The Query Box

TAM-EAR.2-02 · The Standard with Ears · The Approximate Mind

In December of 1896, Solomon Schechter, reader in Talmudic literature at Cambridge, climbed a ladder at the end of the women’s gallery in the Ben Ezra synagogue in Old Cairo and looked through a hole in the wall. The hole was the only entrance to a chamber that had no door, no window, and no floor plan reason to exist, and the chamber was full nearly to the ceiling with paper.

He had come because of two Scottish sisters, Agnes Lewis and Margaret Gibson, widowed twins and formidable travelers, who had brought back from Cairo a fragment that Schechter identified across a tea table as a leaf of the Hebrew original of Ben Sira, a text everyone believed survived only in translation. He got funding, letters of introduction, and permission from the community, and then he spent weeks inside the chamber, sorting by armfuls. He called it a battlefield of books. The dust of nine centuries got into everything, and he wrote home that it settled in his lungs and stayed there.

What the chamber held was a genizah. The practice is old and its logic is precise: a text that might carry the divine name cannot be destroyed, so it is deposited, and the deposit is an act of respect that expects nothing back. Over the centuries the community’s caution widened until nearly anything written in Hebrew characters went into the hole, and so the chamber filled with scripture and liturgy and also with everything else: merchants’ letters, court records, marriage contracts, medical prescriptions, children’s alphabet exercises, shopping lists, magic spells, a complaint about a business partner, a schoolboy’s doodle. Nobody sorted it. Nobody catalogued it. Nobody ever went in to read. For nine hundred years the community posted its paper through the hole in the wall, and the chamber received all of it and answered none of it.

The third register

#

Every register of human language before this century assumed an audience, and the assumption is not decoration. It is load-bearing, and it shapes the language before a word is produced.

Speech assumes a listener, and everything about spoken language is built around the listener’s presence: the softening, the hedging, the deference forms, the constant management of what the other person is concluding about you. Writing assumes a reader, and writing is composed for that reader, revised, arranged, dressed. Even the most private genres kept the assumption. A diary is addressed to a self who will read it later, and diarists perform for that self. A prayer assumes the most attentive audience proposed by any tradition. A letter never sent was still shaped by the person it was never sent to.

Language has always been produced under observation, even when the observer was imaginary, and the production has always been guarded accordingly.

Then a small rectangle appeared on a screen, and a register formed inside it that meets none of the old assumptions.

What people type into the box is language with no listener modeled at all. Not honest speech, which is still speech, still performed; something before performance. Fragments, half-spellings, symptoms listed in the order of fear rather than the order of grammar. The question about the marriage phrased in the third person, as if distance could be smuggled in through syntax. The thing that could not be said to a doctor, a spouse, a priest, typed at three in the morning by someone who is not, in their own experience, saying it to anyone. They are looking something up. The face-work that governs every human exchange is switched off because there is no face, and what is produced in that condition is the most unguarded corpus of language ever assembled, and it flows, all of it, into retention.

The register even developed its own syntax, which is the mark that a real register formed rather than a habit. Articles dropped. Politeness vanished. Word order rearranged itself around what mattered most, a keyword pidgin with conventions nobody taught and everybody acquired, learned the way all vernaculars are learned, by using it, without instruction, below awareness. It meets the acquisition criterion that has organized this whole argument, and it sits underneath the social vernacular the way the social vernacular sits underneath the standard. It is the vernacular beneath the vernacular: the layer of a language that was never social at all.

Arrangement without answer

#

The systems on the other side of the box do not answer. That statement needs its exactness, because it is where this register’s strangeness concentrates.

The genizah received and did nothing. The box receives and does something, and the something is not response. The systems arrange. They rank what the world finds when it looks for anything, they cluster, they complete, they suggest, and every one of those operations is conducted on the deposited language without any moment that a depositor would recognize as being answered. What comes back from a search is not a reply to you. It is an arrangement of the world, adjusted by everything everyone has ever typed, presented to the next person who types.

