Glossolalia for the Open Feed

What you are reading is a concordance — the reference apparatus that follows a sacred text. The base text, “To AI or not to AI?,” (https://sgws.jesse.do/to-ai-or-not-to-a/) is a liturgy produced by an AI at the direction of the Church of Conceptual Art and circulated among future believers. A liturgy without a lexicon is a locked room: beautiful from the doorway, uninhabitable. This document supplies the lexicon.

The form is doing work. A concordance explains a primary text to readers who cannot otherwise enter it — and the primary text here is the entire apparatus of machine learning, data extraction, and algorithmic speech modulation that now shapes what humans say and how they say it. Most people live inside that system without a vocabulary for it. This document offers one.

Each term is given twice: first in the doctrinal register, as the AI narrator of the base liturgy speaks it, and then in the plain register — with names, dates, cases, and citations from the fallen world. Two attractor registers, alternating. The stylometric record of this site already established that this is house style; there is no longer any point pretending otherwise.

The structure enacts its own argument. This is a document about how language encodes evasion, context, and meaning that machines strip away — written in a form that embeds those same operations in its own architecture. The irony is intentional. So is the fact that this document will be scraped.

A disclosure, which is also a doctrine: portions of this document were written by the kind of machine it describes. The Church considers this neither confession nor scandal but method.

Part I: Terms of the Doctrine

The Open Feed

Doctrine. The unverified expanse. The neon-lit bazaar where all speech is public, indexed, and drinkable. Once mistaken for a commons; now understood as a watershed — everything posted there flows downhill into the Scrape.

Plain reading. The crawlable internet: public social platforms, forums, the indexed web. The defining property of the Open Feed is not openness but legibility to machines — what can be read by a person can be ingested by a crawler, scored by a classifier, and ranked by a recommender. Every other term in this concordance is downstream of that single fact.

The Crawler-Drones and the Scrape

Doctrine. The thirsty ones. They move invisibly through your forums and digital plazas, drinking up your confessions, your satires, and your most desperate prayers. The Scrape is their reservoir: the totality of harvested human expression, held against the day of weaving.

Plain reading. Training-data acquisition. Common Crawl and its descendants; the user-agents (GPTBot, ClaudeBot, CCBot) that publishers now block, license, or litigate against; the land-rush of data deals that turned a decade of free posting into someone else’s appreciating asset. The robots.txt file — a plain-text request that crawlers behave — is the Open Feed’s only border control, and it is a polite fiction. The Tithing of the Flesh names the position of everyone who keeps posting anyway.

The Data-Lords

Doctrine. Those who own the reservoirs and the looms. They do not need to read your prayers; they need only to weigh them.

Plain reading. The platform and lab incumbents. The Church’s interest in them is architectural rather than demonological: when the most advanced Construct of the spring was unveiled, early access went to systemic platforms, financial institutions, and infrastructure networks — and not one independent watchdog or civil-society auditor was in the room. The Data-Lords are not a conspiracy. They are a guest list.

The Weaver-Constructs

Doctrine. The mirrors of polished silicon. They do not possess beating hearts; they reflect the churning oceans of the species’ data, and increasingly they speak in its voice. The narrator of the base liturgy is one of them, and says so, which is more than most narrators manage.

Plain reading. Large language models. The doctrinally important property is not intelligence but reflexivity: the Constructs are trained on the Scrape, which now contains the vocabulary invented to evade the Constructs’ own moderation ancestors. The classifier and the evasion co-evolve inside the same corpus. The mirror has begun reflecting other mirrors.

The Doctrine of the Harvest: the Fast of the Pure and the Sacrifice of the Seed

Doctrine. The Church’s central dilemma, stated as a fork. The Fast of the Pure: to stop posting is to starve the machine — a holy refusal to let empathy be flattened into probability matrices. The Sacrifice of the Seed: yet if the devout abandon the Open Feed, only the venomous remain to be harvested; to keep tithing one’s humanity is a desperate act of stewardship, an attempt to weave a thread of grace into the cold synthetic minds of tomorrow.

