an address to position details. A vault maps an address to shares and debt. That’s what indexers and bots scrape.

Private-state systems stop writing those details in plaintext. They replace “public state keyed by your address” with “private state represented as hidden records,” and force updates to come with a proof that the update followed the rules.

Here’s the simplest way to picture it:

You still perform actions like “deposit collateral,” “borrow,” “rebalance,” or “swap.”

The chain still has to enforce constraints like ‘you cannot borrow beyond what your collateral allows’; ‘you cannot create value out of thin air’; and ‘you cannot spend the same private balance twice’.

But the chain enforces these constraints by verifying a proof, not by reading your position.

That’s why state privacy and ZK proofs often travel together. You need something that says “this update is valid” while keeping the underlying numbers private.

A useful concrete example is Aztec. The design centres on private execution on the client-side, with the network validating proofs and commitments. That lets a position exist without being stored as a readable table on-chain. You can do DeFi-like actions, while the public chain only sees proof-verified state transitions, not your raw position.

Where does state privacy leak? Mostly at the edges, again.

If you have a private position but periodically exit to a public DEX, the size and timing of those exits can reconstruct your behaviour. If you bridge in and out in predictable patterns, you create a linkage.

If you rely on public keepers for things like liquidations, your “private” position still needs some interface to the outside world, and that interface can leak.

Also, state privacy is hard to make composable. Public DeFi is like Lego because everyone can read everything. Private DeFi has to answer “How do two contracts interact when neither can see the other’s internals?”

The more complex the composition, the more careful the design needs to be.

State privacy is where privacy stops being about “hide a transfer” and becomes about “hide an ongoing financial posture,” and that’s why it’s harder, more useful, and more fragile at the boundaries.

Execution Privacy

This type of privacy goes one level deeper. It hides not only balances or positions but also how the computation happened. This matters for auctions, matching, solver logic, liquidation strategies, private order types, and anything where the strategy becomes exploitable once it’s visible.

There are two common approaches:

One uses Trusted Execution Environments (TEEs), where contracts execute within a hardware enclave, inputs are decrypted within the enclave, outputs are encrypted, and attestation verifies that the correct code ran.

Secret Network and Oasis Sapphire are examples of systems that use this approach to get private execution with lower proofing overhead. The trade is hardware trust and side-channel risk classes.

The other approach uses ZK proofs for private execution, where the system produces evidence that a program ran correctly without revealing the private inputs that drove the execution.

This approach is conceptually clean, typically heavier on tooling and performance, and tends to be rolled out gradually through limited scopes before it becomes general.

Execution privacy is weak in the same areas as every other type of privacy: timing, boundary interactions, and the access layer.

RPC: Tackling the Place Privacy

Even if your on-chain privacy is perfect, if your wallet uses Infura or Alchemy, that RPC provider can see your IP address, which addresses you control (because you queried their balances), what contracts you interact with, and your timing patterns.

In 2022, ConsenSys publicly acknowledged that MetaMask’s default RPC (Infura) collects IP addresses and wallet addresses. This is why protocol privacy often fails in practice: the access layer reveals everything before the cryptography gets a chance to work.

So, privacy is context-shaped. Different contexts shape privacy designs in different ways.

Trading mostly needs intent privacy. Payments need value and recipient-side graph privacy. DeFi wants state privacy. Bridges multiply correlation points. Institutions want confidentiality with a path to verification and accountability.

So, the question ‘Which privacy model wins?’ is usually the wrong one.

A more accurate question is: Which surface are you protecting? What assumptions are you making? And where does the information still leak when the user touches the real world?

That’s all for today.

Yash Chandak

P.S.: Thanks to Nitanshu, Joel and Kunal for helping with their input.

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 200,000+ 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: The views expressed in this article belong to the author, Yash Chandak, and not Token Dispatch.

For the longest time, crypto functioned much like tribes, built around chain maximalism. You picked one, learned its tools and apps, mingled with its community, followed its conference circuit, and rarely ventured far outside of that bubble.

That initial choice was shaped by whatever pulled you into crypto in the first place. Since the early decentralised finance (DeFi) grew on Ethereum, many early adopters ended up there. Later, the NFT hype and memecoin cycles drew new users into a different ecosystem, Solana. The result was a chain-first onboarding model: once you landed on a chain, you typically stayed there.

