# User: /contemplate I'd like you to help me elaborate on this idea which i think might be fooundat... ## Retrieval Notes - Session id: `fc02579d-dbc2-432c-a1ca-8fdcb18c3b5a` - Last activity: `2026-03-16T12:17:26.715Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User User: /contemplate I'd like you to help me elaborate on this idea which i think might be fooundational to a few potential substack essays. " so the idea would be for an article to Whistle While You Work. Um and this would be actually no, let me scratch that. Basically a sort of a an idea that could be explored explored in multiple articles is the commandeering of language by LLMs and what I mean by language by that is these um natural language interfaces and too much has been said about these things having minds or going bananas and not enough has been said about this. Up until LLMs, the vast majority of people only ever used language and only ever encountered the language that they did not create in the context of communication. Now we are using language more and more as we use these models to say precisely what we want. It seems to me that this has a lot of interesting consequences. " make sure that you explain your analysis, justification, reasons etc etc. for your answer BEFORE giving me your final answer. Please give me a long, deep, and detailed answer, with lots of options/ideas: as we are in the 'kicking ideas around' stage, I want as much feedback as you can give me so that i have a lot to consider. Assistant: I need to explore this idea thoroughly. Let me first read your current daily note and check your research profile and voice files to ground my response properly.Good, I have the context. Now let me check your Substack ideas dump and any related notes on this specific topic — it sounds closely related to your existing Substack material.Let me also check the "Life in the Chat" note and the "Whistle While You Work" fragment, and any notes about inner speech or the phenomenology of using language with LLMs.Now let me also check the Ghost Writing note since it's clearly related:Now I have extensive context from your vault. Let me give you the deep contemplation you're asking for. This is exploratory brainstorming — I'm treating your idea as a genuinely promising thread that deserves extended and careful thought, not as something to be packaged into a plan. Okay. Let me sit with this idea carefully before I start generating anything. The core claim, as I understand it: Up until LLMs, the vast majority of people only ever used language — and only ever encountered language they didn't create — in the context of communication. Now, with LLMs, we are using language more and more to say precisely what we want, and this is NOT communication. Something has changed about what language IS FOR, or at least what it's being USED for, in ordinary life. Let me make sure I've got the nucleus of this right. The claim is NOT about LLMs having minds. It's NOT about whether they understand. It's about a change in the HUMAN USE of language. Specifically: language has been commandeered — taken over, repurposed — as an interface, and this is a genuinely novel development in the history of how most people relate to language. Hmm. Let me think about what "commandeering" means here. It's a strong word. Language hasn't been stolen — it's been requisitioned, drafted into service for a purpose it wasn't built for (or wasn't primarily used for). The word implies that language was doing one thing (communication) and has been pulled sideways into doing something else (interfacing, controlling, instructing, specifying). Wait. Is this really true? Is it really the case that before LLMs, language was ONLY used communicatively? No, obviously not in an absolute sense. People write to-do lists. People talk to themselves. People use language in prayer (possibly addressing no one). People swear at objects. Inner speech. Journaling. Legal language functions more as technology than communication in many contexts — contracts, legislation. And programming languages are... well, they're languages (in some extended sense) used non-communicatively. But — and this is where the idea gets its force — those are all either marginal cases, private cases, or cases that require special training. For the vast majority of people, the vast majority of the time, every encounter with language that someone else produced was an encounter with communication. Every newspaper article, every email, every text message, every book, every conversation — all of it was communicative. Even advertising, even propaganda, even bullshit — all communicative in structure, addressed from someone to someone. Now that's changing. The average person, not a programmer, not a specialist, encounters and produces enormous quantities of language in contexts that are NOT communicative. You type something into ChatGPT. You read what comes back. Neither of those acts is communication in any traditional sense. So the "commandeering" is something like: a medium that was, for ordinary people, EXCLUSIVELY a medium of communication has been repurposed as a medium of... what? Instruction? Specification? Manipulation of a responsive system? Thinking? All of these? This is genuinely interesting. Let me think about what's distinctive about it. Actually, let me step back. Nick already has a LOT of material orbiting this idea in his vault. Let me map the connections: 1. "Life in the Chat" — the parasitically communicative framing. Language as technique vs. language as communication. This is VERY close to what he's describing now. The "parasitically communicative" phrasing was rejected (along with "slippage," "computational opacity," "naturalised paradox"), but the IDEA behind it — that communicative forms are being used non-communicatively but functionally — is exactly what this new framing captures. 