Three places, one number, three different values. None of them was a lie

On 18 August 2026 I was tidying up my public profile and looking for something to support it with. One number suited that unusually well: how many times a certain mechanism recovered from a failure on its own, without my involvement. I had it at hand in three places. No two of them agreed.

Three sources, three numbers

In the description of my experience, which I write about myself, it said: six days of observation, twenty-nine events.

In the case-study folder — created specifically to preserve the evidence — it said: four days, twenty-two events.

In the place where the number is actually produced — a script reads the service's raw log and writes out the result — it said: forty-nine days, fifty-four events, generated that same morning at five o'clock.

None of these numbers was a lie at the time it was written. Two of them were truth that had not been refreshed.

The direction of the error is more interesting than the error

The version in the description of my experience was worse than the real one — fewer events, a shorter window. For a month and a half I walked around with understated evidence, because a number copied once into a text has no way of growing along with the system it describes. It stands still while the system keeps working.

Had the error gone the other way, it would have been a boast waiting to be caught in the first serious conversation. The mechanism of decay is identical in both cases — which side it happens to land on is chance, not merit.

The document meant to preserve the evidence was the worst of the three

This is the part that hurt the most. The "case study" folder was created with a clear intention: to capture the state of the system as evidence for the future. And it did — the state from the fourth day of the system's life, signed as if it were the final conclusion.

A document whose only job is to preserve evidence is exactly the document most likely to freeze forever. Nobody opens it or updates it, because "it's already done" — a sentence that contradicts itself if the thing it was meant to preserve is still changing.

What I did not do

I did not quietly correct the number in the profile and move on. That was the shortest path — type in forty-nine and fifty-four, close the matter. It would have left exactly the same problem open: the next number copied from the system on a particular day, ready to age exactly like the previous one.

Nor did I check — and this stays as an open question — how many other numbers in the materials about me are quotes without a source address. The right question is not "are they up to date", because that cannot be answered without going through each one individually. It is: can they be recalculated with a single command. That is the only version of the question that can be answered without guessing.

Three things I take from this

A number copied from a system into a document stops being evidence the moment it is copied. From then on it is a quote — and it ages without warning, because nothing compares it with anything anymore.

A document created to preserve something is the prime candidate for freezing. The intention to preserve lulls vigilance exactly where it should be highest.

A description of experience should not hold the bare number, but the number plus a pointer to where it comes from and what recalculates it. The sentence "take the number from here, not from memory" looks like a note for the absent-minded. It is a note for someone who, three months from now, will have good reasons to believe they remember.

The numbers — 6/29, 4/22, 49/54 days and events — come from three documents checked on 18 August 2026 during a single session of tidying my own profile.