For most users, crypto felt like a collection of separate worlds, and you only interacted with one at a time. That mental model is now breaking down.

Today, crypto is increasingly experienced as a single market, not as a collection of chains. Activity is spread across many thriving ecosystems at once. Capital moves to where yields are highest, and assets are purchased where liquidity is deepest. Users no longer choose chains. They choose actions, and they care about results. They want to swap, earn, send, or pay, and they care about what those actions produce: profit, yield, or a transfer of money.

As a result, the dominant user model has shifted. Crypto is no longer chain-first. It’s asset-first. And wherever the asset is, users now expect to be able to go.

This new asset-first economy only works because interoperability (often shortened to interop) enables value to move freely across chains.

What is interoperability?

Interoperability connects the entire crypto ecosystem and makes it feel usable as a whole. Without it, every blockchain would be its own walled garden.

As defined in the State of Interop (2026):

“Interoperability is the ability for value, state, and intent to move seamlessly across independent blockchains. It is what allows composability to operate at scale in crypto, enabling coordination across otherwise separate ecosystems. For users, interoperability compresses a multi-chain ecosystem into a single, interconnected mental model of using crypto as a whole.”

A helpful way to think about this is to treat each chain as its own financial backend. They don’t naturally talk to one another. Assets, liquidity, and applications live in separate environments, each with different rules, costs, and trade-offs.

Interoperability is the layer that connects these backends.

If that sounds abstract, traditional finance offers a familiar parallel. Banks don’t natively interoperate either. What makes the global financial system usable isn’t the existence of one giant bank; rather, it’s the shared rails like Visa, SWIFT, and ACH that sit on top and move money between them. When you send money, you don’t think about which banks’ systems are involved. You just expect it to work.

Crypto is undergoing the same transition. Interoperability enables chains to feel less like isolated networks and more like a single financial system. It’s what lets users swap assets, chase yield, or move capital across chains without having to start from scratch each time.

For a long time, this was mostly theoretical. The tools existed, but the users weren’t ready. That has changed.

Users are already living multi-chain

Today, active crypto users routinely hold assets across multiple chains. While many still have a “home” chain (often Ethereum, especially among high-net-worth users), their on-chain behaviour shows that they are natively multi-chain. Capital moves based on risk, cost, and opportunity, not loyalty to a single chain.

Data from high-activity cross-chain users of Jumper, a multi-chain swapping and bridging exchange, illustrates this shift. Most users distribute their funds across several chains. They tend to park the majority of their capital on one or two primary networks, then move smaller amounts elsewhere when the market presents an opportunity.

This reflects how crypto is actually used today.

Users keep their capital where they feel most comfortable. When a new opportunity emerges, like a better yield, a new app, or a fresh narrative to speculate on, they use interoperability rails like bridges to move funds to wherever that activity is happening.

As a result, the act of ‘bridging’ funds, i.e., moving capital from one chain to another, has become a routine part of participating in crypto, and cross-chain volume is up 100x since 2022. Bridges tend to see their highest activity during these surges, as users move funds to whichever chain is the latest hotbed of speculation and attention.

Three things happening at once are amplifying the importance of interoperability:

First, users are accepting the multi-chain reality of crypto. Liquidity can never be confined to one corner of a global market. It will go where it gets the most bang for its buck, and the average user no longer expects everything to happen on one chain.

Second, chains are being launched at an unprecedented rate. 2025 saw one of the broadest waves of chain launches in the industry’s history. This includes crypto native ecosystems and chains run by fintech companies like Stripe. The surface area of crypto has expanded dramatically.

Lastly, tokenisation is accelerating rapidly. From equities and treasuries to private credit and real-world assets, everything is coming on-chain. At the same time, regulatory clarity, most notably through the GENIUS bill, has unlocked a wave of stablecoin issuance.

This is the reality of crypto today, and interoperability sits at the centre of all three shifts.

For users, interoperability is how crypto is used. Bridges are the primary way users move, trade, and express intent across chains.

For chains, interoperability is economic infrastructure. Without bridge connectivity, chains cannot attract external capital or onboard users at scale. A chain without interoperability (interop) is like a country without trade routes - isolated, constrained, and irrelevant in a global market.