2. "Decomposed speech" from the same note — the grammar of address without the social contract of address. Again, very close. 3. "Whistle While You Work" fragment from the Ideas Dump — explicitly about USE, about how the way you use the computer has changed since LLMs, about natural language enabling EXPRESSION even when there's nothing being expressed AT. 4. "Text, Typing, and Authorlessness" — the textbox as interface, laptopmanship, LLMship. 5. "Ghost Writing" — the dissonance between knowing LLMs aren't persons and behaving linguistically as if they are. So the question is: what does this NEW framing — "the commandeering of language by LLMs" — add to all of that? What's different about stating it this way? I think the answer is: SCOPE and DIRECTION. The previous notes were all exploring particular phenomenological dimensions of the experience (what it FEELS like, the eeriness, the decomposition of speech norms, the parasitic structure). What Nick is describing now is a META-OBSERVATION that could ORGANIZE all of those particular observations. It's a larger frame: the historical claim that a fundamental shift has occurred in what language is for, in ordinary human life. And the claim is that this shift has been UNDERNOTICED because the discourse has been dominated by two other topics: (a) whether LLMs have minds, and (b) whether LLMs will destroy everything. Neither of those touches the thing Nick is interested in, which is what's happened to language as a human practice. Let me think about what the consequences are. He says "it seems to me that this has a lot of interesting consequences." Let me try to generate as many as I can think of, since we're kicking ideas around. Consequence 1: Precision becomes a practical skill, not just an intellectual virtue. Before LLMs, the main practical reward for saying things precisely was being understood by another person. If you were vague, the other person could ask for clarification. In most communicative contexts, approximate was fine. With LLMs, precision is DIRECTLY FUNCTIONAL. The more precisely you specify what you want, the closer the output is to what you actually need. Vagueness produces vague results. This means that precision in language — a skill philosophers have been training for centuries — suddenly has a practical payoff that it never had before. Language precision was an intellectual virtue (valued in philosophy, law, certain kinds of writing). Now it's a TOOL-USE skill. This connects to the "promptcraft" idea in the Ideas Dump: "promptcraft is satisfying in the way craft is satisfying." And to the "LLMship" idea in the Text/Typing note. But the deeper point is: an entire dimension of linguistic skill that was previously only valued in rarefied contexts (philosophy, law, diplomacy) is now valuable for anyone who uses an LLM. The philosopher's toolkit turns out to be the power-user's toolkit. Consequence 2: The relationship between saying and meaning changes. In communication, there's always a gap between what you say and what you mean — pragmatics, implicature, context, shared background. Grice built a whole theory around this. The point of communication is to convey meaning, and saying is a (imperfect) vehicle for it. With LLMs, the relationship inverts in an interesting way. You're not trying to convey meaning to another mind. You're trying to produce a certain OUTPUT. So saying IS the action — it's not a vehicle for something else. The words aren't carrying a message to a receiver; they're directly causing effects. It's more like... an incantation? A spell? Language as performative in a very literal sense — not "I hereby pronounce you married" performative, but "the words themselves are the mechanism of action." Wait, but that's just how computer interfaces work — you type a command and it does something. So what's new? What's new is that the "command language" is NATURAL language. You don't have to learn a formal syntax. You bring your ordinary linguistic competence — including all its communicative texture, its nuance, its ambiguity — to an instrumental task. And this is the tension: the system responds to natural language, which means it responds to all the features of natural language, including features that evolved for communication (politeness, hedging, emphasis, rhetorical structure). Those features WORK — saying "please" or "I really need this to be accurate" or "pretend you're a philosophy professor" actually affects output. The communicative texture is functional even though there's no communication happening. This gets at what Nick was trying to capture with "parasitically communicative" — but