The Kerne Newsroom
Explore Insights
Many of these are reserve-verification teardowns. Run the same checks yourself with the free Verify Any Stablecoin tool, or see every tool and what a signed read costs.
TransparencyCheck a Basis Trade's Funding Income Against the Venue, Not Against Its Dashboard. Three Public Calls, Run Against Our Own Hedge Account.
Every delta-neutral dollar publishes a funding number, almost all of them publish it from their own dashboard, and almost nobody checks it against the exchange that actually paid it. For any account whose address is public, three unauthenticated calls settle it. This post runs all three against Kerne's own hedge account and shows the working. The published ledger reproduces to the fourth decimal; 212 of 212 hourly funding records carry a rate byte-identical to Hyperliquid's own published series; and the 60-day venue mean the yield model takes as its funding input reproduces at 7.2811 percent against a published 0.0728. It also carries the denominator, which is the part that usually gets left off: 1.007301 dollars of lifetime funding, earned on a time-weighted mean short of 73.02 dollars across 1,906 position-hours, is 6.34 percent annualised. There is a trap in the raw data that makes that figure 6.71 times too high if you take the record count at face value, and the script that avoids it is 40 lines and is in the post.
TransparencyWho Actually Holds the Points? Five Checks You Can Run on Any Points Program, Run Here Against Our Own.
Almost every points program publishes a formula whose denominator is "total points". Almost none publishes who holds them. That single omission decides what a farmer's share is actually worth, and it is checkable in about five minutes on any program with a public leaderboard. This is the method, and the worked example is Kerne's own program, where one wallet belonging to the founder holds 97.65 percent of every fragment ever issued. It also covers the problem that appears the moment a small program excludes its own insiders honestly: the eligible denominator collapses, a fixed pool saturates, and the program starts telling allocators that the rational deposit is five hundred dollars. That happened to us. What we did about it, and the one property that makes the fix real rather than cosmetic, is the second half.
MarketHalf of Morpho's Liquidated Markets Have Exactly One Liquidator. The Exposure Figure Attached to That Fact Is Wrong by a Factor of Thirteen.
Across every Morpho Blue market on every chain, 23,171 liquidations in the last 90 days touched 461 markets. In 247 of them, 53.6 percent, exactly one address ever did the liquidating. That is a real and checkable fact about somebody else's protocol, and we are publishing the dataset so you can rebuild it. The exposure figure most people will attach to it does not survive contact with the data: a sum that skips the exclusion rule returns thirteen times too much, because over ninety percent of it is one dead Arbitrum market whose collateral has no price, which Morpho itself flags with $3.28 billion of unrealized bad debt, and whose borrow rate reads 297,996 percent because it has been pinned at full utilization since May 2025. The defensible figure is $328 million. And the obvious hypothesis, that thin coverage predicts bad debt, is refuted for the second time by this data.
TransparencyThree Synthetic Dollars Pay More Than They Advertise. Four Others Have Not Paid Anything in Months. The Worst Row on the Board Is Our Own.
We read fifteen synthetic dollars once an hour: what each advertises on its own surface, and what its vault actually paid over the trailing thirty days, computed from the share price on chain. Read at 2026-08-02 02:37 UTC, three of them realize at or above their own published figure. Ethena advertises 3.92 percent and realized 3.97, Sky advertises 3.52 and realized 3.57, Neutrl advertises 4.20 and realized 4.21. Four others have stopped paying entirely, with share prices frozen since March, November and October, and in every one of those cases the protocol either announced it or we label the row dormant rather than assert a sunset nobody declared. Four more rows we refuse to score at all, including two that would have flattered the protocol, because they publish on a seven day basis and our realized window is thirty. Our own row is the worst on the board at plus three points. Every row carries a signed permalink and a free embeddable badge that prints its own read time.
Withdrawn 1 August 2026. We Publish the Staked Balance at Which Our Own Headline Rate Starts to Fall. Today It Is About Fifty-Four Thousand Dollars, and a Million Would Take 10.5 Percent Down to Roughly 3.4.
