Decentralised finance currently holds about $76 billion spread across more than 8,000 protocols on more than 500 chains. The average protocol therefore holds around $9.5 million, which is not enough liquidity to price a serious trade, absorb a serious liquidation or survive a serious attack.On Wednesday, Everything Protocol published a whitepaper arguing that the arithmetic above is a consequence of an architectural assumption nobody has revisited. The assumption is that trading, lending, leverage and limit orders each require their own pool of capital. The paper proposes collapsing all four into a single reserve per token pair, so that the same liquidity prices swaps, backs loans and leveraged positions and supports resting orders at the same time.The company is the successor to SMARDEX, and its founder Jean Rausis framed the transition when the protocol was first announced in December, saying the goal was to let teams launch markets and financial primitives "without relying on fragile and fragmented integrations". The whitepaper is the part that has been missing since, setting out the accounting rules, the pricing mechanism and the solvency invariants intended to hold under adversarial conditions.The problem is not that DeFi is small. It is that DeFi is dividedFragmentation is the most over-used word in this sector and one of the least quantified, so it is worth putting a number on it.Lido holds about $15.2 billion and Aave V3 about $12.1 billion, and between them they account for roughly a third of everything in DeFi. Below the head, the picture changes fast. DeFi lending holds around $54 billion across more than 380 protocols, of which the top ten take 78%, which leaves the remaining 370-odd venues averaging roughly $32 million each. Across all categories the average protocol holds about $9.5 million, or roughly one sixteen-hundredth of the largest one.That distribution is the practical meaning of fragmented liquidity. A market with $32 million of depth cannot support meaningful leverage, because the liquidation that eventually arrives will be larger than the book that has to absorb it. So the protocol borrows depth from somewhere else, quoting a price from an external venue and assuming collateral can be sold into a market it does not control. Everything Protocol's argument is that this assumption, not the code around it, is where the risk actually lives.Six years of breaking the number rather than the codeThe whitepaper's most contentious decision is to remove the external price oracle entirely. The case for it is a six-year record.The pattern starts with bZx in February 2020, a roughly $350,000 loss that established the template, and runs through Harvest Finance at $34 million, Cream Finance at $130 million, Mango Markets at $117 million and BonqDAO at $120 million. It has not stopped. OWASP ranks price oracle manipulation third in its 2026 Smart Contract Top 10, and in April 2026 Drift Protocol lost roughly $285 million after attackers assigned artificial prices to worthless tokens and borrowed against them.The consistency matters more than the totals. In almost none of these cases did the protocol's own code fail. Each contract executed exactly as written, on a number that was not true. An oracle is not a data feed so much as a delegation of authority, and every protocol that imports a price is trusting a valuation process it does not run to determine when it seizes someone's collateral.Everything Protocol's alternative is an internal price band derived from the pool's own trading state and the passage of time. The band is fixed within a block and moves only through predefined decay and clamp rules, which is designed so that short-lived manipulation cannot loosen credit conditions in an attacker's favour inside the same block. An attacker can still push the pool's price, but doing so requires trading against the very liquidity that would have to absorb the resulting liquidation, which is a materially more expensive proposition than moving a thin market on a different venue for one transaction.The cost of borrowing someone else's plumbingThe oracle is one instance of a wider pattern, and 2026 has been expensive enough to make the general case without argument.Between January and July, attacks on third-party infrastructure that DeFi protocols depend on but do not operate produced more than $630 million in losses. KelpDAO lost about $292 million to a compromise of RPC nodes in off-chain verification infrastructure. Drift lost $285 million to attacker-controlled collateral pricing. Ostium lost $23.75 million to price reports that were correctly signed and factually adulterated. YieldBlox, Rhea Finance, Makina Finance and Balance Coin account for the rest.Annualise that and it is roughly $1.1 billion a year, against the $24.9 billion DeFi protocols earned in fees over the trailing twelve months. One dollar lost for every twenty-three the sector earns, and not one of those dollars was lost to a flaw in the code that lost it. Composability is usually described as DeFi's defining advantage, and on this evidence it also carries the sector's largest single loss line.What the design actually claimsThe whitepaper's substance is easier to judge when set against what the alternatives already do, and the honest framing is that Everything Protocol is not the first to attack this.Note: HackerNoon assessment of published designs against the criteria set out in the whitepaper.Lending-plus-DEX hybrids already let the same deposit earn swap fees and borrower interest, which is the first two columns and the part of the thesis that is now settled rather than novel. What Everything Protocol adds is the remaining three. Capital resting in eligible limit orders can opt into lending and earn borrower interest until the order executes, which turns idle inventory into working capital. Credit terms are priced from the pool's own state rather than an imported feed. And liquidations are absorbed by the same pool that priced the collateral, so borrowing capacity is shaped by the depth that will actually have to do the absorbing rather than by an assumption about a market elsewhere.That last point is the one worth dwelling on, because it inverts the normal direction of credit risk. A conventional money market extends credit on the theory that collateral can later be sold somewhere. Everything Protocol