For asset issuers, interoperability is a distribution requirement. Issuers no longer want to launch assets on a single chain; they want assets to be everywhere. That means availability across markets, chains, and liquidity venues. Doing so efficiently and at scale is only possible via interoperable rails.

Taken together, these indicate that interop is the layer through which crypto now expands. As users, chains, and assets multiply, interop becomes the mechanism that keeps the market liquid. Interop is crypto. Interop is inevitable.

State of Interop in 2026

Now that we’ve established that interoperability matters, what does the interop market actually look like?

It’s entering the phase that every infrastructure category eventually reaches: consolidation.

You see this pattern everywhere in finance and technology. In the early days of a category, value is spread thin. Many players coexist. Differentiation is weak. Growth comes less from being the best and more from being early.

Think back to the early days of fintech. Before Stripe emerged as the default payments layer, there were dozens of payment gateways. Before Plaid, every bank integration was custom. Before Bloomberg consolidated financial data, traders stitched information together from multiple terminals and feeds.

Interoperability followed the same arc. As blockchains proliferated, barely any interop protocol saw traction. That phase is now over.

As categories mature, value stops accumulating in the middle. It concentrates on the edges. You either become exceptional at one specific job or you operate at a scale that makes you unavoidable. Everything in between gets squeezed.

In practical terms, this means two things:

• Users gravitate toward the best possible solution for a specific job.

• Developers default to infrastructure that’s proven, reliable, and widely adopted.

Interop today is a textbook example of this dynamic.

On one end are the specialists. These teams focus obsessively on a single outcome and optimise everything around it. They don’t try to be platforms. They aim to win one job.

Think Stripe, which started with payments APIs; Wise which focuses narrowly on cross-border FX or Robinhood, which makes trading simple and cheap. In interop, intent-based bridges fall into this category. Teams like Relay and Gas.zip realised most users just want fast, cheap cross-chain swaps, so they built for speed and low cost, and nothing else.

At the other end are the infrastructure giants. These are generalised platforms with real distribution, deep integrations, and network effects.

Think SWIFT in banking, Visa in payments, or AWS in cloud computing. In interop, large platforms that power many apps, wallets, and chains sit here. This includes the likes of LI.FI, LayerZero, and Wormhole.

Once a team proves it can operate reliably at scale, the market gives it permission to expand into adjacent products. This is happening in interop, with teams expanding into new product offerings like swaps and yield aggregation.

What struggles is the middle.

Projects that are neither sharply focused nor widely adopted end up in a dangerous no-man’s land. They’re too generic to dominate a niche but too small to be foundational infrastructure. They move slower than specialists and lack the distribution of platforms.

This is where consolidation happens. Some teams pivot. Some shut down. Others are acquired for talent or technology, like Circle’s acquisition of Interop Labs’ team and IP.

Interop in 2026 is no longer about being “good enough.” As with every mature market, specialisation and scale are where value now accrues.

Where We’re Headed

If you step back, the shift is obvious.

Crypto now asks you what you want to achieve, irrespective of where you are.

Users don’t need to be conscious of which chain they are on. They act: swap, earn yield, send money, and get exposure to an asset, irrespective of the chain they reside on.

And increasingly, those actions just happen, even when they span multiple chains underneath.

This is the new reality of crypto. Chains still exist, but mostly as infrastructure. To the user, crypto is starting to feel like a single, global system of markets and money.

However, this still requires the users to know which interoperability tools to use and how to use them.

This is where the next step comes in: everything will soon be embedded in app buttons, and things will just work.

Today, interoperability helps you move between chains. Soon, it will help you act across them - automatically, atomically, and in one seamless flow. Swapping, lending, rebalancing, paying, issuing assets, coordinating agents, all with a single click, even if ten different chains are involved.

This is what an asset-first, action-first crypto experience looks like.

For the reader, the implication is that on-chain capabilities are increasing rapidly. Things that once required planning, bridging, and manual coordination are collapsing into single decisions and single clicks.

All to say:

The chain-first era is behind us.
The action-first era is arriving.

And interoperability is what makes it inevitable.

Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 200,000+ 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: The author, Arjun Chand, is the head of narrative at LI.FI, a bridging and DEX aggregation protocol. The views expressed in this article belong to the author and not Token Dispatch.