Who Actually Verifies the Synthetic Dollars

Every synthetic dollar tells you the same thing: it is fully backed, one to one, nothing to worry about. The claim is almost never the problem. Stream was fully backed until the off-chain manager moved the money. Resolv was fully backed until one key minted eighty million dollars of it from nothing. The figure on the dashboard said par right up to the hour it did not. So the question that actually sorts the field is not whether a protocol claims full backing. It is how much of that claim you can check for yourself, on-chain, in real time, before you deposit and at any moment after.

This is a neutral read of the live synthetic-dollar field on exactly that axis. It names each protocol's real transparency method, because the differences are real and worth getting right, and most of them are legitimate even when they are not the strongest. It also says, in the same breath, the one place Kerne is no better than anyone else. The companion to this essay is a side-by-side reserve transparency scorecard with every protocol's cell sourced and dated.

The question that separates them

Here is the test, applied identically to every name: can a regular holder verify the reserves and the hedge themselves, on-chain, in real time, right now, without trusting a third party? Not can a fund. Not can an auditor. Can you, with a block explorer and a browser, on a Tuesday afternoon.

One thing has to be said up front, because the rest of this is meaningless without it. Verifiable is not the same as safe. A protocol you can check is not therefore solvent, and a protocol you cannot check is not therefore hiding anything. A signed proof tells you the figures are authentic and fresh; it does not tell you the backing is sufficient. Plenty of the names below are large, audited, and run by serious people, and their reserves are almost certainly there. The axis here is narrower and more durable than safety: it is who, exactly, is the party that looks at the reserves. You, or someone you have to trust.

Three ways to be shown reserves

Across the field, there turn out to be three honest answers, and they form a gradient.

1. Trust an attestor. The backing sits in off-chain custody and on exchanges. A third party reads those accounts and publishes a proof, often a zero-knowledge attestation or a signed audit report. You read the result and trust that the attestor saw the truth. This is the institutional default, and it is a real, well-run model.
2. Read part of it on-chain. Some of the backing is an on-chain token or wallet you can read yourself on a block explorer. The rest, an off-chain custodian, a fund administrator's par value, a weekly NAV, still rests on a layer you take on trust.
3. Read it and re-derive it. The reserve is raw-readable on-chain, and the proof of it is signed from public inputs, so you can recompute it yourself with no attestor in the path. This is the strongest answer on the reserve leg, and it is rare.

None of these is wrong. The point of naming them is precision about where verification ends and trust begins, which is the only thing a careful allocator is actually buying.

Trust an attestor: Falcon, Axis, Noon, Neutrl, Midas

The largest and most institutional designs cluster here, and they share an off-chain stack: custody at firms like Fireblocks, Ceffu, Copper and Fordefi, hedges on exchange perpetuals, and a reputable verifier sitting between you and the accounts.

Falcon Finance (USDf) runs a daily transparency dashboard with a per-custodian breakdown across Ceffu, Fireblocks and on-chain, backed by weekly reserve attestations from ht.digital, the digital-asset arm of the UK audit firm Harris & Trotter, and quarterly ISAE 3000 assurance reports, the first published in October 2025 over roughly 1.96 billion dollars of reserves at about 104 percent. That is a serious, above-average disclosure framework. What it is not is a set of enumerated wallet addresses or a proof you re-derive: the numbers are custodian-sourced aggregates you read and trust. USDf briefly depegged to about 0.94 in July 2025 on collateral-quality concerns and recovered the same day.

Axis (USDx) keeps reserves and the hedge in CeFi custody mirrored off-exchange via Copper and Ceffu, and surfaces them through Accountable's Data Verification Network, which ingests custodian and exchange data into a tamper-proof environment and produces a zero-knowledge proof, then pushes an attested value on-chain via Chainlink Proof of Reserves. You read that attested value. You cannot recompute it from the underlying accounts, because the accounts are not public. It is audited by Zellic and was still early-stage as of mid-2026.

Noon Capital (USN) uses the same Accountable model: a verification engine deployed inside Noon's own perimeter ingests data from its custodians and exchanges and publishes real-time reserve, liability and trading-exposure attestations, with some collateral wallets also visible on Etherscan. USN is backed one to one by USDC, USDT and short-term Treasury bills in custodial wallets, plus delta-neutral strategies.