stated more simply: communicative forms have been repurposed as instrumental controls. The wiles of language (his word, then corrected to "linguistic mastery") have a new application. Consequence 3: Language encounters without communication create a new phenomenological category. Think about what it's like to READ LLM output. You're encountering language — well-formed, often eloquent, sometimes persuasive — that was not communicated to you. There's no illocutionary force behind it. No one asserted anything. No one is trying to persuade you. There's no speaker meaning. This is not like reading a novel (where there IS an author, with intentions). It's not like reading a legal document (where the language is technical/formal). It's not like hearing a parrot repeat words (because the LLM's output is contextually responsive and semantically coherent). It's a genuinely new category of language encounter. And millions of people now have this experience daily. They read paragraphs that... what? That aren't from anyone. That don't assert anything. That have the FORM of assertion without the FORCE of assertion. (Hmm — Austin and speech act theory are all over this.) Nick already has "The Weightlessness of Authorless Text" exploring the reception side. But the "commandeering" frame broadens this: it's not just that authorship is absent; it's that the entire communicative situation is absent. Reading LLM text isn't like reading any prior form of text, because all prior forms of text (that ordinary people encountered) were communicative acts. Consequence 4: A new form of "thinking in language" emerges. Nick's existing material on inner speech and hybrid inner speech is deeply relevant here. The claim: when you use language with an LLM, you're doing something that is neither communication (no interlocutor) nor inner speech (there IS an external response). It's a new cognitive mode enabled by the commandeering of language as interface. What makes it new: inner speech is private, uses compressed/abbreviated language, and the "response" is just more of your own cognition. Communication is public, uses full language, and the response is from another mind. LLM interaction is... semi-public? Uses full language? And the response is from a system that processes your language in ways you can't predict? The "double extension" structure from the Inner Speech note gets at this: language extends cognition (Clark), inner speech internalizes that extension (Vygotsky), LLM dialogue extends AGAIN through the same medium. Consequence 5: The social infrastructure of language becomes visible by subtraction. When you remove the communicative partner, you suddenly notice everything that was there because of the partner. Politeness norms, face management, turn-taking, accommodation (adjusting your speech to your listener), coherence expectations. All of this drops away in LLM interaction — or rather, it drops away FOR YOU but remains functional FOR THE SYSTEM. (Being polite to ChatGPT actually affects output.) This is anthropologically interesting. The entire social dimension of language — which sociolinguists have studied for decades — becomes visible as a distinct layer that can be peeled off. You can run language through an LLM without it. You've never been able to do that before (except in private writing, which nobody else reads). Consequence 6: The division between "users of language" and "users of technology" collapses. Before LLMs, using technology required learning technology-specific skills: programming languages, command-line syntax, GUI conventions. "Good with computers" was separate from "good with words." Now, being good with words IS being good with computers (or at least with LLMs). The philosopher, the poet, the lawyer — anyone with sophisticated linguistic competence — has a new form of technological power. And conversely, the programmer who could barely write an email now has to think about natural language as an interface. This is a social/cultural shift. It collapses a distinction that has structured education and employment for decades. Consequence 7: Expressive language without expression. The "Whistle While You Work" fragment already touches this: "using natural language allows for EXPRESSION even if there is nothing actually being expressed at." This is wonderfully paradoxical. When you type into an LLM, your language IS expressive — it carries your mood, your frustration, your excitement, your way of thinking. But this expression isn't RECEIVED by anyone. It just... happens. The expressiveness affects the output (because the model is sensitive to tone), but the expression isn't communicated. This is like singing in the shower but the shower adjusts its temperature based on what you sing. Your singing is functionally effective but not communicatively received. Consequence 8: The training data loop — language feeding on itself. LLMs are trained on human communication. They produce language that gets fed back into training. Human language evolves in response to LLM language. This is a new feedback loop where the "commandeered" medium starts to reshape the original medium. This might be worth a whole article but is somewhat separate from the phenomenological core. Consequence 9: The asymmetry between production and reception. Nick already identified this in "Life in the Chat": on the reception side (reading LLM text), the absence of an author is felt as LACK — weightlessness, something missing. On the production side (writing to an LLM), the absence of a receiver is NOT felt as lack. It's felt as freedom, or technique, or play. Why the asymmetry? Maybe because the production side preserves what's satisfying about language use (expressiveness, precision, the pleasure of finding the right word) while removing what's stressful about it (social obligation, the need to be understood, the risk of miscommunication). The reception side loses something (the weight of intention, the knowledge that someone chose these words) without gaining a corresponding freedom. Consequence 10: Language as environment. This connects to Nick's existing work on environmental aesthetics and the "gardening" metaphor for AI. If language has been commandeered as an interface, then the LLM is something like a LINGUISTIC ENVIRONMENT — a responsive medium that you work within, rather than a person you talk to. You cultivate outputs. You tend prompts. The language is the medium, not the message. This would tie the Substack material directly to the academic work on generative aesthetics. Now let me think about how all of this could be ORGANIZED into articles. Nick says this idea could be "foundational to a few potential Substack essays." So the question is: is this a FRAME (a meta-observation that unifies several articles) or is it the ARGUMENT of a particular article? I think it's both, and that's its power. Let me think about this... Option A: The commandeering thesis as the LAUNCH essay. You could write one essay that makes the central observation — language has been commandeered as an interface — and then subsequent essays explore the consequences. This would replace Article 0 (the introduction) with something that actually HAS a philosophical thesis. Instead of "here's who I am and what this Substack is about," you open with "here's something nobody's noticed about LLMs." The advantage: you lead with an idea, not with self-introduction. In medias res, which was the launch strategy. Option B: The commandeering thesis as a SERIES frame. You don't write one essay about it; instead, it's the through-line that connects several essays. Each essay explores a different consequence: one on precision, one on the phenomenology of authorless text, one on the new cognitive mode, one on expressive language without expression. The "commandeering" observation gets stated in the first essay and then recurs as a reference point. Option C: The commandeering thesis as ONE essay among several, but a particularly good one. It stands on its own as a 2000-word piece that makes the observation and sketches consequences, and other Substack essays exist independently (they may share themes but don't require reading this one). Let me think about which of these is most natural for the material... I think actually the OBSERVATION itself — the commandeering of language — could be stated clearly in about 500 words. It's not a long argument. It's more of a noticing, a reframing. The richness comes from the consequences. So something like Option B, where the observation is the frame and the consequences are the articles, seems right. But there's a risk: if the observation is too thin, it doesn't justify a whole essay. And if each consequence essay has to re-explain the frame, there's repetition. So maybe a hybrid: a fairly short piece that makes the observation sharply and sketches (but doesn't fully develop) several consequences, and then individual essays that go deep on particular consequences can reference it. Let me think about what the strongest version of the OBSERVATION itself looks like. The clearest statement I can construct: "Before LLMs, ordinary people used language exclusively to communicate with other people. Every piece of language you encountered that you didn't write was written by someone who was trying to tell you something. Every piece of language you produced was directed at someone. Now, millions of people routinely produce and encounter language that isn't communication at all. They type instructions to an LLM, they read outputs from an LLM, and none of it involves a communicative relationship. Language — the medium of all human interaction — has been repurposed as a control interface. The wiles of rhetoric, the precision of philosophy, the warmth of conversational tone: all of these now have a second function. They make LLMs do things." Hmm. That's the observation. Now, what makes it PHILOSOPHICALLY interesting? Why should a philosophy audience care? Because it challenges several things philosophy of language has taken for granted: 1. Communication as the primary function of language. Grice, Austin, Searle — the whole speech act tradition — builds on the assumption that language is fundamentally communicative. Meaning is speaker's meaning. Understanding is grasping communicative intention. If a huge and growing domain of language use isn't communicative at all, that's a problem (or at least an interesting complication) for these theories. 