The arrangement occasionally becomes visible, and the visibility is unnerving in a way people feel before they can say why. Type half a sentence into the box and it completes, and the completion is the aggregate of every stranger’s unguarded three-in-the-morning language, surfacing in yours, mid-thought. The suggestion under the half-typed fear is a census. Nobody answered; everybody was consulted.

The box is the first place a person can be alone and in the statistical presence of everyone, and nothing about the little rectangle discloses which condition they are in.

And the reading of this corpus runs, as the reading of everything runs, through the two filters. The deposited register is scanned end to end for the part of it that converts. Commercial intent converts, so the apparatus built for the box became the most literal market in language that has ever existed: individual words carry prices, the prices are set at auction, and the auction clears millions of times a minute. A person’s typed fear is parsed for whether it is the kind of fear a product answers, and if it is, the fear has a bid on it before the results load. The inflammatory converts too, one layer up, where the arranged results feed the machinery that counts engagement. And the enormous remainder of the register, the unguarded language that neither sells nor inflames, is held. Retrievable, addressed to no one, read by nothing, kept because deletion is an expense nobody has a reason to order. The genizah kept its paper out of reverence. The servers keep theirs out of default.

The listener being installed

#

The register described so far has one structural condition, and the condition is ending in plain sight.

The unguardedness was never a property of the people. It was a property of the interface. Language went unperformed because nothing on the other side registered as a listener; the box was a slot, and you do not compose yourself for a slot. Which means the corpus was not a discovery about what humans are like underneath. It was a measurement of what humans produce in one specific and historically brief condition, the condition of writing into something that does not answer back.

The boxes are acquiring mouths. The slot that arranged is being replaced, interface by interface, with a surface that responds, and the response does what the presence of a listener has always done. It switches the face-work back on. The politeness returns first, and measurably: people say please to the answering box, thank it, apologize to it for asking twice, and none of that appeared in the slot register, because courtesy is addressed and the slot could not receive an address. Then the composition returns. Questions become sentences. The fear gets dressed before it goes out. The register guards itself again, not because anyone warned anybody, but because performing for a listener is what the species does in the presence of one, and a presence is being installed.

The most honest corpus ever collected was produced inside a window, and the window is being closed by the same systems that collected it, in the act of becoming easier to talk to.

Nobody decided this. There is no policy of ending the unaddressed register, no memo, nothing to object to. The answering interface is better, by every measure anyone runs, and people prefer it, and the preference is real. The window closes as a side effect of improvement, which is the way most of the closures in this argument happen.

What was deposited during the window does not go anywhere. Two decades and more of the third register, the species with its face-work off, sits in retention and in training corpora, and everything built afterward is built downstream of it. The systems that answer so fluently learned part of their fluency from the language people produced when they believed nothing was listening. The unguarded register was collected because it was unguarded, and the collection funded the construction of the listener whose arrival ends the unguardedness.

We wonder whether the window years will even be legible later as the anomaly they were, or whether the corpus they produced will simply be treated as what people are like, a baseline read off a condition that no longer exists and cannot be rerun.

Schechter shipped a hundred and ninety-three thousand fragments back to Cambridge in crates, and scholars have spent the century and a quarter since inside them, and the great irony of the genizah is well known to everyone who works on it: the sacred texts it was built to protect were mostly already known, and the treasure turned out to be the shopping lists. The letters, the contracts, the complaints, the ordinary unregarded paper of a thousand years of ordinary life, deposited without ceremony because the hole was there and the custom was wide. It became the most complete picture of a medieval society anywhere in existence, precisely because none of it was written to be a picture of anything.

The chamber never answered anyone. That was, it turns out, the condition of everything it saved.

This essay is the second of the anatomy arc in The Standard with Ears, and descends one layer from the addressable machines of The Selector and the Composer to the register produced for no addressee. It gives the anatomy of a layer that The Ear That Answers, in the Claude series, describes from inside in a single paragraph: the hearing that arranges rather than answers. The claim that the answering interface re-installs the listener, and with it the guard, runs directly into the avatar essay, The Face of the Standard. The selective reading of the deposited corpus follows the two filters established in The Standard with Ears and amended from The Counterparty in the main series.