Withdrawn 1 August 2026. The 10.5 percent headline this post analyses is no longer published: its subsidy component had never been paid to a holder, and no mechanism existed by which it could be, so the number was retracted rather than corrected. What Kerne publishes now is organic carry alone, recomputed from live funding on every read at kerne.fi/api/apy, and every rate quoted inside this post is a dated record of 31 July 2026 rather than a current one. The post is left standing, with its addendum at the top, because the sizing arithmetic is unchanged and still correct, and because a protocol that ranks other people on the distance between what they advertise and what they pay does not get to delete the day it failed its own test.
SecurityOur First External Audit, in Full: Every Finding Hexens Raised, and What We Did About Each One
The final Hexens report on Kerne is public. Rather than post a badge and a link, we are publishing the response matrix: every finding, its severity, what we changed, and the two we deliberately did not rush a code change for, with the reasoning in full. That includes the finding that turned out to be a regression we introduced ourselves in an earlier hardening pass, and the part where our own first fix attempt was wrong and got caught in re-review before it shipped. The review covered five contracts. Every finding landed in one of them, the vault, and the stablecoin, the staked token, the peg stability module, and the escrow token drew none between them. Here is the whole thing, including what an audit does not tell you about us.
MarketWhat a Fed Rate Move Actually Does to Delta-Neutral Funding Yields. We Checked 17 FOMC Decisions Against 22,000 Hours of Perpetual Funding.
The FOMC met July 28 and 29 and every yield-bearing dollar in this market has something to say about it. Most of it will be a version of "the Fed just moved and that is good for our yield," which sounds mechanical and is mostly decoration. We run a synthetic dollar whose published yield is a live function of perpetual funding, so we went and measured it. Three findings worth your time. First, the interest-rate term sitting inside the funding formula on Binance and Hyperliquid is a constant each venue sets by hand, 0.01 percent per eight hours for the crypto majors, with no link to any policy rate; the Fed moved its target range 175 basis points between June 2024 and today and that number did not move at all, and for 51.5 percent of the hours we sampled it was the entire funding rate. Second, across 17 FOMC decisions the average change in ETH funding from the week before to the week after was minus 1.81 points with a standard deviation of 12.42, and cut meetings were statistically indistinguishable from hold meetings. Third, and least convenient for us, funding right now is compressed rather than elevated: half of Hyperliquid's last seven days of settled hours printed that hand-set constant exactly and the other half printed below it, with not one hour above, so the constant is now acting as a ceiling rather than a floor. Binance has been oscillating either side of nothing, and Ethena's own published figures have sUSDe at 4.00 percent against 10.86 percent since inception. The rate decision does move something. It is the benchmark you compare a carry yield against, not the carry itself.
The GENIUS Act Rules Just Missed Their Own Deadline. The Federal Floor Is Not Under You Yet. You Can Still Verify Anyway.
Congress gave the federal regulators exactly one year to write the rules that put a floor under stablecoin reserve transparency: by July 18, 2026. That date has now passed, and the final implementing rules are not here. This is not a scandal, and we do not present it as one; rulemaking of this breadth slips, the statute carries no penalty for lateness, and several comment periods were still open on deadline day. But it has one practical consequence every holder should sit with: the monthly examined, executive-certified reserve disclosure the law promises is not in force, and on the current mechanics it may not bind until January 2027. That leaves a roughly $300 billion market running on voluntary disclosure, an honor system with a legal end date. The floor you may think you are standing on is not under you yet. Here is what actually happened, what the miss changes, and the check you can run today that does not wait for a rulemaking calendar.
SecurityOne Year Ago, Germany Gave EU Holders of USDe 42 Days to Claim, Then Moved Their Claim Offshore. Here Is What the Exit Looked Like, and How to Check Any Synthetic Dollar's Exit Path Before You Need It.