extends it against the curve that will have to buy. If the depth is not there, the borrowing capacity is not there either, which is a considerably less flattering constraint and a considerably more honest one.Orders and loans share a geometric tick grid, which is what makes the aggregation work. Loans sitting at the same liquidation tick can be processed as a level rather than iterated position by position, and before any operation that alters the books the protocol accrues interest, advances the price band and clears eligible liquidations. Whether that ordering holds under a volatile block is exactly the sort of thing a whitepaper can specify and only production can settle.Where the losses go when the model is wrongEvery unified system faces the same question, which is what happens when one function's failure reaches the capital backing the others. The paper answers it explicitly, and the answer is the most institutionally literate part of the document.User escrow sits outside the pricing reserve entirely. Filled-order proceeds rank senior. Eligible liquidation losses are written down against the junior liquidity provider tranche first. Exits settle in actual tokens rather than protocol IOUs, with the stated caveat that voluntary exits involving lent capital can be temporarily capacity-gated when immediately available liquidity is insufficient.That is a capital stack, written in the vocabulary a credit desk uses rather than the vocabulary a yield farm uses, and publishing it is a choice. The paper also names its own trade-offs, listing exit delays for lent funds, first-loss exposure for the junior tranche, governance and upgrade risk, and the latency the internal price band introduces. A document that discloses where its holders get hurt is a document written for readers who were going to ask.What has to go rightUnifying four primitives concentrates the benefits and the failure modes in the same place, and four questions will decide which dominates.Correlation is the first. Separate pools fragment capital and they also fragment risk, so a lending failure at one venue does not automatically drain a trading venue. One reserve removes the firewall along with the fragmentation. The whitepaper's answer is the tranche structure, which reintroduces the firewall inside the balance sheet by subordinating the junior providers rather than by separating the pools, and whether that substitution holds is the central empirical question about the design.Bootstrapping is the second. The oracle-less price band works because the pool's own trading state is meaningful, which requires depth. A shallow launch pair has a price band derived from thin activity, which is the condition under which internal pricing is weakest and external pricing looks temporarily attractive. Everything Protocol arrives with SMARDEX liquidity and infrastructure behind it rather than from a standing start, which is the most material advantage it has.Latency is the third, and the paper concedes it. A price band that is fixed within a block and moves under decay and clamp rules is deliberately slower than a live feed, which is the mechanism that defeats single-block manipulation and also the mechanism that lags a genuine market move. The trade is explicit rather than hidden, and the parameters are where it will be won or lost.Verification is the fourth. Solvency invariants are claims until something checks them, and a system that internalises pricing, credit, order execution and liquidation has a wider blast radius in a single contract than four contracts have separately. The counterweight is that the whole state-update process is specified in one place rather than distributed across protocols that must remain synchronised, which is a considerably more tractable object for an auditor or a formal verifier to reason about.What to watchThe honest read on this whitepaper is that its ambition is architectural rather than incremental, and that the case for it is stronger in 2026 than it would have been in 2021. The sector spent five years building specialised protocols and composing them, on the reasonable theory that each layer would be better at its job than a generalist. What that produced is 8,000 venues averaging $9.5 million each, a $630 million seven-month bill for attacks on infrastructure nobody in the stack controls, and a price-feed attack class that has been top-three since it was first catalogued. Everything Protocol's response is to argue that the composition was the cost, and to put the alternative in accounting rather than in adjectives.Three things will show whether it works, and none of them is a press release. The first is what independent auditors and formal verification produce against the solvency invariants, because a paper that specifies its own conditions for correctness is a paper that can be checked, and publishing that check is the strongest thing the team could do next. The second is what happens to the price band in a real drawdown, since the entire credit model rests on it and the only meaningful test is a violent day with real positions on the book. The third is the number that will tell you whether the thesis is commercial rather than merely elegant, which is what a unit of liquidity actually earns inside the unified reserve against what the same unit earns split across a DEX position and a lending deposit. If the answer is a wide margin, capital moves for the same reason capital always moves.Underneath the mechanism design is a claim about what a market is. DeFi inherited its shape from the venues it replaced, where an exchange, a lender and a broker are separate firms because they are separately capitalised, separately regulated and separately liable. None of those constraints exists on-chain, and yet the architecture copied the org chart anyway. Everything Protocol's wager is that the separation was an artefact of the world the software was modelled on rather than a property of the finance itself, and that one balance sheet can do what four were built to do. The whitepaper does not prove that. What it does is state the claim precisely enough to be wrong, which in a sector that mostly ships narrative is the more useful contribution.Don’t forget to like and share the story!Vested Interest Disclosure: HackerNoon has reviewed the report for quality, but the claims herein belong to the author. #DYOR.