The funding market for programmable capital.

In the fourteenth and fifteenth centuries, commercial activity across Europe was expanding faster than the tools available to finance it. Capital existed — merchants held real wealth in goods, land and credit — but markets were fragmented. Florence, Venice, Rome, Bruges and London each ran their own currency, their own legal customs and their own settlement rhythm. Moving coin between them was slow, costly and dangerous.

Merchant banking networks, correspondent relationships and increasingly sophisticated accounting practices grew up to answer that problem. Claims could move between cities on paper faster and more safely than specie could move by road. Trust, once local, became portable.

Today's markets look nothing like fifteenth-century Florence, and it would be a mistake to pretend otherwise. But a structurally similar problem is emerging. Financial assets are beginning to exist as programmable instruments — tokenized equities, tokenized commodities, stable-value cash — while the infrastructure to finance, collateralize and price those instruments remains fragmented across exchanges, custodians, brokerages, chains and wrappers.

Tokenization creates the asset. Funding turns it into capital.

That is the argument this paper makes, and it is the reason this project carries the name it does. Not because a fifteenth-century bank invented modern finance — it did not, and this paper is careful not to claim otherwise — but because the underlying pattern, financial infrastructure arriving to make a new form of property genuinely useful, is old, well documented, and worth understanding before building the next version of it.

The Medici Bank was founded in Florence in 1397 by Giovanni di Bicci de' Medici^[2], who had trained under his uncle Vieri at the Rome branch of an earlier Medici banking venture before striking out on his own^[2]. Under Giovanni and later his son Cosimo, the bank grew into a federation of partnerships operating branches in Florence, Rome, Venice, Bruges, London and other commercial centers^[1], becoming one of the largest and most respected banking networks in fifteenth-century Europe.

The instrument at the center of that network was the bill of exchange: an order, written by a merchant in one city, instructing a correspondent in another city to pay a named party a given sum, after an agreed interval, at an agreed rate of exchange^[3]. An exporter could sell his bill to a banker and receive cash before his buyer had actually paid — the banker, in effect, financed the exporter on the security of the bill^[3]. No coin needed to physically cross the Alps or the Channel for the trade to settle. What moved was a verified claim.

This is the useful parallel, and the paper will not stretch it further than the history supports. Bills of exchange did not eliminate risk or replace judgment — a bank still had to trust its correspondents, price the exchange rate, and absorb the occasional default. What they did was let a claim on value travel faster and more safely than the value itself, across a market that had no single settlement authority. That is a financing-infrastructure problem, not a currency problem, and it is the same category of problem tokenized assets raise today.

A merchant ledger recorded who owed what to whom, when it was due, and what security stood behind it. In 1494, Luca Pacioli published the Summa de Arithmetica in Venice, which included the first printed, systematic description of double-entry bookkeeping — a technique he called "the method of Venice," already in use among Italian merchants for roughly two centuries before he wrote it down^[6][5]. Pacioli did not invent the method; he made it legible and transferable, which is arguably the more important act. A technique that only lives in one merchant's head is not infrastructure. A technique that can be taught, copied and audited is.

A blockchain is also, fundamentally, a ledger: a record of who holds what, updated according to agreed rules. What it adds to Pacioli's method is not better arithmetic — it is shared state that many parties can read without trusting a single custodian, programmability that lets the ledger itself enforce rules rather than merely record outcomes, atomic settlement where a trade and its payment either both happen or neither does, and continuous verifiability, so the state of the ledger is never a matter of waiting for someone's monthly statement.

None of that makes a blockchain a better bookkeeper than a well-run merchant house — Pacioli's method, applied with discipline, is sufficient for its purpose. What a public, programmable ledger changes is who can build on top of the record, and how quickly. That is the opening this paper is written into.

Take a tokenized stock. What can it actually do? If the honest answer is buy, sell, transfer, the market around it is incomplete — a spot market, not a capital market. A traditional security participates in a much wider set of financial activities: it can be posted as repo collateral, pledged for margin, lent out through securities lending, hedged with derivatives, transformed through structured products, and financed by a dealer's balance sheet^[7]. Those activities, taken together, are most of what makes a security useful to an institution that holds it.

Tokenization ≠ financialization. Tokenization is the substrate. Financing is the next layer.

