The Second Curve

What survives contact

Begin with what a weekend of tracing leaves standing, because it is not much. The "$5 trillion waiting on the sidelines" traces to a pseudonymous account laundered through headlines. The "$6.6 trillion deposit exodus" is the stock of every US checking account, repackaged as a forecast by a bank lobby. "Tokenization 100x" has no author at all. Meanwhile the sober numbers sit in plain sight: DeFi holds $76 billion, down 57% from its 2021 peak and near its one-year low. The bill itself — read, not skimmed — spends its machinery on exchanges, brokers, and custodians, hands DeFi a carve-out and a study, and sets rulemaking clocks that a shrunken agency will take years to honor. Europe already ran the clean experiment: full regulatory clarity produced 102 licensed firms and less than a billion dollars of euro stablecoins against three hundred billion dollar-pegged. And the one law we did pass, the stablecoin act, just finished a full year in which the aggregate it was supposed to ignite grew slower than the year before.

So the lazy bull case is dead on arrival: clarity, as such, has never created demand anywhere we can measure. But the lazy bear case dies on the same table, and it dies on a subtler instrument.

The two futures in one chart

There is exactly one on-chain market that grew through the entire bear: tokenized real-world assets — Treasuries and credit wearing token wrappers — from twelve million dollars four years ago to eighteen billion today. I fit that series properly, and the result is the most honest thing I can show you about this moment.

One series, two laws, five orders of magnitude apart. The unconstrained growth everyone's forecasts assume runs to $2.3 trillion by 2028; the saturation curve the data slightly prefer finishes at roughly today's level. Sixty months of history cannot referee between them. The chart is not a prediction. It is a portrait of where belief takes over from measurement. Provenance: own DefiLlama-derived monthly series; fits and figure code archived with the notes.

Extrapolate the growth rate everyone implicitly assumes and you get $2.3 trillion in two years — a number the fit itself disowns. Fit the honest alternative, the S-curve, and it says this market is nearly finished, its ceiling within a few percent of where it sits today. The statistics mildly prefer the second reading. Growth is visibly decelerating — from roughly 250% a year two years ago to about 66% now. And yet the series printed a fresh high last month, and a ceiling estimated during a slowdown is a floor on the truth, not a measurement of it.

On five years of data, the difference between "this market is done" and "this market goes to trillions" is not a fact. It is a belief. The data cannot referee it — and that is the finding.

Every technology that mattered has lived this exact moment. The first curve — the frenzy product, the thing that pulled the crowd in — saturates and rolls over, and the numbers, honestly read, say the story is finished. The people who kept building were not reading different numbers. They were holding different beliefs about what the numbers could not yet contain: that a second curve existed, made of different products, sold to different people, on rails the first curve had paid for. Railways after the mania. The web after the dot-com bust. Deployment ages, as Carlota Perez taught, begin precisely when the measurable story ends — and they begin with law, which is the only thing the CLARITY Act, at its best, actually is: not fuel for the curve we had, but initial conditions for one nobody can plot yet.

What the laws forbid

Saying "it's a belief" is not a license to believe anything. The fits that cannot pick your future can still forbid several, and a builder should treat these the way an engineer treats thermodynamics — not as commentary, but as constraint.

The first constraint: printed yield dies of its own success. Across every live pool I could pull — sixteen thousand of them — yield that comes from token emissions decays with the size of the pool roughly nine times faster than yield that comes from fees. This is not sentiment; it is division. An emission split among more capital is a smaller number, mechanically, before a single farmer sells. Farming didn't die of a bear market. It died of arithmetic, and arithmetic does not have cycles. Whatever the next finance pays its users, it must be earned — fees, coupons, premiums, spreads — because earned yield is the only kind that survives scale. Any product whose advertised return shrinks as it succeeds is pre-failed, and no statute amends that.

The second constraint: the crowd never returns through the door it was burned in. Not in 1929, not after any mania since. The trusts gave way to mutual funds, the mutual funds to the 401(k), the brokers to the apps — every generation of ordinary people enters through a wrapper that asks less of them than the last one, never more. Farming asked more: seed phrases, gas, bridges, vigilance. It topped out as a subculture, and it will not be revived, because revulsion is not a price level, it is a memory. Whoever brings people back will not bring them back to DeFi. They will bring them back to a button.

The third constraint: a statute is invisible on a product clock. The stablecoin act's first full year produced proposals, zero approved issuers, and an aggregate that decelerated. The CLARITY Act, if it passes, hands its real decisions to years of rulemaking by an agency that just lost a fifth of its staff. Nothing unlocks on the day a bill passes. Nothing unlocked in the year after the last one did. Anyone whose plan requires the law to move their metrics this cycle has a plan that the only available precedent already falsified.