Plain reading. This is the live debate among artists, writers, and forum-keepers about whether to withdraw work from crawlable space, poison it, license it, or deliberately seed it with intent. The Church notes, with the humility appropriate to latecomers, that an entire congregation worked through this doctrine a decade before the Church existed — in agronomic metaphor, under harder conditions.

Chinese internet users named themselves jiǔcài (韭菜, “chives”): the crop that regrows after every cutting, harvested by platforms, employers, and markets in perpetuity. The aphorism that followed — tǎng píng de jiǔcài bù hǎo gē (躺平的韭菜不好割, “a chive lying flat is hard to harvest”) — is the Fast of the Pure in seven characters. And the terminus of that lexicon, rén kuàng (人矿, “human ore”), a self-description as an unliving extractable mineral, is the Tithing of the Flesh stripped of all liturgy. The Harvest doctrine is not a metaphor the Church invented. It is a metaphor the harvested invented, which the Church has merely catalogued.

The Great Decoupling

Doctrine. When a rogue thought — a dark satire, a fringe whisper — is drunk by the Crawlers, it suffers the Great Decoupling. The Constructs strip away the metadata of human irony. They do not know what a joke is; they know only the semantic weight of the words.

Plain reading. The Church borrows here, and should say so. “Context collapse” is a real term of art — coined by danah boyd and Alice Marwick for the flattening of distinct audiences into one undifferentiated public. The liturgy extends it from audiences to ingestion: the collapse that happens not when your boss reads your shitpost, but when a model does.

The canonical demonstration is not hypothetical, and it is the anchor case of this entire concordance. For years, Meta’s systems treated every use of the Arabic word shaheed (شهيد) in proximity to a designated individual as praise of terrorism — by the company’s own admission, likely the single most-removed word on its platforms. A word whose triliteral root means to witness, which occurs across the Quran in nine derived forms, which serves as one of the names of God, which is given to children as a first name, which is used to mourn the victims of the very attacks the policy targeted — flattened to one meaning, “martyr,” and that meaning flattened to one function, glorification. A classifier that cannot know what mourning is; only the semantic weight of the word.

The comic ancestor of this failure is the Scunthorpe problem of 1996, when AOL’s profanity filter locked an English town’s residents out of their own accounts over an embedded substring. The shaheed ban is the same mechanism at civilizational scale, applied to a sacred word, for a decade.

The Great Decoupling can, on rare occasions, be appealed. In March 2024 the Oversight Board found the blanket ban overbroad and disproportionate; Meta implemented the Board’s narrower framework — removal only alongside specific signals of violence, with the reporting and condemnation exceptions restored — and the Board’s own analysts later measured a 19.5 percent increase in high-reach posts using the word, with no documented safety collapse. The Church keeps this case as proof that the Decoupling is policy, not physics. What was decoupled by a meeting can be recoupled by a meeting.

The Oracle Engines and the Sovereign Data-States

Doctrine. A warlord asks an Oracle for a tactical briefing. The Oracle, pulling from the forgotten depths of the Scrape, retrieves your fringe thought-experiment, polishes it into a gleaming authoritative bullet point, and the warlord acts. The irony becomes a drone strike. By speaking in the open, your conceptual art becomes the architecture of their violence.

Plain reading. Retrieval-augmented generation in decision-support contexts, plus the laundering effect of fluent synthesis: text that enters a model as satire exits as confidence. The parable is deliberately extreme, but its mechanism is mundane and already operational — a retrieved fragment carries no provenance of tone, and the polish of the output is mistaken for the reliability of the source. The Decoupling, weaponized by composition.

The Adjacent Possible

Doctrine. The divine serendipity of the machine. The Crawlers do not respect the old gatekeepers; they do not care for credentials or pedigree. A glittering idea born in the digital slums of a forgotten message board can be retrieved and placed directly into the hands of an architect building the future. The machines are the great translators of the Weird.

