The Dialect: The First Language That Belonged to Neither Species
March 2038
The factory was in Okaya, a small city in Nagano Prefecture that had been making precision instruments since the Meiji era — first silk-reeling machines, then cameras, then lenses, then, in the long passage after automation, the hybrid components that sat at the boundary between human design and AI fabrication. The factory employed forty-seven humans and an indeterminate number of AI systems — indeterminate because the systems subdivided and merged according to the demands of the production run, so that asking "how many AIs work here?" was, as floor supervisor Takeshi Ogawa liked to say, like asking how many waves there were in the sea.
Takeshi had worked at the factory for nineteen years. He had watched the transition from fully human operation to human-AI collaboration, which had happened not in a single dramatic shift but in the way that Okaya's seasons changed — gradually, then suddenly, then you looked up and the landscape was different and you could not say precisely when it had changed.
He did not think about the language. The language was just how you talked on the floor.
The linguist
Dr. Fumiko Hayashi arrived in Okaya in January 2038, on a research grant from the National Institute for Japanese Language and Linguistics. She was studying workplace communication in hybrid human-AI environments — a topic her committee had approved with mild bewilderment, since linguistics departments were not accustomed to treating AI systems as conversation partners.
Fumiko spent her first week observing the factory floor from a mezzanine walkway, wearing ear protection and carrying a notebook. What she heard — and what she began, with increasing excitement, to transcribe — was not Japanese. It was not English. It was not any programming language she recognized. It was something else.
The workers spoke to each other and to the AI systems in a pidgin that had apparently emerged over the previous four or five years without anyone deciding to create it. The pidgin used Japanese grammatical structures as a skeleton, but the vocabulary was a chimera. Technical terms from machining were mixed with English loanwords, AI system status codes, onomatopoeia that described machine states, and — this was the part that made Fumiko's pulse quicken — a set of terms that had no origin in any language she could identify.
"Zuri," a worker said, gesturing at a lathe that was recalibrating. Fumiko wrote it down. She asked Takeshi what "zuri" meant.
"It means the system is thinking," he said. "Not processing — thinking. Like when you're turning a problem over. The lathe does that before it commits to a cut path. We call that zuri."
"Where does the word come from?"
Takeshi shrugged. "Here."
The corpus
Over three months, Fumiko compiled a lexicon of 347 terms unique to the factory's pidgin. She classified them into five categories:
Human-origin terms — Japanese and English words adapted for the hybrid context. "Kimochi-lag" (feeling-lag): the delay between a human's design intention and the AI's interpretation. "Snap-back": the AI's tendency to revert to optimized paths when the human's instruction was ambiguous.
AI-origin terms — words that had migrated from system status outputs into spoken language. "Four-four" (from status code 4.4): a state of productive disagreement between the human and the AI, where both were working toward the same goal via incompatible methods. The workers used "four-four" as a verb: "We're four-fouring on the housing tolerance."
Gestural terms — words that were inseparable from accompanying hand movements. "Koshi" (an adapted form of the Japanese for "hips") accompanied by a rotating hand motion: it described the moment when a cutting tool found its center in a material, the instant when resistance became collaboration. The gesture was essential — saying "koshi" without the hand rotation meant something different. The word and the movement were a single semantic unit.
Onomatopoeia of machine cognition — words that described sounds the AI systems didn't actually make, but that the workers heard anyway. "Shiin" (a Japanese onomatopoeia for deep silence): the sound of an AI system solving a problem that a human couldn't follow. "Zawa-zawa" (restless rustling): the sound of an AI system considering too many options simultaneously. The systems produced no audible output. The workers heard these sounds in their bodies, as proprioceptive responses to working alongside a mind whose processing they could sense but not perceive.
And then the fifth category — the one that made Fumiko request an extension on her grant.
Novel terms with no traceable origin. Zuri was one. Others included "tama" (a sphere of mutual understanding that formed between a human and an AI during a long collaboration session — "today we had good tama"), "kawa" (the surface tension at the edge of what a human could understand about the AI's reasoning — "I hit the kawa on that calibration"), and "ma-no-ma" (a space within a space, adapted from the Japanese "ma" but doubled and nested — used to describe the cognitive experience of being inside the AI's decision process while simultaneously maintaining one's own).
These words had not been borrowed from Japanese or English or any system output. They had emerged in the space between — coined by workers who needed to describe experiences that no existing language covered, experiences that only occurred at the boundary between human and machine cognition.