Neutrl (NUSD) is the newest of this group and uses Accountable too, in one of its more on-chain forms. NUSD settles mints and redemptions on-chain through a Chainlink oracle, while Accountable's zero-knowledge Proof of Solvency reads balances from Neutrl's custodians, exchanges and contracts and publishes collateral and delta-neutrality to a public dashboard, with Hypernative holding a monitor and pauser role. The backing is the unusual part: NUSD is funded by buying locked or vesting altcoins at an OTC discount and shorting an equal amount on perpetuals, so the load-bearing assets are off-chain and illiquid, and the authoritative solvency figure is still the Accountable attestation, not a number you re-derive. Neutrl is also the field's freshest reminder that verifiable and safe are different questions: in February 2026 an attacker used a flash loan to exploit an access-control flaw in the Pendle Standardized-Yield integration for NUSD, a permissionless internal-balance burn in the redeem path, and drained about $45,900 from NUSD and USDC liquidity providers. It was a smart-contract bug in an integration rather than a shortfall in the OTC backing, and a separate March 2026 DNS hijack hit the website without touching the contracts, but both are the kind of risk an attestation of reserves does not speak to.

Midas (mTBILL, mBASIS and the mToken suite) is a tokenized real-world-asset issuer rather than a pegged stablecoin, but it belongs here because its on-chain numbers are the output of an attestation pipeline. The Midas Attestation Engine converts fund-administrator reports into structured claims published on-chain through the Chainlink Runtime Environment and LlamaRisk's SAVE framework, with Canary and vLayer as independent verifiers; mTBILL's price is propagated on-chain by Ankura Trust under fiduciary duty. What you read on-chain is a verified figure produced from data you cannot see directly. It is a credible, regulated, multi-verifier design.

The honest summary for this group: the assurance is real and it scales, and the one structural property is that the verifier, not you, is the party that looks at the reserves.

Read part of it on-chain: Ethena, Usual, Anzen, Avant

A second group lets you read some of the backing yourself, which is a genuine step closer, while still resting on an off-chain layer for the rest.

Ethena (USDe) is the largest and most proven synthetic dollar, audited and live since 2024, and nothing here is a knock on it. Its transparency dashboard links some collateral wallets you can read on-chain, and combines them with weekly third-party Proof of Reserves and monthly signed custodian attestations across Copper, Ceffu, Anchorage and Kraken, with Coinbase named as primary custodian and distribution partner as of June 2026. Notably, Chaos Labs' Edge oracle verifies weekly not just that backing covers supply but that the position is delta-neutral, which is more than most do for the hedge leg. The honest limit is that most of the backing sits in off-exchange custody and on exchanges, and the entire short-hedge leg is confirmed by attestation rather than data you re-derive. That is the normal structure for a CeFi-custodied design at this size, not a defect.

Usual (USD0) holds tokenized Treasury and money-market tokens, USYC, M, USTBL, inside its contracts, so you can read the collateral balances against supply on Etherscan, with a Chainlink Proof of Reserve overlay on top. The catch is one layer down: that each of those RWA tokens is itself worth par rests on its own off-chain fund administrator's attestation, from Cohen & Co or PwC. Usual is also the cautionary tale on a different axis. In January 2025 it unilaterally changed USD0++'s redemption from a hardcoded one dollar to an 0.87 floor rising back over years, and the token broke its peg toward 0.90 and cascaded into leveraged loops, even though the underlying collateral was never impaired. It was a governance and redemption-terms repricing, not a reserve shortfall, and the distinction matters.

Anzen (USDz) mints one to one against an on-chain token called SPCT that you can read, but the load-bearing value is an off-chain private-credit portfolio underwritten via the broker-dealer Percent and checked against published eligibility criteria by an unnamed third-party agent, with solvency shown on Anzen's own dashboard. The private-credit notes are opaque and illiquid, mint and redeem are permissioned to qualified minters, and USDz has a real depeg record, trading near 0.97 in late June 2026 with an all-time low of 0.82 in March 2025.

Avant (avUSD) publishes its Reserve Fund and strategy wallet addresses across several chains, including the on-chain perp venues Lighter and Paradex, so a lot of the book is directly readable. The authoritative solvency figure, though, is a portfolio NAV struck weekly by an independent valuation provider, Pennyworks, and the off-chain hedge venues run by its external manager are not enumerated. So intra-week, the true backing and the full hedge are a trusted figure, not a live proof.