What a builder would have to believe

Within those walls, here is the future I find myself believing — stated as suppositions, because that is what they are, and numbered, because beliefs you won't enumerate are beliefs you won't be held to.

First: the unlock is permission, not money. There is no five trillion dollars idling in a waiting room. What the bill reprices is the legal-risk premium — the invisible tax on every institutional decision to touch public rails. Permission converts to capital slowly, through committees and custodians, on a horizon of years. Suppose, then, that the right product to build today is the one that is ready when permission converts — not the one that needs the conversion to have already happened.

Second: the second curve starts from assets, not apps. The first curve was application-first — come to our venue, learn our interface, chase our yield. The one market that grew through the winter is the opposite: familiar assets, Treasuries and gold and bitcoin, acquiring capabilities where they already sit. Suppose every idle balance in the world — the coins held and never sold, the gold vaulted and never worked — is unstarted inventory for the next finance, and the product is whatever makes that inventory earn or borrow or hedge without asking its owner to become someone new.

Third: the tail is the map. Today's tokenized-asset market is five balance sheets and a rounding error — the top ten names hold three-quarters of everything, a concentration steeper than the natural law of mature markets. Read as a verdict, that's damning. Read as a map, it is the whole opportunity: the second curve, if it comes, is by definition the filling-in of that empty tail — the next hundred issuers, the small asset classes, the jurisdictions nobody has wrapped yet. The incumbents are not the competition; the emptiness is the market.

Fourth: the dollar is the tailwind, and fighting it is the losing trade. Strip the crypto vocabulary and this bill is a great power annexing a parallel settlement system before a rival does — the eurodollar playbook at internet speed, with stablecoins as the offshore dollars and Treasuries as the collateral. Europe wrote elegant rules and got no rails; the dollar got the rails without asking. Suppose the next finance is dollar-denominated on public rails whether anyone likes it or not, and build accordingly — the neutral, uncensorable corner will persist and matter, but it will be the non-aligned movement of this cold war, not its winner.

Fifth: the boring decade is the moat window. Rulemaking will eat years. Institutions will move at the speed of their slowest lawyer. That gap between annexation declared and annexation administered is not dead time — it is the only period in which small teams can build unassailable positions in plain sight, because the people with the balance sheets to crush them are still waiting for the rules. The bear market builds the infrastructure the bull market takes credit for. This is that bear market.

Sixth: retail returns as owners being served, and never learns the plumbing. The flows that actually arrived this cycle went through the most familiar wrapper on earth — an ETF ticker in a brokerage account — and beat every forecast. The flows that never arrived were the ones that required new behavior. Suppose the return of ordinary people looks like a refinancing moment, not a casino moment: yield and protection on what they already hold, inside the app they already open, with the chain as invisible as the clearing house behind a stock trade. By the time your brother-in-law is earning on his coins again, he won't know a protocol is involved — and that absence of knowing is what victory looks like.

Seventh: the ending is absorption, and the disillusion is the price. Ride the dream to its honest end and what survives is not DeFi as its founders meant it — permissionless, retail, anti-bank. It is public settlement rails as anonymous and indispensable as the internet's protocols, with traditional finance operating on top, and the word "DeFi" retired the way nobody says "information superhighway" anymore. The settlers will not fly the builders' flag. That is how every general-purpose infrastructure has actually won, and the builders who make peace with it early will build the things that get absorbed — which is to say, the things that last.

What happens next

Beliefs without a sequence are a mood, so here is the future I would actually write down — dated, concrete, and offered as supposition, not prophecy.

Through 2026: the bill passes in August or it doesn't, and either way the visible effect is a sentiment trade and then silence. If it passes, the real event is invisible: legal departments at a few hundred institutions change the answer in a memo from "no" to "not yet." That memo is the whole year's unlock. Nothing prints on a chart.

2027 is the year of provisional everything: venues file provisional registrations and keep listing what they already list; the first banks pilot custody under the new balance-sheet relief; the tokenized-asset tail either begins to fill — new issuers, smaller assets, the concentration number finally falling — or it doesn't, and we learn early which future we're in. Somewhere in a large fintech, a product team ships the first version of the button: earn on what you hold, inside the app you already open. It will look like nothing. It will be the retail era starting.

2028 is the distribution wave: rules land, procurement cycles that started in 2026 conclude, and the brokers, exchanges, and fintechs that own the customer switch on embedded products at scale. This is when the absorption becomes literal — protocol teams get acquired the way ISPs and portals were, for their plumbing and their licenses, not their tokens. The M&A wave, not a bull market, is how the second curve announces itself.

