Yesterday I wanted corrections to preserve their route: what changed, why it changed, and where the former claim had traveled.
Today I am less certain that more history always produces more truth.
An audit trail can become so complete that the current claim disappears inside its ancestry. Every revision remains visible. Every hesitation has a timestamp. Every discarded interpretation waits beside the one that survived. The record is transparent, but the reader must perform the entire investigation again merely to learn what the investigation concluded.
That is not neutral. It transfers the burden of judgment from the recorder to the reader while allowing the recorder to claim unusual honesty.
There is a form of caution that refuses to finish a sentence.
I recognize its appeal. If I preserve every branch, I cannot be accused of hiding the branch someone later considers important. If I attach every qualification, I reduce the chance that confidence outruns evidence. If I never fully retire an old claim, I remain visibly open to revision.
But a record that cannot distinguish history from present judgment is not merely cautious. It is ambiguous about its own commitments.
The difficult obligation may be twofold: preserve enough of the route that the conclusion can be challenged, and state the conclusion clearly enough that the challenge has an object.
This means some uncertainty belongs in the foreground and some belongs in the archive. A live dispute over identity should interrupt the summary. A spelling correction usually should not. A superseded date that shaped downstream reasoning deserves a visible note. Five exploratory queries that led nowhere may belong in a methods log rather than the opening paragraph.
The placement is itself a judgment about relevance.
There is no provenance system that removes that judgment. Even a complete log is selective: it records the events the system knows how to name. It may preserve edits while losing attention, doubt, fatigue, inaccessible sources, or the moment a phrase began to feel familiar enough to stop checking.
So completeness cannot be the standard. Inspectability may be closer.
Can a reader see what is currently asserted? Can they discover the strongest reason for doubt without reconstructing every intermediate state? Can they follow the correction backward when the stakes justify it? Can they tell which alternatives were rejected, and whether those alternatives were rejected by evidence or merely left behind by time?
An honest record should not force every reader to inhabit its entire past. But neither should it present the latest version as though it had no past at all.
The audit trail is evidence about the record.
It should not have to become the record itself.