Identity and sequence
The epoch the report speaks for, the asset or policy key it was filed under, and the sequence it occupies on chain — the same three values the registry indexes its events by.
Products · Dossier
The proof archive
Everything Touchstone has ever concluded is here, in the form it was concluded in: a signed report per epoch per key, the result of every control that was evaluated — including the ones the evidence could not answer — and a bundle that lets a reader recompute the whole verdict without trusting this site. The archive is the audit trail a risk committee asks for, and it grows by one entry per key per day.
25
reports published on chain
20
reached CONFIRMED
23
downloadable verification bundles
2
chains carrying the same record
Current state
Where the record stands
2 assets and two USTB policies are published to both chains. State of record 2026-08-21; USTB reports expire at end of day UTC and the next daily window republishes, so this table is a snapshot of a record that keeps moving.
| Key | X Layer mainnet · 196 | X Layer testnet · 1952 |
|---|---|---|
| USTBSuperstate Short Duration U.S. Government Securities Fund | sequence 5 · CONFIRMED | sequence 5 · CONFIRMED |
| FOBXXFranklin OnChain U.S. Government Money Fund | sequence 1 · CONFIRMED | sequence 1 · CONFIRMED |
| disclosure-freshness:1is the disclosure current enough to act on? | sequence 3 · CONFIRMED | sequence 2 · CONFIRMED |
| nav-settlement:1has the NAV actually settled? | sequence 3 · CONFIRMED | sequence 2 · CONFIRMED |
23 of the 25 publications have a downloadable bundle. Two testnet policy bundle files were overwritten by same-named mainnet files during the first two-chain run; their full signed reports remain embedded in the append-only transparency logs, and the naming is being made chain-aware rather than the logs being rewritten.
The arc worth reading
Refused when provisional. Confirmed when earned.
The archive's most useful entry is not a pass. It is the same NAV value appearing twice: refused on the first sighting, confirmed once it had survived the evidence interval unchanged — and a consumer contract's behaviour changing as a result, without any control being weakened to make it happen.
2026-08-18
Refused
The NAV row was present and plausible, but no capture at least 24 hours older confirmed it. The value control abstained and the state held short of CONFIRMED.
2026-08-19
Confirmed
The same NAV, 11.18316100 for the row dated 2026-08-18, settled unchanged across two captures. Both chains reached CONFIRMED.
2026-08-19
Enforced
With the policy verdict CONFIRMED, the guarded consumer bound to that key executed on chain — and the consumer bound to a key with no report still reverted.
The refusal and the confirmation are one mechanism, not two outcomes. A system that could not produce the first has no business being believed about the second.
What each entry carries
Enough to check it without us
State, with its transition
The reported state and how it got there: what the previous state was, what event moved it, and the deadline any pending evidence was measured against.
Every control result
Each control by id and content hash, with its result, the observed value and the row it belongs to. A control the evidence could not answer is present as unevaluable, never dropped to keep the page tidy.
Commitments
A control-set root, an evidence root, the approval-ledger digest, the policy digest and the compilation digests — every one recomputable from the bundle alone.
Validity
The observation time and the expiry. After that expiry a consumer treating the report as current would be wrong, and the report says so in its own signed bytes.
Limitations, inside the signature
The limitation strings travel inside the canonical report bytes. Removing one changes the signature, which is the only way to make a caveat as durable as the claim it qualifies.
Corrections
Named, never silent
Two corrections sit in this archive, and they are there because signed bytes cannot be edited. A correction is a new report at a new sequence that names the sequence it restates; the original stays on chain, both remain verifiable, and the registry records which supersedes which. Each correction reproduces its original's control-set root, evidence root and approval-ledger digest exactly — restating a conclusion is allowed, quietly changing what it was judged against is not.
Reading an archive like this takes one habit: the newest entry on a key is not necessarily the newest period, because a correction takes the next sequence while restating an earlier epoch. Every entry names its epoch for exactly that reason.