This is not a criticism of tokenization — quite the opposite. Making an asset programmable is a genuine and necessary precondition. But it is a precondition, not a conclusion. The financing architecture that traditional securities markets built up over a century of practice, repo, margin, securities lending, structured credit, prime brokerage, does not arrive automatically the moment an asset becomes an ERC-20 token. Someone has to build it.

Start with the plain version. Someone owns a valuable asset. They need cash. They do not want to sell — maybe they expect the asset to appreciate, maybe selling triggers a tax event, maybe they simply want to keep the position. So they borrow against it instead, using the asset as security, and agree to repay on a set date.

That is, in economic substance, a repurchase agreement. In its traditional form, a repo is a sale of securities coupled with an agreement to repurchase them at a specified price on a later date — "economically similar to a loan collateralized by securities," in the Federal Reserve Bank of New York's own description^[7]. The proceeds of the initial sale correspond to the loan principal; the difference between the sale price and the repurchase price corresponds to interest^[7]. Repo is not an exotic corner of finance — it is one of the core funding mechanisms beneath modern securities markets, used constantly by dealers, funds and banks to finance inventory and manage short-term liquidity^[8].

Medici brings the same economic structure onchain, for tokenized collateral, with five parts:

• Collateral — the tokenized asset posted into an isolated market.
• Settlement asset — USDG, drawn against the collateral.
• Haircut — the buffer between collateral value and the amount advanced.
• Rate — fixed the moment the position opens.
• Maturity — an explicit date, agreed before a dollar moves.

Most DeFi lending is perpetual and variable: a position can sit open indefinitely, and its rate drifts with pool utilization every block. That design is well suited to buy-and-hold borrowing, where the borrower wants continuous liquidity and does not mind an unpredictable carrying cost.

A treasury desk planning next quarter's cash needs, a fund pricing a hedge, or a market maker financing inventory overnight wants something different: certainty. A fixed rate and a known end date let a borrower plan a liability the way a lender plans an asset. Overnight, 7-day and 30-day terms are simple examples of the same idea at different horizons — and each one, priced consistently, becomes a point that later terms can be compared against.

Worked example. $V_c=\$1,000,000$, $h=18\%$, $r=5.10\%$, $T=1$ day, $D=365$: $A=\$820,000$, $C\approx \$114.58$, $R\approx \$820,114.58$.

A haircut exists because collateral value can move between the moment a position opens and the moment it must be unwound. The size of that gap depends on how volatile the asset is, how liquid it is under stress, how much the underlying market can gap between price checks, how confident the protocol can be in its oracle, and how concentrated or correlated the position is with the rest of the book.

A sufficiently deep, sufficiently trusted funding curve is designed to create the basis for fixed-rate lending, principal/yield-token structures, rate swaps, structured products, collateral pricing, portfolio financing, risk systems and institutional analytics. None of that exists yet as an established benchmark — it is a design ambition this architecture is built to support, not a claim that other protocols currently reference it.

The mechanism this is meant to describe:

An investor holding $1,000,000 of SPY exposure has, broadly, two options if they need cash: sell the position, or finance it. Selling realizes the exposure change immediately and may trigger tax or timing consequences. Financing at an 18% haircut instead raises $820,000 in USDG while the position, subject to the market's structure, keeps its economic exposure.

That is not a claim that financing is always the better choice. It carries its own costs and risks: the funding cost itself, continued exposure to price risk on the collateral, and liquidation risk if the buffer is breached before maturity. The point is narrower — financing and selling are genuinely different decisions with different risk profiles, and a market that only supports the second one is missing half the toolkit.

A normal funding curve rises with tenor: locking a rate for longer typically costs more, because the lender is giving up optionality for longer. An inverted curve, near-term funding scarcer or more expensive than longer-term funding, could indicate acute short-dated demand, stress, or a temporary supply imbalance rather than a durable repricing of risk.

Each market has independent collateral, oracle, haircut, caps, liquidity and liquidation rules. A default in one market cannot, by construction, draw on the collateral or reserves backing another.

Tokenized equities and ETFs introduce a category of risk crypto-native collateral does not always have: an underlying reference market with its own hours, halts, corporate actions and the possibility of a stale or gapped price outside those hours. A protocol that stays technically online 24 hours a day cannot manufacture a verifiable price for an asset whose primary market is closed.