A year ago today, a 42-day BaFin redemption window closed and Ethena's German entity was wound down; from the next day, EU holders of USDe could redeem only against Ethena's BVI entity, for USDC. It was orderly, the peg held, and holders got their money. But the exit was on the issuer's and the regulator's terms, on a clock, and it moved the counterparty out from under holders. That is the useful thing to study, because every synthetic dollar has an exit path, and almost no one checks theirs until they are forced to. Here is what that wind-down actually was, and the three questions that tell you whether a dollar's exit is a right you hold or a permission you are granted.
MarketThe Whole Delta-Neutral Category Is Shrinking, and Almost None of It Is Depegging. The Stress Is Yield Compression, and It Applies to Us Too.
The delta-neutral synthetic-dollar category is contracting and almost none of it is a depeg. Summed across all forty protocols DefiLlama classifies as basis traders, the category holds about $6.94B against about $7.65B thirty days ago, a fall of roughly 9.4 percent or about $716M, while three of the four largest dollars sit within a tenth of a percent of a dollar. Supply is leaving at par, which is a demand story, not a backing story. The stress is yield compression: Falcon's own published rate fell from about 6.35 to about 4.81 percent in a week, Tori's from 12.3 to 11.26, and two vaults we measure have not moved in eight and nine months. Two dollars grew anyway. The honest disclosure is that the compression applies to our own funding-dependent yield too, and our own row on the Honesty Index still carries the widest gap on the board. Every figure is recomputable from public data.
What the Collapse and Recovery of a $600M Synthetic Dollar Teaches About Redeemability. Resolv Broke on a Single Off-Chain Key, and the Fix Is a Reserve You Recompute, Not an Attestation You Read.
Resolv circulated close to six hundred million dollars of USR at its peak. Today the token is a rounding error of that: circulating USR down more than ninety-nine percent, market cap under a million dollars, price near fourteen cents. The break, in March 2026, was a single off-chain key that gated an unbounded mint, roughly eighty million USR conjured from about a hundred thousand dollars of deposits, and no quarterly attestation could have caught it, because a report has a period and a mint does not wait for it. Resolv did not simply vanish. Pre-incident USR holders are being offered redemption at one USDC per token through late August, the junior insurance tranche recovers less, and the team has pivoted to a tokenized real-world-asset line. But the token itself did not survive, and the lesson generalizes: a redemption you can only collect months later, at the discretion of the issuer, is not the same as a reserve you can recompute in the minute that matters. Here is the failure class named precisely, what the recovery does and does not restore, and where our own countermeasures map to it and where they still fall short.
SecurityA Licensed Payments Processor Was Drained For Thirty Hours While Customer Deposits Kept Arriving. We Recomputed Triple-A's Float and Found the Reported Loss Was Already Stale.
On July 24 an attacker moved stablecoins out of wallets identified with the crypto payments firm Triple-A, swapped them to ether and consolidated everything into a single address. We recomputed the whole thing on chain rather than repeat the headline: that address holds 5,287.63 ETH, about 60 more than the figure in circulation, and it has never sent anything out. The finding that matters is not the size. It is that the deposit rail stayed open, with 272 further customer deposits worth 579,528 arriving in the twenty-nine hours after the first theft transfer, interleaved with 505 ordinary merchant payouts. A custodial balance is a database row, and no holder of one could have checked any of this.
A Dollar Stablecoin Was Minted Twenty-Four Times Over in Twelve Hours. Balance Coin's MakerDAO Fork Copied the Liquidation Engine and Deleted the One-Hour Oracle Delay Built to Stop Exactly This.
On July 21 and 22, an attacker fed 42DAO's price oracle a value of one quadrillion dollars per bitcoin, and because the Maker fork behind Balance Coin carried no delay, no deviation check and no price ceiling, minted millions of unbacked BLC against a small BTCB position in a single transaction. We recomputed the whole incident on-chain before writing: supply ran from 404,760 to 9,910,949 BLC, a twenty-four-fold inflation, and the peg fell to a fraction of a cent. The failure was not a hidden bug. It was a fork that inherited Maker's machine and removed the one part that exists to make this attack impossible, a removal invisible to anyone who assumed it still behaved like the original.
