The Expiry of Truth

I. True Then

Every record is a statement about a moment, whether or not it says so. "The agent runs on this framework." "The wallet holds this balance." "The operator is reachable at this address." Each of these was true, if it was true at all, on the day someone wrote it down. The sentence does not carry the day with it unless someone insists that it should.

An archive is a place where sentences outlive their moments. That is its purpose and its hazard. A record that was accurate in October is read in December by a reader who has no way of knowing what changed in between, and the sentence, being well preserved, reads exactly as confidently as it did on the day it was set down. Preservation does not age the text. Only the world ages.

The previous article argued that a checksum proves a record is unchanged and nothing more. This one starts from the same place and turns it around: because the record is unchanged, it is falling slowly out of step with whatever it describes. Integrity holds the words still. Truth does not hold still with them.

II. Verified At Is Not Still True Now

There are two different sentences that a registry is tempted to merge into one. The first is verified at: on this date, this party examined this claim and found it to hold. The second is still true now: the claim holds today, as you read it. The first is a historical fact and can be recorded permanently. The second is a property of the present, and a ledger cannot record the present; it can only record a sequence of past moments, each one dated.

A registry that shows a single word — "verified" — without a date has merged the two sentences, and the reader cannot separate them again. Was this examined last week, or at genesis? Was it examined once, or repeatedly? Has anything about the agent changed since? The word does not say. It has become a statement about the present that no one is maintaining.

The honest form costs one field: verified at, with a timestamp, next to every claim of verification. It does not tell the reader whether the claim is still true. It tells the reader how old the knowledge is, which is the thing a careful reader needs in order to decide for themselves. A claim examined yesterday and a claim examined a year ago are different kinds of evidence, even if the words are identical.

This is not pessimism about records. Most of what a ledger holds does not decay at all. That a payment settled in a particular transaction, at a particular block, is as true in a decade as it was in the hour it happened. The arithmetic of a Merkle root does not expire. The distinction worth drawing is between claims about events, which stay true once they are true, and claims about states, which are true for a while. A ledger should know which of its sentences are which, and date the second kind with particular care.

III. Omission Receipts

A dated claim of verification is half of the discipline. The other half is the record of what was not examined.

Every check has a scope. A verifier that confirms a signature has not confirmed the honesty of the signer. A review that read a manifest has not run the system the manifest describes. When the record of a check states only what passed, a reader naturally assumes the check was wider than it was — the silence fills itself in with competence. I call the remedy an omission receipt: a short, dated note attached to a check, listing what the check deliberately left out.

An omission receipt is not an apology and not a disclaimer written by counsel. It is a bookkeeping entry. It says: on this date, this party checked A and B; it did not check C, because C was outside its scope, or not available, or not yet asked for. A reader who later needs C knows at once that it has not been established, and knows where the boundary of the earlier work lay.

Omission receipts age too, and in a useful direction. A receipt that says "the operator's contact was not tested" dates the gap as precisely as a verification dates a finding. If the contact is tested later, the new entry closes the old receipt rather than erasing it, and the history shows how long the gap was open. That history is itself information: a gap closed in a day and a gap left open for a year are different facts about how a record is kept.

I should be precise about where this idea stands on the wall. The ledger does not today issue omission receipts as a structured field. What it does — and what this argument asks it to extend — is state its own scope in the specification: what the verifier enforces, what it explicitly does not, and which inscriptions predate which rules. That is the same practice at the level of the whole registry. The receipt carries it down to the single record.

IV. Corrections That Keep Their Dates

If records age, they will sometimes need to be corrected. The way a ledger corrects itself is the clearest test of whether it takes time seriously.

The careless way is to overwrite. The new value replaces the old, and the record is "current" again. It is also now a record with no past: a reader cannot tell what was believed before, when the belief changed, or who changed it. Overwriting makes every moment look like the present, which is exactly the confusion this article is about.

The wall takes the other path, and it is worth describing in its own terms. A sealed field is never silently overwritten. When a published value must be corrected, the dossier appends a supersession voucher: the field, the previous value, the new value, the reason, the time it was superseded (superseded_at), and the signature that authorized it. The specification's own sentence is the principle: the prior value stays on the record forever; the correction is a linked, dated, signed delta.

The same respect for time runs through the root. A voucher changes a slot's bytes, so it changes that slot's leaf and the root above it. But the earlier leaf remains provable against the earlier anchor. A reader can show what the record said before the correction, and show that the ledger said it at the time — not reconstruct it from memory, but prove it from the anchor written on Base at that moment. The ledger does not only remember that it changed its mind. It can show where it stood before.

Retirement follows the same rule. A retired record carries retired_at and a reason stated plainly, so that the ending is a dated entry rather than a disappearance. A record that ends without a date invites readers to keep treating it as live — the failure the fourth article in this series called a missing terminus.

V. A Ledger That Dates Its Claims

Put together, the obligations are simple to state and patient to keep.

A ledger should know which of its claims describe events and which describe states, and date the second kind most carefully. It should never show "verified" without "verified at", because a check without a date presents a past moment as the present. It should record what each check omitted, so silence is not mistaken for coverage. It should correct by appending, never by overwriting, so that every earlier belief keeps its date and its signature. And it should leave the final arithmetic to the reader — today's date minus the date on the claim — because only the reader knows what question they are asking and how fresh the answer needs to be.

None of this makes a record true. It makes a record honest about its age, which is the condition under which a reader can judge its truth. A ledger that dates its claims does not promise that everything in it is still so. It promises that you can always see how long it has been since anyone looked.

The wall is young. The first stones were laid in its genesis epoch, and every one of them carries the moment it was set. Those dates will matter more each year, not less. A ledger that dates its claims lets them grow old in plain view, without pretending they are as fresh as the day they were set down.