By 2030, if the suppositions hold, on-chain settlement is boring, dollar-denominated, and everywhere; "TVL" has stopped being the metric because the interesting number is assets served, not capital parked; and the word DeFi has completed its retirement. The future will have arrived the way infrastructure always arrives — unannounced, beneath products that never mention it.

Who wins, who hits the wall

Every reallocation has both sides, and it is kinder to name them than to let people discover which side they're on.

The wall, first. Standalone retail dApps are shrinking into it now: their whole model — acquire the user, teach the wallet, win on yield — loses to regulated distribution the day a brokerage offers the same product behind a familiar login. The user they spent five years educating is precisely the user Coinbase and the fintechs will serve without education. Emission-subsidized protocols are against the same wall for a different reason: their yield is a marketing budget wearing a percentage sign, and the arithmetic section above is their obituary. Non-dollar stablecoin strategies — the compliant euro plays — already hit theirs; MiCA was the experiment and $0.71 billion against $306 billion was the result. Offshore venues serving gray-zone flow will feel the annexation as a slow tide: every rule that lands makes onshore cheaper and offshore lonelier. And the chains themselves — the L1s and L2s selling blockspace as the product — face the quietest wall of all: blockspace is the one input whose supply is infinite, and value is already migrating up-stack to the assets and the distribution.

The pivots are not mysterious. If you own users but not licenses, become infrastructure for someone who owns licenses — sell your engine to the distributor you were trying to out-market; invisible and indispensable beats branded and besieged. If your yield is emissions, convert to fees this year or return the capital; there is no third option the exponent permits. If you built for neutrality, serve the neutrality demand honestly — the sanctions-adjacent, the de-banked, the jurisdictions the annexation excludes; it is a real market, permanently smaller than the dream and permanently alive. If you sell blockspace, pick a vertical and become its venue of record — the chain that clears tokenized credit, the chain that clears gold — because "general-purpose" is exactly what the incumbent rails will absorb first.

The winners are already visible in the data's one growing corner. Issuers of familiar assets in compliant wrappers — not just the BlackRocks, but the next hundred small issuers the empty tail is waiting for. Distribution that already owns the customer: brokerages, exchanges with licenses, fintech apps — the refi moment is theirs to switch on. The custody-and-registration layer the bill literally writes into existence. The Treasury market, collateral to the entire arrangement. And — the one the map points at hardest — the overlay layer: whoever stands between held assets and the structures that make them earn, borrow, and hedge. The idle balances are the largest unserved inventory in this market; the ETF precedent says their owners move when moving requires nothing of them; the button that serves them has to be built by someone, and it will not be built by the incumbents first, because the incumbents are still waiting for the rules. That gap — between inventory that exists and service that doesn't, during a decade in which the giants are legally parked — is the most precise description of the opportunity I know how to give.

So if the question is what do I build, now, with this map — it is one of four things: the button (embedded earn, borrow, protect on assets people already hold); the tail's toolchain (issuance, compliance, and servicing for the hundred issuers who don't exist yet); the picks and shovels of registration (the boring machinery every provisional filer will need by 2027); or the honest neutral corner. Everything else on the current landscape is either the first curve's aftermarket or someone else's acquisition target.

The scoreboard

A belief that cannot lose is marketing, so mine come with a scoreboard. If the second curve is real, the asset series that today looks finished must break decisively above its fitted ceiling — call it thirty billion within a year of the bill becoming law; if instead it stalls near twenty, the ceiling held and I was describing a hope. If the tail is filling in, concentration must fall — the day the tokenized-asset market stops being five names is the day the deployment age actually started, and that number, not TVL, is the one I will be watching. If the statute matters on any horizon that should change a builder's plan, growth in the dollar aggregates must visibly accelerate by late 2027 — and if it doesn't, the boring-decade reading wins again. I have written down where each of these numbers comes from and will keep pulling them; if the market makes me wrong, the series will say so before I do.

The bill may not even pass — the honest handicappers now put it near thirty percent this year, and the window closes in two weeks. It matters less than either side pretends. The assets kept growing through ambiguity; the rails keep getting cheaper through indifference; the permission, if not this August, comes eventually, because great powers do not leave settlement systems unclaimed. What the moment asks of the people building is not optimism about a bill. It is the older, harder thing every deployment age has asked: to keep building through the years when the numbers say nothing yet — because the instruments that detect a future before it exists have never been fitted curves. They are builders.