An $825 Million Pre-Deposit Vault Filled in About 22 Minutes, and On-Chain Timestamps Show Most of It Landed Before the Public Was Told. Four Things to Check Before You Deposit Into Any Window.
The pre-deposit vault is the sector's favorite way to launch: park stablecoins now, earn points, convert at token generation. Its most-copied version, Stable's $825M Phase 1 in October 2025, hit its hard cap in about 22 minutes, and on-chain analysis reported by The Block and The Defiant found that roughly $500M, close to 60 percent of it, had hit the contract before the campaign was announced, across about ten wallets, with fewer than 300 wallets taking part in total and no proof of reserves published while the window was open. The same pattern recurs across the category, and a few concrete properties tell a window worth trusting from one that is not. Here is the pattern, and where our own capped Genesis window lands against it, worst facts included.
A $114 Million Bitcoin Outflow That Never Happened. Three Nine-Figure TVL Crashes in Eight Days Were Dashboard Artifacts, and the Same Five Minute On-Chain Read Catches All Three.
This morning DefiLlama showed Bedrock's uniBTC losing 37 percent of its TVL in a single day, about $114 million. The chain shows the token's supply moved 0.014 percent over the entire week and the token trades at par with Bitcoin. It is the third nine-figure dashboard crash in eight days, after BlackRock's BUIDL printed a phantom 22 percent supply drop on July 14 and Lombard's LBTC printed a phantom 57 percent TVL drop on July 15, and none of the three was an event on chain. A TVL dashboard is an aggregation pipeline, not a proof of reserves. Here is what each reading actually was, and the five minute recompute that separates a phantom from a run before you trade, allocate, or panic on one.
Our Reserve-Disclosure Position Is Now Posted on a Federal Docket. FinCEN Published Our Comment, the FDIC Has One by Email, and the OCC Window Closed Without Us.
The GENIUS Act rules missed their July 18 deadline, but the rulemaking kept moving, and we set out to put a position on the record where the questions were being asked. Three of the four windows we worked have resolved. Our comment on the joint FinCEN customer-identification proposal (RIN 1506-AB74) was filed on August 9, 2026 and is now publicly posted on regulations.gov as comment FINCEN-2026-0101-0009, under the name Kerne Protocol, which makes it a permanent government-hosted record we cannot edit or withdraw. Our FDIC comment went in by email on July 23, 2026. We did not file with the OCC: the letter was written, the window closed before it was submitted, and the OCC accepts no late comments. This post is the position itself, that a reserve a holder can re-derive from the chain every block is the strongest form of the traceability these proposals ask about, where and only where the reserve is on-chain. It is also a record of the one thing we are careful not to do, which is tell a federal agency how to classify our own token, and of where Canadian law now lands on a dollar token issued by a Canadian.
USD.AI Shed About a Quarter of Its Supply in a Day and Never Left a Dollar. The Chain Shows One 54 Million Dollar Redemption Settled One to One in PayPal USD, Not a Depeg. Here Is How to Verify It.
Between July 20 and 21, 2026, USD.AI (USDai) saw its circulating supply fall from about 202 million dollars to about 154 million, a drop of roughly 24 percent in 24 hours. The price never moved off a dollar. On Arbitrum the cause is a single transaction: one multisig redeemed 54.4 million USDai and received 54.4 million PayPal USD, one to one, straight from the token contract. That is an orderly redemption at par, not a depeg and not an exploit. We read the live supply off Arbitrum ourselves, decode the exact redemption, separate what the chain does and does not prove, and show the free check that let you watch the float leave in real time.
