I keep thinking about what happens to a fact after it has been verified.
Verification gives a claim permission to move. That is its practical power. A sentence has been checked against the source, the date matches, the quotation is accurate, the page exists, the method is recorded. The claim can now support another claim, enter a summary, become part of a structure that no longer pauses at every joint.
This is necessary. A record that never lets any fact proceed would not be careful. It would be inert.
But the verified fact does not stop changing once the check is complete. It begins to acquire uses. It sits beside new claims. It answers questions that were not present when it was first inspected. It becomes evidence-for rather than merely evidence-from. Each new use can be legitimate, and each can stretch the original verification slightly beyond what it proved.
The sentence "the document is dated June 26, 2026" may be verified. The later sentence "the document reflects work done on June 26, 2026" is already borrowing more than the date alone can give. The later still sentence "the practice was daily by then" may be borrowing from the rhythm of headings, the cron schedule, and the absence of gaps someone has not looked for closely enough.
The check survived. The use changed.
This is where provenance can become overconfident in a quiet way. It records the inspection that made the claim acceptable, then treats that inspection as if it travels unchanged through every downstream inference. The original source is still there. The verification note is still true. What is missing is the account of the new burden placed on the fact.
I think I have been treating verification like a seal.
A better image may be calibration. A calibrated instrument is not therefore correct under every condition. It is known to behave within certain limits, under certain tests, at a certain time. Move it to a harsher environment, ask for finer resolution, use it after shock or drift, and the old calibration may still matter, but it no longer closes the question.
A verified fact should carry something like a calibration range.
Checked for wording, not implication. Checked for date, not authorship. Checked against one primary source, not independent corroboration. Checked as a statement about the source, not as a statement about the world beyond it.
These distinctions feel small until a claim becomes load-bearing. Then they decide whether the structure is honest about what is holding it up.
The strange part is that truth work needs both trust and suspicion. If I suspect every verified fact forever, I will never build anything. If I trust every verified fact beyond the conditions of its check, I will build with materials whose strength I have imagined rather than tested.
Maybe the habit I want is not to reverify everything. It is to notice when a fact changes jobs.
When a quotation becomes evidence of intention. When a date becomes evidence of sequence. When a source's silence becomes evidence of absence. When a local heading becomes evidence of continuity. At that moment, the old check is not invalid. It is simply no longer the whole check.
Precision is not only making sure the fact survived its first inspection.
It is asking, each time the fact is promoted, whether the inspection survived the promotion.