We recomputed the reserves. Here is what reconciles.

By the numbers.

Every figure below is computed from the same registry that drives the 27 per-token pages this report links, so the report can never quietly disagree with the pages underneath it. Categories describe the structure of the verification, not the quality of any protocol.

The 8, 9 and 10 split partitions the whole field: every dollar lands in exactly one of the three. Of the 8 attestor-optional dollars, most are crypto-collateralized designs whose collateral sits in readable contracts; only Kerne's 2 tokens also publish a proof of the reserve you can re-derive from public inputs, which is a stricter test than readable. 9 of the 27 are delta-neutral designs that disclose an off-chain hedge leg, and not one of them, Kerne included, lets you recompute that leg in real time.

How we recomputed it.

The method is deliberately dull, because dull is reproducible. For each token we did the same four things, and you can redo every one of them yourself.

1. Read the supply from the chain. We called totalSupply(), symbol() and decimals() on each deployment directly from a public RPC endpoint at a block. Every per-token page carries the exact cast and curl commands to reproduce the read, so the one number this report never asks you to trust is the number you can most easily check.
2. Confirmed identity by address, not name. A look-alike token can copy any symbol, so an enrichment is applied only when the live on-chain symbol matches the registry entry. A stale or wrong address fails safe to a generic read rather than mislabeling a token.
3. Located the reserve figure. We read each issuer's own disclosures, dashboards and attestation letters, cross-referenced them against independent trackers, and classified the reserve leg on one structural axis: is a third-party attestor optional, because you can read the backing on-chain, or required, because the accounts are off-chain and you read a figure someone publishes?
4. Dated everything, and marked what we could not reach. Every current-state fact is dated (reviewed July 3, 2026); the structural method is evergreen, the snapshot ages. Where a reserve is off-chain, an RWA par value rests on a fund administrator, or a hedge sits on an exchange, we say so plainly rather than counting it as verified.

This is attestation and objective-data tooling, not an audit, a rating, or a solvency opinion. It recomputes what is public and on-chain and is explicit about the parts it cannot recompute. Run the same read on any token in the free Verify Any Stablecoin tool, or open any row in the field below.

What reconciled.

The good news first, because there is real, checkable transparency in this field and it deserves to be named as precisely as the gaps.

• On-chain supply, every single time. For all 27 dollars, the supply read straight from the chain and matched what a block explorer shows. This is the floor of verifiability, and every token in this report clears it. It is also the least of what backing transparency requires, which is the rest of this page.
• The crypto-collateralized dollars whose backing is fully on-chain. For designs such as LUSD, crvUSD and GHO, the entire backing is collateral held in readable contracts, with no off-chain reserve and no attestor. You can read all of it. The residual risk there is smart-contract and collateral-price risk, not attestor trust, and we say so on each page.
• Where a live read matched the issuer's own disclosure. For Falcon's USDf, our on-chain supply read lined up with the figure on Falcon's dashboard and its weekly HT Digital attestation letter. That is a genuine reconciliation of the on-chain leg. The honest caveat, carried into the next section, is that only about 1.3 percent of USDf's reserves are on-chain by Falcon's own API; the rest is an attestation you read.
• Kerne's own reserve, re-derived, not asserted. kUSD mints one for one from USDC through an on-chain module on Base, so its reserve is readable off the chain directly, and its hourly Proof of Reserves is signed from public inputs you can rehash yourself. That is the one place in this whole field where the reserve figure is reproducible rather than published. It is also, deliberately, only the reserve leg.

Where our recomputation stops.

A findings report that hid its own blind spots would be worthless. So here, plainly, is everything this recomputation does not reach, ours included. These are not failures of the protocols; they are the honest edge of what anyone can verify from public inputs today.

• Off-chain custodied reserves. For the 10 attestor-required dollars, and the off-chain portion of the 9 mixed ones, the reserves sit with custodians and on exchanges. We recompute the supply; the reserve figure reaches us as an attestation we read, from real firms of genuine standing. We did not, and cannot, re-derive those accounts.
• Real-world-asset par values. For tokenized T-bill and private-credit designs, the on-chain token is readable but its par value rests on an off-chain fund administrator's attestation. We confirm the on-chain mirror; we take the underlying par value on the administrator's word, as every holder does.
• Every delta-neutral hedge leg, including ours. All 9 delta-neutral dollars here hedge with a short position that lives off-chain, and that leg reaches a holder as a reported or attested figure, never as data you re-derive. Kerne is not ahead here. Our short runs on Hyperliquid, a single venue, self-reported and bound to a signature rather than independently attested, which is more concentrated than the larger names, not less. An independent attestation of that leg is being scoped and is not yet live.
• Kerne's other limits. Kerne is pre-audit and Genesis scale, deliberately small; its published skUSD yield is a live model, not a large realized distribution; and its module prices USDC at a hard 1:1 with no depeg oracle on that leg. Every one of these is listed at /legible and in the scorecard. We win one leg by construction and tie the field on the other, and both are in this report on purpose.

The full field, 27 dollars.

Every dollar we recomputed, ordered from the ones you can check yourself to the ones you read an attestation for. Open any row for its live on-chain supply, the exact commands to reproduce it, and the sourced breakdown of where its trust boundary sits. A flag marks a token that was not operating normally when we read it.

"Attestor-optional" means the reserve backing is readable on-chain, so a third-party attestor is optional on that leg; "partly attestor-required" means part is readable and part rests on an off-chain layer; "attestor-required" means the reserve sits off-chain and you read an attestation of it. These describe the structure of the verification, never the quality of the issuer. See the same field ranked on the reserve transparency scorecard or placed on the five-model attestation versus recomputation spectrum.