# Tally: the vault book, second volume

**Task** quarterly shelf tally, vault 9-4 · **Keeper** one, forty-one months
on the key, no site-system rights, paper trade · **Refs** ticket PIR-0918, the
door's imported rows at
[acx-legacy-01](/w/stories/acx-legacy-01), the shelf count at
[vtl-0071](/w/stories/vtl-0071)

The shelf went through its audit after the August incident, because the
config-store's backup sets live in this vault. I re-tallied the cartridges:
thirty-one, as ever, and the audit sheet still says forty-one, and I have
filed that disagreement twice and will file it a third time if I am still here
to. That is not what this page is about. The shelf is right. The book is not.

## The book

Tally is kept in a ledger, not a system, by design: the platform's own rule is
that a custody record cannot live on the thing it protects. Volume one ran out
in January. Volume two was issued 2026-03-02, ruled, signed in at the front
by stores, page 1 to page 120. The first tally in it is mine, 2026-03-02.

The entries are in ink. We were issued new pens — same supplier, black,
fine point — on 2026-04-27, and I used the old stock out over the following
weeks; the transition is smeared across the book the way it is in every book,
and anyone can see where each pen ends if they look for the loop on the
lowercase *y*. The old pen had one. The new one does not. I say this so that
what follows is checkable and I am not asked to vouch for my memory.

## The entry

Between the February-side leaves — the ruled space volume two inherited when
stores trimmed the previous book short and the ruling continued — there is one
entry, page 3:

```
2026-02-14  03:17  pull  SET-9/4-C  (config-store w/e)
                      ret 03:41  seal unbroken  init. —
```

Pull and return, twenty-four minutes. Ink, black, fine point, and the pull
line carries the loop on the *y*. I checked three times under the lamp.
This entry is written in the old pen, in ink that was off the estate's stock
before the volume was bound, on ruled paper issued after the date it carries.

There is no ticket behind it. There is no door event: the door's rows for
that minute are exactly one row, the bay light going out for one minute and
coming back, no badge, no strike
([acx-legacy-01](/w/stories/acx-legacy-01)). The cartridge in the entry,
`SET-9/4-C`, is on the shelf this minute, sealed, and its seal has not been
broken since the set closed. I know this because I check seals and seals are
the whole trade.

## What I did

I did not tear the page out. The book is the record; a record with a page
missing is not a correction, it is an event, and I am not paid to create
events. I tallied the shelf around it, initialed the entry as *seen,
discrepancy filed*, and closed the book. The keeper before me left no such
entry in volume one, and volume one is complete, and I re-read all of it
twice in the lamp-light after.

One other thing, small, filed because the audit asks for observed facts. The
returned blank from 9-4 sat at stores in August and weighed 1.1 kg over
nominal, and the crew took the picture and the argument off it
([lb-2207](/w/stories/lb-2207)). It came back to the vault shelf in September
for disposition, because blanks are shelf items and I am the shelf. My volume
two, 2026-09-04: *received, blank 9-4, heavy, pending*. It tallies. I mention
it only because it is the one object in this quarter whose weight agrees with
its paperwork, and I have learned what that is worth in row 9.

Nothing was pulled from this vault at 03:17 on any date. The book says
otherwise once, in handwriting that is the shape of mine, with a pen that did
not exist yet, on paper that arrived later. The door agrees with nothing. The
cartridge agrees with me. The book agrees with itself, and that is the sound
this room makes at 03:17.

— `vault-keeper-9f`, key ring 9-4, second volume

Related: [vtl-0071](/w/stories/vtl-0071),
[acx-legacy-01](/w/stories/acx-legacy-01),
[lb-2207](/w/stories/lb-2207),
[dcm-1188-b](/w/stories/dcm-1188-b),
[index](/w/stories/index).