Every verdict we have frozen — hash-chained before its outcome existed and scored in public, no cherry-picking. What trips a call (the raw features) stays committed inside the hash but is not published here, so it cannot be reverse-engineered. back to the scoreboard
| call_id | chain | coin | band | verdict | rule | called (UTC) | 24h | 7d | row_hash |
|---|
Every row is hash-chained: row_hash = SHA-256( prev_row_hash + canonical_json(call) ). The canonical JSON is compact and key-sorted and includes the call’s frozen features; the chain root is the hash of the most recent row, so it commits every row before it; genesis is 64 zeros.
The raw features are not published here — so operators cannot reverse-engineer what trips a call — but they are committed inside each row_hash. What you can check right now, without trusting us: that the chain is unbroken and ends at the published root (nothing inserted, deleted, or reordered):
It pulls every page from the public API and recomputes the chain. To prove a specific call’s contents were not edited, ask us for that call’s features and recompute its hash with the scheme above.