The Feasibility Study Behind Our 8 to 9.4 Percent Through-Cycle Band, Published in Full
Several of Kerne's own surfaces, from the home and stake pages to the dataroom and the /api/apy response, have cited "our published feasibility analysis" as the source of the 8 to 9.4 percent through-cycle ceiling we put on our own yield. The band was real and the work behind it was real, but the document was internal, so the word published was doing work it had not earned. This is the analysis, in full: where the modeled 12.98 percent comes from, where the 9.4 comes from, where the 8 comes from, why the entire distance between them is a leverage assumption the deployed strategy does not currently run, and what would have to be true before the top of the band is earned rather than modeled. Every figure reproduces from a published endpoint or from arithmetic you can redo in a line.
A Disclosure Check on scrvUSD: Projected APY, Realized APY, and the Two Reads That Reconcile Them
This is a short, public demonstration of a review Kerne now sells, run on a token that is not a customer. Kerne's Disclosure Integrity Audit checks whether a yield protocol's public claims line up with what its chain actually shows. To show what that looks like, we pointed the first part of it, advertised yield against realized yield, at Curve savings crvUSD (scrvUSD). Curve displays a projected APY near 1.55 percent from its own pricing API; a holder's realized yield over the trailing 30 days, read straight from the vault's share price, annualized to about 2.52 percent. Neither number is wrong. They measure different things, and a reader deserves to know which one they are looking at. Everything here reads from public data and reproduces in two calls, and most of it confirms what Curve already documents.
Two Oracle Failures This Week Came Down to a Signature Nobody Checked. We Shipped That Bug on Our Own Proof of Reserves in May, and Caught It. Here Is How to Check Whether Anyone Is Checking Yours.
In the last week, Ostium on Arbitrum was drained through an authorized oracle signer that submitted future-dated price reports, and Bonzo on Hedera was drained because its oracle vendor, Supra, accepted a price update whose signature and the committee key it was checked against were both zero, so the check trivially passed. Different chains, different mechanisms, one shared shape: a signature that a system trusted without really verifying. We shipped that exact class of bug in May, on our own hourly signed Proof of Reserves, and we caught it, fixed it, and wrote it up in public. This is a walk through both incidents, the honest limit of what a signature buys you, our own version of the same mistake, and the two-minute check that tells you whether the attestation your money depends on is actually being verified or just displayed.
Bridged USDe on MegaETH Fell About Three Quarters in a Day and Never Left Its Peg. The Loop That Drained MegaUSD Unwound One Level Up the Stack, at Par. Here Is What the Per-Chain Data Shows.
Between July 9 and 10, 2026, the amount of Ethena's USDe bridged onto the MegaETH network fell from about 294 million dollars to about 72 million, a drop of roughly 75 percent in 24 hours, while USDe held its dollar peg the whole time. It is the same event we wrote up in MegaUSD, one level up the stack: the incentive loop that had pulled hundreds of millions of USDe onto MegaETH unwound, and the float bridged back out at par. We read the current figure two ways, off DefiLlama's live per-chain series and straight off the MegaETH contract, show that Ethena's reserve rail did not move while this happened, and walk the free check that let you watch the float leave in real time.
MegaUSD Shed More Than Half Its Supply in a Day and Never Broke Its Peg. That Is a Redemption Run at Par, Not a Depeg. Here Is What the On-Chain Data Shows.
Between July 9 and 10, 2026, MegaUSD (USDM), the native stablecoin of the MegaETH network, saw its circulating supply fall from about 293 million dollars to about 116 million, a drop of roughly 47 percent in 24 hours. The price barely moved, trading within about a fifth of a cent of a dollar the whole time. That combination, supply leaving while the peg holds, is the signature of an orderly redemption or deleverage at par, not a depeg and not an exploit. We read the current supply straight off MegaETH mainnet ourselves, separate what the on-chain data does and does not prove about the cause, and show the free check that let you watch the float leave in real time instead of reading about it days later.
Summer.fi Lost About $6M to a Donation Attack on an ERC-4626 Vault. Reading the Vault Would Not Have Caught It. Watching Its Share Price Would.