No liquidation without a price the protocol can verify.

A stock split changes a share count without changing economic value; a token multiplier has to reconcile with that split before price-dependent logic can trust the feed again. A merger or delisting can eliminate the reference market entirely. None of this is handled by a single universal rule — it is handled by pausing the affected market until the reconciliation is confirmed.

Medici is designed to be capitalized from two directions. The first is protocol cash fees — revenue earned from financial activity in the repo market itself. The second is the $MEDICI transaction tax — a defined fee on eligible token activity. Both can strengthen protocol resources, but they are not the same thing, and this paper treats them separately rather than blending them into one undifferentiated number. Developer or protocol revenue is not the same balance-sheet line as reserve assets set aside against future losses.

A recurring pattern in earlier DeFi cycles was using token emissions to manufacture activity that would not otherwise exist — paying users to show up, in a currency the protocol could print for free. That approach can bootstrap volume; it does not, by itself, build a durable business, because the moment the emissions stop, so does the incentive to stay.

Medici's design goal is revenue from financial activity that would exist whether or not a token existed at all:

• origination and matching fees on every repo
• rollover and refinancing fees at maturity
• liquidation economics on positions that fail their terms
• treasury income on protocol-owned reserves
• future curve-data and benchmark licensing, and institutional routing
• future structured-product issuance and credit-market fees, as those markets are built

The $MEDICI tax described in §22 complements this revenue; it does not replace it. New tokens that merely rename themselves as revenue are still emissions.

A defined transfer tax of 5 to 10% on eligible $MEDICI activity contributes to protocol reserves. This is not framed as holder rewards, farming yield, or a reflection-token redistribution mechanism, and it is not something this paper will describe that way. The premise is specific: economic activity in the token should strengthen the balance sheet behind the financial system it funds, not simply pay out to whoever is holding at a given moment.

Real financial activity: more assets move onchain → more eligible collateral → more repo demand → more funding activity → more protocol revenue → larger reserves → greater market depth → better funding markets → more integrations → more collateral.

Token economy: $MEDICI activity → transaction tax → protocol reserves.

The two loops meet at the same balance sheet. The interactive version of this diagram, with a interactive stage inspector for each step, is built into the homepage — see medici.markets/#reserve.

Robinhood launched the public mainnet of Robinhood Chain on July 1, 2026, introducing Stock Tokens to eligible users in more than 120 countries, alongside lending, borrowing and decentralized-exchange infrastructure and a stated focus on tokenized real-world assets^[10][12]. That combination, EVM compatibility, an RWA orientation, and native decentralized-exchange liquidity for tokenized equities, is the strategic reason Medici is building here first: it is one of the first environments where tokenized securities and programmable DeFi infrastructure are designed to coexist natively, rather than living in separate, incompatible systems.

USDG, the settlement asset for Medici's repo markets, is a USD-backed stablecoin issued by Paxos and part of the Global Dollar Network, a group of financial and crypto institutions that includes Robinhood^[14]. USDG is intended to be redeemable 1:1 for US dollars, backed by cash and short-term US Treasury holdings in segregated accounts^[15].

Stock Tokens, as issued on Robinhood Chain, are tokenized debt securities issued by Robinhood Assets (Jersey) Limited, structured as standard ERC-20 tokens that can be held, transferred and composed into onchain applications^[11]. They provide economic exposure to the price of an underlying security but do not grant the holder any legal or beneficial rights in, or against the issuer of, that underlying security^[13]. There are no voting rights and no direct ownership claim on the referenced shares^[13].

This document uses the term "Stock Token" deliberately rather than the looser phrase "tokenized stock," and treats collateral built on Stock Tokens as collateral against a debt instrument tracking economic performance, not as collateral against equity ownership. Medici's risk framework, haircuts and Charter apply to what the instrument actually is, not to a simplified label for it.

Everything from here to the end of this section is vision, not roadmap with dates. It describes the shape of the stack Medici is designed to support if the repo market succeeds — not commitments to ship any of it on a particular timeline.