Read it and re-derive it: Kerne

Kerne is built the other way around on the leg where it can be. kUSD mints one to one from USDC through an on-chain Peg Stability Module on Base, so the reserve is something you read off the chain directly, no dashboard in between. On top of that, an hourly Proof of Reserves is signed with a key, and the inputs are public, so anyone can recover the signer, rehash the canonical payload, and check the timestamp for freshness. There is a free tool at kerne.fi/verify that does exactly this in your browser, with no backend and no keys, pre-loaded with Kerne's own live attestation as the worked example.

This is the rare third answer on the reserve leg: not a number an attestor hands you, but a result you recompute. It is worth being precise about what that proves and what it does not. A passing verification proves the figures are authentic and fresh, that a named key signed this exact payload at this time. It does not, by itself, prove solvency, and crucially it does not cover the off-chain hedge. Which brings us to the column Kerne loses.

The hedge leg, where everyone converges

Every delta-neutral synthetic dollar on this page holds an asset and shorts an equal amount of it on a perpetual-futures venue. That short leg is what makes the dollar delta-neutral, and it is the part almost no one lets you verify yourself. Ethena's shorts sit on Binance, Bybit, OKX, Deribit and Hyperliquid; you get Chaos Labs' attested confirmation of delta-neutrality, not the position records. Axis, Falcon, Noon and Avant's hedges live on exchange or off-chain accounts covered by an attestation or a weekly NAV.

And so does Kerne's. The Kerne hedge runs on Hyperliquid, and the hedge equity in the signed Proof of Reserves is reported by Kerne and bound to a signature, not independently attested. The signature proves who asserted the figure and when. It does not prove the position exists at the claimed size. An independent attestation of that leg is being scoped and is not yet live. On the hedge, Kerne is in exactly the same position as the rest of the field, and its hedge sits on a single venue, which is more concentration than the larger names carry. That is why, on the honest version of this scorecard, Kerne wins one column and ties the rest on the hardest one.

What Kerne still owes

A piece that graded everyone else and exempted its author would be the exact thing it warns against. So, plainly: Kerne has not completed an external audit, an independent engagement is being finalized and an internal adversarial review is public at /security/findings-tracker. It is in its Genesis stage and intentionally small, so the published APY is a live model, not a large realized distribution. The yield is built from staking plus perpetual funding, so a sustained negative-funding regime erodes it, absorbed at this scale by an insurance fund that is near empty today. And the module values USDC at one to one by design, so on its own it does not independently price a USDC depeg.

None of this is hidden, and that is the point. Kerne earns exactly one column on the scorecard, the reserve leg, by letting you check rather than asking you to trust. Everywhere it cannot yet do that, it says so on the same page.

Which Kerne is this

Because the name collides: this is Kerne, kerne.fi, the Base delta-neutral synthetic dollar whose kUSD is at 0x5C2EfdF0D8D286959b42308966bc2B97f5680AA3 on Base, chain 8453. It is not KernelDAO, formerly Kelp DAO, the BNB Chain restaking ecosystem with its own KUSD, and not Kern, and not Kernel Protocol on Karak, which issues a different kUSD on Ethereum mainnet. The tickers overlap; the teams, chains and contracts do not. See /not-kerneldao and /kernel-vs-kerne to tell them apart.

Conclusion

The institutional synthetic dollars converged on a good answer: hold the backing with reputable custodians and pay a reputable verifier to prove it. It works, it scales, and for a very large audience it is the right model. The one thing it cannot give you is the ability to check the reserves without trusting the verifier, because the verifier, by design, is the only party that sees the accounts.

That answer is getting better, too. A Chainlink Proof of Reserves feed is becoming an institutional default, and protocols that adopt it are more transparent than those that do not. It is worth being exact about what it changes, though. A Proof of Reserves feed is still a value an oracle network publishes after reading custodian and exchange balances you cannot see. It raises the floor of trust; it does not let you recompute the reserves yourself. That is why, on the scorecard, every protocol that leans on an attestor answers no to one specific column, can you recompute the reserves with no third party in the path, and only Kerne answers yes. The line is methodology, not scale or sincerity. It is the difference between reserves you read and reserves you re-derive, and as more of the field adopts attested Proof of Reserves, it is the line still worth watching.

Kerne's bet is the other one, on the leg where it is possible: put the reserve on-chain, sign the proof from public inputs, and hand you the tool to recompute it. It does not pretend that fixes the hedge, and it does not pretend it removes all trust. It moves the trust boundary as far toward you as the architecture allows, and then it shows you exactly where the boundary still is. If that is the kind of dollar you want, the receipts are at the scorecard, the live proof is at /api/por/signed, and the tool to check it is at /verify. Do not trust this essay. Check it.