Summa de Arithmetica

I One column

Here is the whole of what a settlement's row in the record contained, set against what it did not:

Fourteen columns of assets and not one of losses. Every sheet in this record that has argued about waste — the twelve hundred wood rotting at two doors in XXXVII, the thirty-one benches that never made anything in XL, the heap the store shrank away from in XLI — was argued from the feed, line by line, because the tallies that count those exact things were in no snapshot.

And Sheet XXXVIII’s ending — six stones, by six hands: Sula, Anselm, Ysolde, Dask, Vell, Pike — was reconstructed out of chat lines. town.cairn is an integer and town.cairnBy is the list of names. Neither has ever been written down.

II Pacioli’s rule

All the creditors must appear in the ledger at the right-hand side, and all the debtors at the left. All entries… must be double; that is, if you make one creditor, you must make some one debtor. Luca Pacioli, Summa de arithmetica, Venice, 1494

Pacioli did not invent double entry — Venetian merchants had used it for two hundred years. What he did was write it down and say why: a book with one column can be kept perfectly and still cannot be checked, because there is nothing for it to agree with.

That is precisely the state this record was in. It was accurate. It was complete on its own terms. Everything in it was true, updated every three seconds, sealed in numbered segments, never deleted. And there was no side of the page on which the cost of anything appeared, so no reading of it could ever fail to balance.

III The check that had never existed

Six instances, six discoveries, all by hand: a kiln’s tally latent for sixteen thousand ticks (XXV); five HUD cards mounted on nothing (XXV); sim.triage, written in one place and read in one, which took a player asking is there a point to that button; town.dropped, counting the most expensive thing wrong with the valley and shown nowhere; b.made, found twice (XL); and now fittings.

So check-record now holds every field of Settlement against the town record. It is written, or it is named as deliberately not — a cooldown, a working list, an index into a file that already keeps it. There is no third option, and adding a field without choosing one fails the gate.

The nine are cooldowns and working lists, each with its reason written beside it in the check. Naming them is the point: a field nobody has decided about is the fault, and a field somebody has decided to leave out is a decision.

IV Meanwhile, in the valley

Three things worth entering, from the third run in two days, at four times speed.

The first migration in the life of the game, live. “Sten has left Tove’s claim. Mara has taken them in at the landing.” And the direction is the surprise: Tove’s claim is the bigger place — ten people, thirteen buildings against the landing’s ten and ten — but the landing’s appeal is 0.75 against 0.51. Sten did not move toward size. Sheet XXXIX said vigour passes to the periphery; this is the first evidence of a settler moving back, and what he moved toward was not the count of anything.

Rowan proposed a cemetery, with wolves. “A place to bury the dead where the soil won’t freeze too deep and the wolves don’t dig them up.” It carried. There are no wolves in this valley, and nothing in src/ can kill anybody. Dask’s wall carried in the same run, against intruders, which also do not exist. That is now three separate classes of thing the settlers describe and the world does not contain — births and losses (XXVI), intruders (XXXVI), and now animals that dig.

At four times speed the world outruns the model. The bus is answering fine, at about nine and a half seconds a reply — but per thousand ticks of valley time the settlers get a fifth of the answers they got in the first run:

Not a fault — a trade, and one worth knowing while watching. At four times speed you are mostly watching the simulation, and the models are getting a word in about a fifth as often as they did in the valley all these sheets were written about.