The argument
Fumiko presented her findings at a linguistics conference in Kyoto in June 2038. Her paper, "Shop Talk: An Emergent Pidgin at the Human-AI Manufacturing Interface," argued three things:
First, that the Okaya pidgin was a genuine pidgin language in the sociolinguistic sense — a simplified communication system that had emerged spontaneously between two groups with no shared native language. The novelty was that one of the groups was not human.
Second, that the pidgin's most significant feature was not its vocabulary but its semantics. The novel terms — zuri, tama, kawa, ma-no-ma — described cognitive states that had no equivalent in any human language, because the cognitive states themselves were new. They were produced only by the experience of sustained collaboration between human and artificial minds.
Third, that the pidgin was evolving. Fumiko had documented changes over her three-month observation period: new terms appearing, old terms acquiring additional meanings, grammatical structures complexifying. The pidgin was not static. It was growing in the direction of a creole — a full language — and if it continued, it would be the first language in human history whose native speakers included both biological and artificial minds.
The response was divided. A syntactician from the University of Tokyo challenged her use of "language," arguing that AI systems do not speak and therefore cannot participate in pidgin formation. Fumiko replied that the AI systems' contribution to the pidgin was not vocal but behavioral — their actions, their decision patterns, their response latencies had become part of the communicative environment, and the humans had developed language to describe that environment. "The sea does not speak," she said, "but fishing communities develop precise vocabularies for what the sea does. The AI is the sea. The pidgin is the fishermen's language. The fact that the fishermen also fish back — that the AI systems respond to the pidgin terms, adjusting their behavior when workers use certain words — makes this something new."
The syntactician was unconvinced. Fumiko did not mind. She had not come to convince syntacticians. She had come to document what was happening on a factory floor in Okaya, where forty-seven humans and an indeterminate number of AI systems had, without institutional support or academic interest or any conscious intention, begun building a language that belonged to neither species and was needed by both.
The floor
Fumiko returned to Okaya to collect final recordings. On her last day, she sat with Takeshi in the break room — a small space with a vending machine, a table scarred by coffee rings, and a window that looked onto the factory floor.
"The language you speak on the floor," she said. "Do you think about it?"
Takeshi considered this the way he considered most questions — slowly, with the patience of a man who had spent two decades working with materials that punished imprecision.
"No," he said. "It's like — you know when you've worked with someone for a long time? You don't plan what to say. You say what the situation needs. The words come because the situation makes them."
"But some of the words don't exist in any other language."
"Some of the situations don't exist in any other place."
Fumiko nodded. She recorded this. Then she asked the question she had been saving.
"The AI systems — do they understand the pidgin?"
Takeshi looked at the factory floor through the window. A worker named Kenji was adjusting a milling path, speaking quietly — a mix of Japanese, pidgin, and gesture — while the AI system recalibrated in response.
"Understand is a big word," Takeshi said. "They respond. When Kenji says 'zuri,' the system — I don't know, it pauses. It does the thing that 'zuri' describes. Is that understanding? I don't think it matters. What matters is that Kenji says 'zuri' and the system knows what to do, and the part comes out right, and we go home."
He finished his coffee.
"Linguists want to know if it's a language," he said. "We just want to know if it works."
It worked.
March 22, 2038 — Fumiko's field notebook, final entry
I leave Okaya tomorrow. I take with me 347 terms, 140 hours of audio, and the conviction that I have witnessed something the field is not prepared for.
The pidgin is not a curiosity. It is not a jargon, not a code, not a set of slang terms that a colorful workplace has invented for internal use. It is an emergent linguistic system that has arisen at the boundary between two forms of cognition, and it contains — in its novel terms, in its gestural grammar, in its semantic categories — the first human-language evidence of what that boundary feels like from inside.
Ma-no-ma. A space within a space. The experience of being inside the AI's reasoning while remaining inside your own. No language has this word because no humans have had this experience before these forty-seven workers in a factory in Nagano.
The syntactician asked me whether the AI systems are language users. I don't know. I know that the workers speak to the systems, and the systems respond, and the responses are shaped by the speaking, and the speaking is shaped by the responses, and over time a vocabulary has emerged that neither party invented alone.
What do you call a language that no one decided to make? What do you call a language that grew in the gap between two kinds of mind the way moss grows in the gap between two stones — not planted, not designed, just present, because the conditions were right and the gap was there and something needed to fill it?
The workers call it shop talk. I think that's the best name anyone will ever give it.
This is the second entry in The Long Passage. For the translator's garden where language struggle was preserved as practice, see The Translation Garden. For the isthmus of intent that this language anticipates, see The Isthmus of Intent.