If the Medici Funding Curve becomes deep and trusted enough, it could theoretically support rate swaps, exchanging fixed funding for floating, forward funding agreements locked ahead of need, and instruments that hedge exposure to the curve itself moving. None of this exists today; it is a natural extension of having a real curve to reference, not a live product.

Splitting a fixed-term funding position into a principal token and a yield token is a well understood pattern elsewhere in DeFi, and Medici's explicit maturities make the underlying instrument a natural fit for it. This paper does not claim a live integration with any specific existing protocol that does this today; it notes the fit as a future direction.

Article V of the Charter is deliberate: liquid, transparently priced collateral comes first. Extending financing to illiquid real-world credit would require solving valuation, legal enforceability, liquidity, underwriting and oracle design for each new asset class before it qualifies — not relaxing the standard to accommodate the asset.

Tokenized gold offers a deep underlying market, close to continuous global price discovery, and store-of-value characteristics distinct from equity exposure — a genuinely different collateral profile from stocks or ETFs. It also introduces its own risks: custody of the physical metal, issuer solvency, and redemption mechanics, all of which a haircut and underwriting process would need to price explicitly rather than assume away.

Consider an 18%-haircut SPY market. A -12% gap leaves a positive buffer; the position remains covered. A -20% gap exceeds the buffer, and the position becomes eligible for liquidation. These are illustrative scenarios, not a statement about the historical probability of either move — see the interactive stress test in §08.

Whether a backstop applies to any given shortfall is governed by the Charter's defined conditions, not by discretion at the moment of loss. A backstop narrows a loss. It does not promise to erase one.

At a flat 10 basis points on annual repo volume, gross protocol revenue scales as follows:

A useful way to reason about Medici's trajectory is along a small set of structural variables: the number of active markets, the number of active tenors, repo volume, protocol fee revenue, and reserve size. A bear case might see one or two markets never gain enough depth to matter. A base case sees the roadmap in §10 of the homepage play out at a measured pace. A bull case sees enough collateral classes and tenors open that the funding curve becomes something other protocols actually reference. This paper does not attach dollar figures or probabilities to any of these paths.

Founder and team allocation, vesting schedule, ecosystem and treasury allocation, liquidity provisioning and governance token mechanics are TBD. This document will not invent allocation percentages to fill the section. What is fixed today is the reserve-allocation philosophy in §22–23; token-supply mechanics beyond the transaction tax are not yet finalized.

Future governance is expected to have a role in market listings, haircut bands, caps, reserve policy, fees and treasury allocation. Critical risk management, the parameters that keep a market solvent under stress, is not a place for symbolic decentralization ahead of the tooling, process and accountability that would make it safe. Governance is expected to widen progressively as the protocol and its risk infrastructure mature, not on day one by default.

Liquidity, curve history, integrations, risk models, protocol reserves, brand and distribution are the layers this design is meant to compound across. Today, none of them are deep — this is a new market. What exists today is the architecture built to accumulate them: isolated markets that can be added one at a time, a curve designed to be published rather than kept private, and a reserve mechanism designed to grow with usage rather than be spent down by it.

For years, DeFi had programmable lending but mostly crypto-native collateral. Traditional securities had deep financing infrastructure but closed, permissioned rails. Tokenized real-world assets are starting to collapse the boundary between those two worlds. The opportunity is not to launch another token against the same crypto-native collateral everyone else already uses — it is to bring mature financing primitives to the assets that are only now becoming programmable.

The question this paper leaves the reader with is not whether every legacy market disappears — it will not, and this paper has not argued that it should. The question is narrower and, we think, more interesting: what financial infrastructure becomes possible once the asset itself can execute code?

This document uses expandable technical notes inline (§08, §13, §17, §18, §28) rather than repeating them here — the print version renders every note already open. Look for the TECHNICAL NOTE label in each numbered section for pedagogical simplifications, and the MARKET NOTE label for securities-finance nuance that a market-structure professional would otherwise ask about.

Smart contract · Oracle · Market · Liquidity · Collateral / issuer · Stablecoin · Chain · Governance · Regulatory / legal · Corporate action · Operational · Economic design. See §35 for the description of each. Nothing in this document should be read as minimizing any of them in service of the thesis elsewhere in the paper — see §33 for what Medici explicitly does not claim to be.