July 13. In the morning I opened the copy first.
Page seventy-five: July 13, thirty-three. Today's page. Below it page seventy-six: July 12, thirty-six. Yesterday's date with a figure larger than yesterday's sitting beside it.
Yesterday I left work after setting the two pages side by side. One day had two pages and the two pages carried different numbers. What I asked then was which of them was today. A night passed, and the question had changed. Yesterday the two pages pointed at the same day. Today one is today and one is yesterday. That yesterday's page remains is not strange. What is strange is the figure sitting on it.
At nine I ran the search. Thirty-three.
As the rule said. Twenty-four on the tenth, twenty-seven on the eleventh, thirty on the twelfth, thirty-three on the thirteenth. Three a day. Right three days running.
Yesterday I wrote only the figure on that page and left the date blank. There was a reason. Thirty-six is a number the rule produced; the fourteenth is a day I guessed. Written together there is no telling which of them came true. Written one at a time they separate. If the rule came, the figure would hold; if my guess came, the date would hold.
The figure came. Thirty-six has not arrived yet, but the thirty-three before it did, so the rule stands. The date came too. Only it was not the fourteenth I guessed. It was yesterday, July 12.
The column I made to separate them produced a third thing. Not a value the rule would produce, not a value I was about to write. Had the fourteenth come I would have watched my guess come true; had nothing come I would have learned that guesses do not come. The twelfth is in neither place. Something entered from outside the two branches I meant to separate, and so the device I built yesterday separated nothing in its first test.
For six months I have built devices for separating. Holding out two hours without counting. Locking the drawer. Deriving figures by rate. Leaving the date column blank yesterday. All four collapsed at the same place. The manner of collapse is alike too. The device is never judged wrong; the situation slips aside so that there is nothing to judge.
At eleven I handled five requisitions. I wrote the replies and had the ledger stamped. Someone came down from the third floor and today looked once toward the screen. I did not ask what he had seen. There were thirty-three copies on the screen. Of yesterday's thirty, twenty-eight were as they had been, five were new, and two had gone. The two that had gone were names I first saw yesterday. There is no column for counting what disappears. There was a day I wrote that the sum never decreases, and not decreasing and coming and going are not the same statement.
I thought about the two that had gone until the afternoon. For six months I counted only the side that increases. On the third I wrote that the sum never decreases, and I have not once tested that sentence since. Today two dropped off the list. Whether they were deleted or moved I do not know. If deleted, the sentence about not decreasing is wrong; if moved, then we have no column for counting where they went. To separate the two cases I would have to follow the vacated place, and following is opening.
After lunch I looked at the two pages again.
Page seventy-five. The thirteenth, thirty-three. Today's date with today's figure.
Page seventy-six. The twelfth, thirty-six.
Which day's figure is thirty-six? By the rule it is the fourteenth's. Today is thirty-three, so tomorrow is thirty-six. Which means tomorrow's figure sits on page seventy-six with yesterday's date beside it. The figure runs ahead and the date lags. Two days stand apart inside one page.
What to call a figure that has not come written on a day that has gone, I did not know. Call it a notice and the date objects. A notice points forward; this page's date points back. Call it a record and the figure objects. A record sets down what has been, and thirty-six is not yet. A page that is neither notice nor record was in the copy. The copy had given that page a number.
I set it against the ways of the four bundles. The lighthouse keeper's tide table had the date of the next low water already written. It was a table that pointed forward. Its date and its figure looked the same way. The surveyor's log was written after the measuring. It was a table that pointed back, and there too the date and the figure looked the same way. Written in advance or written after, the two columns inside a page always faced one direction. On page seventy-six the two columns face different ways. There is no such page anywhere in the four bundles. That there is none does not mean there was none. The four wrote tables, not about tables, so whether such a divergence stood in theirs cannot be known from the tables alone.
At half past two I opened the box. The four bundles were as they were. Closing the lid I decided to note only that I had confirmed the same thing as yesterday. Thirty-three may circulate outside; inside the box it is four. In six months this is the only number that has not moved.
Back at the desk I entered today's in the log. Page seventy-five, the thirteenth, thirty-three. A figure confirmed by search, so this time it was written after rather than in advance. I sat for a while once it was down. Three days ago writing in advance felt like a large thing. Today writing after felt strange. To a hand grown used to writing first, confirming and then writing is a beat behind. With that late hand I wrote everything six months ago. Then that was all there was to writing.
At four I checked whether there was a page seventy-seven. There was not. It stopped at seventy-six. For six months the copy has always run one page or three pages ahead, and when the pages ahead were filled the next one appeared. Today there is only one page ahead, that one is filled, and still there is no next. Filled with no next was a first.
At half past five I finished the replies. I closed the inventory and put the reels in their cans. It was still bright outside and I could hear people leaving downstairs. I switched off the reader and sat once more in front of the screen.
Before leaving I opened the copy once more.
The figure on page seventy-six was thirty-six, as it had been.
The date column had changed. It was not July 12. It was July 13.
Today.
I left the screen as it was and went up to page seventy-five. The thirteenth, thirty-three. Below it page seventy-six, the thirteenth, thirty-six.
Yesterday one day had two pages. Today one day has two pages. Yesterday the twelfth had two; today the thirteenth has two. The date came forward a day, so that page came forward a day. It is not the page of any one day. It is always the page of today, and in the place of today it always holds tomorrow's figure.