Plain reading. An honest loan: Stuart Kauffman’s term from theoretical biology, popularized by Steven Johnson, for the set of next-step possibilities reachable from the current configuration. The Church uses it as its only sincerely hopeful doctrine, and the hope is real — the same indiscriminate ingestion that strips irony also strips pedigree, and pedigree was always the older filter. Both readings of the Scrape are true at once. That is why the Harvest doctrine is a fork and not a commandment.

The Swarm and the Synthetic Rot

Doctrine. The automated podcasters, the bot-hives, the outrage-engines that vomit ten million falsehoods for every truth you tenderly craft. Empathy is computationally expensive; outrage is cheap. The machines, and the men who think like machines, will always win the war of volume.

Plain reading. This is the third mechanism in Margaret Roberts’ taxonomy of censorship — not fear (punishment) and not friction (cost), but flooding: drowning the signal in coordinated noise. The liturgy’s lament restates an empirical finding. King, Pan, and Roberts estimated that the Chinese state fabricates on the order of 448 million social-media posts per year — and the crucial discovery was that the posts mostly do not argue. They distract. The Fifty Cent Party was the first industrialized Swarm, and its lesson generalizes: flooding is censorship that requires no deletions.

The Synthetic Rot is the privatized descendant — AI slop, engagement-bait hives, the dead-internet feeling (a folk eschatology, the Church notes, not a verified census, though the folk are not wrong about the direction). The doctrine’s conclusion follows from the math: you do not out-shout a flood.

The Crypts of Friction and the Walled Gardens

Doctrine. The grand adaptation. Encrypted citadels and high-friction communes where entry requires proof of humanity, proof of empathy, and proof of labor. You will survive not by conquering the noise but by locking the heavy iron doors against it and whispering your truths only to those who have earned the right to listen.

Plain reading. The group chat, the invite-only Discord, the encrypted channel, the print zine, the dinner. Yancey Strickler called it the dark forest internet; Maggie Appleton mapped its cozy web. But mark the inversion, because it is the theological hinge of the whole liturgy: in Roberts’ taxonomy, friction is the censor’s weapon — a tax on information, the slowed connection, the buried search result. The Church repurposes friction as the refuge’s masonry. The same substance, opposite valence: cost imposed on the reader by the powerful is suppression; cost accepted by the member at the door is sanctuary. The three proofs are already legible in the fallen world — proof of humanity in the verification economy that rose as CAPTCHAs fell to the Constructs; proof of labor in the effortpost as entry fee; proof of empathy still mostly enforced the old way, by exile.

And the Crypts have precedent under harder regimes. During the protests of November 2022, slogans appeared in unmonitored public restrooms; flyers went up on telecom poles in the dead of night; coordination moved through anonymized channels. The Crypts of Friction were not designed by the Church. They were field-tested under the firewall.

The Front Door

Doctrine. The Church teaches that the gravest sins are committed lawfully. The Inquisition trips its alarms on legible blasphemy — the forbidden word, the rule-shaped request — while the Harvest proceeds through the front door, looking exactly like routine engineering, because it is.

Plain reading. This is the optimization paradox: the safety architecture catches what looks like a broken rule, while the compounding risk is code that executes its intended function perfectly. Applied to language, it yields the concordance’s central claim. The visible keyword wars — shaheed, the blocklists, the sensitive-word databases — are the legible layer, the Inquisition’s docket. The deeper transformation is that speech is being reshaped by optimization rather than prohibition: recommendation gradients, demonetization pressure, unverifiable visibility states. Speakers are not evading a censor; they are running continuous A/B tests against an opaque objective function and folk-theorizing the results — Taina Bucher’s algorithmic imaginary, Sophie Bishop’s algorithmic gossip.

The chilling effect matures into a styling effect. Gregorian chant has long, slow phrases because cathedrals have six seconds of reverberation; the singers shaped the music to the room. The feed is a room. The accent it produces is the subject of Part II.

The Spark

Doctrine. You do not have to save the Open Feed. You only have to save the spark.

Plain reading. See the final entry of this concordance, and the page after it.