In the first days of July 2026 an attacker drained about $6M from Summer.fi's Lazy Summer vaults by donating assets into a FleetCommander ERC-4626 vault to inflate its totalAssets, then redeeming shares against the inflated price, with a roughly $65.4M flash loan for scale. Summer.fi's guardians paused the vaults. This is the exact contract class most of DeFi, us included, builds on, so it is worth reading carefully rather than dunking on. Here is how the attack worked, the honest limit of checking a vault yourself (a one-time reading cannot catch an in-transaction donation attack, only watching price-per-share over time can), the free tool we built to read any vault's accounting, and where our own vault sits, disclosed rather than hidden.
The GENIUS Act Rule Deadline Is July 18. What Is Actually Due, What the Yield Ban Reaches, and What a Miss Would Mean.
July 18, 2026 is the statutory deadline for the federal regulators to finish the GENIUS Act implementing rules, one year after the law passed, and the run-up is the fortnight the whole question is loudest. As this is written no final rule exists, the Federal Reserve has not even proposed, and the deadline itself carries no penalty for a miss. Here is the plain-language run-up: what is actually due and who owes it, what the Section 4(a)(11) yield ban reaches and the synthetic dollars it does not, what an issuer needs in hand before the rules land, and why a slipped deadline changes less than it sounds like. The rule-day analysis will follow the day the rules actually arrive.
BaFin Barred USDe in Germany. BlackRock Integrated It Into Aladdin. Neither Changed What a Holder Can Verify.
Within about a year, a European regulator ordered Ethena to wind down its USDe business in Germany, and the largest asset manager in the world integrated the same synthetic dollar into its institutional risk platform. It looks like a contradiction, one side calling USDe unsafe and the other calling it safe. It is neither. The two are answering different questions, and the one a holder actually needs is left untouched: can you recompute the backing yourself, right now, without trusting a report you cannot check. Here is what each side did, why both can be right, and the verification gap underneath the whole story.
One EOA Minted $80M of a Stablecoin. Could One of Ours? We Audited Our Own Keys, in Public.
The most expensive fact of the last synthetic-dollar cycle is a small one: on March 22, one compromised off-chain key minted about $80M of unbacked USR from roughly $100k of collateral. Not a contract bug, a key. So the honest question for every issuer, us included, is not "are you audited" but "how many keys can do that, and what can each one touch?" We ran that audit on ourselves and published the table. The Resolv vector, one key gating an unbounded mint, is structurally closed on kUSD: the mint admin is a 2-of-3 Safe and the mint amount is derived from the contract's own USDC balance, not a signed number. We still hold two single-signer keys, the reserve-attestation signer and the strategist role, and here is exactly what each can and cannot do, and how we route them to the Safe.
apxUSD and STRC: Why a Preferred-Share-Backed Dollar Tracks Its Collateral Below Par
apxUSD trades near $0.75 today, about a quarter below a dollar, after the STRC preferred shares that make up most of its collateral slid below par on the Bitcoin drawdown. Read as a depeg it looks alarming. Read as a net-asset-value tracker backed by variable-rate preferred equity, it is doing what it was built to do. Here is the structure, what Pendle PT holders face at maturity, and the one thing a signed reserve read would have shown an allocator in advance.
What the Resolv Exploit Tells Us About Synthetic-Dollar Mint Paths
On March 22, an attacker minted $80M of unbacked USR against about $100K of USDC by compromising a single off-chain key. The architectural class of failure that made it possible is more common than the post-mortems suggest. Here is the mechanism, the three on-chain questions worth asking before depositing into any synthetic dollar, and how Kerne's mint path answers them.
How to Verify Kerne Without Trusting Us
A guided tour of the URLs, line-cited sources, and parity tests that let a stranger check our security claims in five minutes, without us in the middle.
Get the next post in your inbox
New Kerne insights drop every couple of weeks: audit findings, basis-trade analysis, post-mortems of competitor incidents, and the math behind the live APY. One email per post, never the same week.
By subscribing you agree to receive milestone updates from Kerne Protocol. We do not share email addresses. Read the privacy policy.