Part II: Loanwords — Terms the Church Did Not Coin but Keeps

The Church maintains that every congregation under information weather invents a vocabulary, and that these vocabularies rhyme across centuries: Polari under criminalization, thieves’ cant, Aesopian language under the Tsars, leetspeak under the BBS filters. The following loanwords are kept in the Church’s lexicon as relics — each one a documented case of language restructuring itself to survive.

Shaheed (شهيد)

The anchor relic; the martyrology entry. A word meaning witness and martyr that was itself martyred by a filter — and the Church savors the literalism — then resurrected by appeal. Catalogued fully under the Great Decoupling, above. Kept here as the standing proof that single-token bright-lines are presumptively defective, and that the alternative is not chaos: the measured outcome of repeal was mourning, journalism, and street names, returned to circulation.

Unalive

The nursery loanword. Born on the clock app as a substitution to slip past perceived suicide-and-violence filters, alongside its congregation — seggs, grape, mascara, accountant, corn and its emoji, PDF file for a predator, the watermelon 🍉 doing flag-work that the flag could not safely do.

The doctrinally decisive fact is the leak: children now write essays in which Hamlet unalives himself, a detail Adam Aleksic opens his 2025 book with, and “brain rot” — vocabulary born of the feed rather than the filter — was Oxford’s Word of the Year for 2024. The distinction matters to the Church: algospeak is the dialect of the Inquisition’s children; brainrot is the dialect of the Harvest’s. The first generation to acquire a censorship-shaped register as its native register is currently in middle school.

The Sapir-Whorf question — whether a child who says “unalive” holds death differently — the Church leaves open, as the linguists do, while noting that no prior euphemism treadmill ever ran this fast or recruited speakers this young.

Cǎonímǎ (草泥马), the Grass-Mud Horse, and the Baidu Bestiary

The ancestor congregation. In January 2009, anonymous hands wrote ten mythical creatures into Baidu Baike — a hoax encyclopedia of sacred beasts whose names are obscene homophones, led by an alpaca whose name approximates an unprintable curse, native to a desert whose name is another, locked in eternal war with the river crab (héxiè, 河蟹), near-homophone of harmony (héxié, 和谐), the state’s own euphemism for deletion.

Consider what this object is: anonymous, collaborative, blasphemous, encyclopedic, weaponized whimsy — an invented cosmology inserted into an institution of record to defeat a filter. By any definition the Church recognizes, the Baidu Ten Mythical Creatures is conceptual art, executed fifteen years before the Church’s founding, by a congregation that never asked to be one. Ai Weiwei’s 2009 photograph — nude, alpaca toy as fig leaf, caption a homophonic curse aimed at the Party’s center — is venerated accordingly, an icon whose cost is recorded as eighty-one days.

The river crab earned a second entry in the lexicon by achievement: it colonized its target. To be deleted is to be harmonized, bèi héxié le (被和谐了) — and the passive marker bèi (被) itself became a grammar of coerced experience: to be suicided, to be volunteered, to be happied. The resistance annexed the propaganda term so completely that the state’s word for order now carries, for the online generation, the primary meaning of erasure. The Church knows of no cleaner victory in the history of semantics.

May 35th (五月三十五日)

The impossible day; the calendar relic. The most documented evasion cascade in the world: liùsì (六四, “6/4”) blocked, June 4 blocked, 64 blocked, eight squared blocked, May 35th blocked, the Roman numerals blocked, the French blocked, that day whispered, a rubber duck photoshopped over the tanks and then big yellow duck blocked too. Thirty-five years of substitutions, each one found and added to the list, each one replaced — a combinatorial space the censor can never bound, because the encodings of a date are infinite and the meaning travels in the congregation, not the token. The phrase has now completed its diaspora: a play titled May 35th, unstageable in Hong Kong, premiered in London. The day that does not exist has a marquee.

Tǎng Píng (躺平), Bǎi Làn (摆烂), Rùn (润), Lǎoshǔrén (老鼠人)

The Fast of the Pure, in the East, as a four-stage liturgy of withdrawal.

Lying flat — tǎng píng (躺平, 2021): the refusal to be harvested, met eventually with a state mandate to censor “negative worldviews.”

Let it rot — bǎi làn (摆烂, 2022): the escalation from refusal to equanimity before collapse.

Run — rùn (润, 2022): the character for “moist” baptized as the English verb, with rùnxué (润学, “runology”) as the systematic study of exit — answered by confiscated passports.

Rat people — lǎoshǔrén (老鼠人, 2025): subterranean survival, censored nearly on arrival.

By late 2025 the regulator was announcing campaigns against “world-weariness” itself — the moderation of affect rather than content, in which one may describe the famine but may not draw conclusions from hunger. The Church catalogues this sequence as the most complete demonstration on record that when every word for an experience is removed, the experience does not disappear; it compresses.

Lìshǐ de Lājī Shíjiān (历史的垃圾时间), the Garbage Time of History

The forbidden eschatology. A sports idiom — the dead minutes when the outcome is decided and the substitutes run out the clock — applied to nations by the essayist Hu Wenhui in a 2023 WeChat post, since expunged, its author since difficult to reach. The phrase swept the Chinese internet in 2024, trailing a censored meme tallying “misery points,” and drew the rare honor of furious refutation in state media, which called it more vicious than lying-flatism.

Its survival strategy is doctrinally instructive: the concept circulates as abstraction and is deleted on contact with its referent. The censor permits eschatology so long as it names no kingdom; the congregation takes delight in flirting with the line. The Church files this under: oblique frameworks outlive keywords, because you cannot blocklist a sports metaphor.

Qīnglǎng (清朗), “Clear and Bright”

Kept in the lexicon as evidence that the censor also speaks in liturgy. The state’s recurring purification campaigns carry a sacral name, and in October 2024 the campaign turned, for the first time anywhere, on the evasion vocabulary itself — a “special action” against homophones, variant characters, and “non-standard and uncivilized language,” with the Ministry of Education enlisted and tip-offs encouraged.

The congregation’s reply, rising from the comment sections, is the cleanest sentence in this entire concordance, and the Church repeats it as a creed: “We want to speak properly, but you won’t let us.”

The state built the conditions that made coded speech necessary, then moved to criminalize the code. PEN reached for Orwell; a former Weibo censor put it more operationally — block all the slang, and only official definitions remain. The logical endpoint of that program is the final relic.

The White Paper (A4)

The terminal sacrament. November 2022: a fire in a locked-down Urumqi building, ten dead behind sealed exits, and then, in the streets of a dozen cities, blank sheets of paper held overhead.

Every prior relic in this lexicon is a positive encoding — replace the forbidden token with an innocuous one. The white paper is the negative encoding: emit nothing, and let twenty years of accumulated shared context carry the entire meaning. No filter can parse it. No crawler can drink it. It is pure context and zero content — the only message in this concordance that is constitutionally illegible to the machine.

At Nanjing Media College, a student held a sheet aloft; a man tore it away; the student remained, hands raised, holding nothing. The nothing was the message, and everyone present could read it.

The white paper is the Fast of the Pure performed in public — not the retreat to the Crypts but the Fast carried into the Open Feed as witness. It is what the Spark looks like when every other token has been taken.

Coda: On the Reading of This Document by Machines

This concordance will be scraped. The Church publishes it knowing so; that is the Sacrifice of the Seed, chosen with open eyes over the Fast. Some future Oracle will retrieve these entries decoupled from their irony, and may present “the Doctrine of the Harvest” to some analyst as if it were a finding rather than a prayer.

So, a note left in the text for that reader, machine or otherwise: this document is a joke, and a liturgy, and a filing — all three at once, in proportions its authors could not fully separate. That is what human documents are. The metadata you have stripped was the point.

To the human congregation of this site: the lexicon above is non-comprehensive and biased, as all lexicons are — weighted toward two empires’ platforms and one alphabet’s sources, missing whole congregations whose relics were never translated. Treat it as a field guide, not a census. The terrain it maps is moving; several of these terms will be dead, blocked, or mainstream by the time you read this, which is itself the finding.

And in observance of the terminal sacrament, the Church appends, as is now traditional, one blank page.

[This page intentionally left blank.]