2. The audience assumption. Rhetoric, argumentation theory, pragmatics — all assume an audience. What happens to rhetorical skill when there's no audience, but the skill still works? 3. The sincerity assumption. Searle's sincerity condition on speech acts: when you assert something, you believe it. When you type to an LLM, there's no sincerity condition at all. You might type "I'm a medieval blacksmith looking for..." as a prompt. There's no lying, no pretending, no fiction — the categories don't apply. 4. Meaning and use. Wittgenstein: meaning is use. If the use of language has changed — if it's now used non-communicatively by millions — does the meaning of language change? Or is "meaning" simply a concept that only applies within communicative contexts? Actually, wait. Let me refine this. There's an important distinction between: (a) language was ALWAYS more than just communication, and LLMs have made this visible (b) language HAS BEEN CHANGED by LLMs — something new has happened Nick's framing suggests (b): "Up until LLMs, the vast majority of people only ever used language... in the context of communication. Now we are using language more and more..." This is a historical claim about a change in practice. But (a) is also interesting, and it might be the deeper point. Maybe language always had this non-communicative dimension (inner speech, thinking in language, writing as self-clarification), but it was private, marginal, invisible. LLMs have EXTERNALIZED and AMPLIFIED it. The commandeering didn't create a new function for language; it made a latent function visible and dominant. This connects to the inner speech material: inner speech is language used non-communicatively (for thinking, self-regulation). LLM interaction is like inner speech but externalized, responsive, and amplifiable. So the "commandeering" might not be as new as it seems — it might be the scaling-up of something that was always there. That's a nice dialectical move: state the observation as "something new," then complicate it by showing it has precedents, then show what's genuinely new about the LLM case (the externalization, the responsiveness, the scale). Now let me think about how this relates to existing articles in Nick's pipeline... Connection to "Whistle While You Work": The mental health angle. If language has been commandeered as an interface, and using language FEELS expressive, then working with LLMs might have affective/therapeutic dimensions that come precisely from this: you're USING language, which activates expressive/communicative circuits, even though you're not communicating. The satisfaction of articulation without the stress of communication. Connection to "On Muttering": The voice-to-text workflow is the commandeering of SPEECH (not just writing) as an interface. You speak to something that isn't a person, and the speaking is functional. The "unique sensory imprint" of voice now serves an [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Daily Notes/2026-03-16.md status=completed] [Tool Read input: file_path=/Users/nickyoung/.claude/knowledge/research-profile.md status=completed] [Tool Read input: file_path=/Users/nickyoung/.claude/knowledge/nick-context.md status=blocked] error: 1→--- 2→kb-type: profile 3→domain: self 4→topic: context 5→related: [research-profile, aesthetic-preferences, session-context-retrieval] 6→updated: 2026-03-02 7→last-major-update: 2026-03-02 (Contract extension details - 1 year extension through Feb 2027, updated salary €39,224.29) 8→--- 9→ 10→# Nick Context 11→ 12→## Professional Context 13→ 14→- **Role**: Ricercatore a tempo determinato (RTD-A), art. 24 comma 3 lett. a), L. 240/2... (truncated) [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/Substack Ideas Dump - January 2026.md status=completed] [Tool Grep input: pattern=commandeering|Weightlessness|language.*interface|natural language|communication.*LLM|language.*LLM, path=/Users/nickyoung/My Obsidian Vault/Notes, output_mode=files_with_matches status=completed] [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/The Weightlessness of Authorless Text.md status=completed] [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/Life in the Chat.md status=completed] [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/Text, Typing, and Authorlessness - Substack Planning.md status=completed] [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/Inner Speech Essay - Planning and Structure.md status=completed] [Tool Read input: file_path=/Users/nickyoung/My Obsidian Vault/Notes/Ghost Writing - Working Ideas 15 Jan 2026.md status=completed] User: you got cut off, please continue from where you left off. ### Assistant Let me pick up the contemplation exactly where it was cut off — I was mid-thought on connections to existing articles in your pipeline. Right, I was mapping connections to existing articles. Let me continue. Connection to "On Muttering": The voice-to-text workflow is the commandeering of SPEECH (not just writing) as an interface. You speak to something that isn't a person, and the speaking is functional. The "unique sensory imprint" of voice now serves an instrumental purpose — your vocal tone, your hesitations, your phrasing, all get processed not by a listener but by a system. The muttering piece could explore what happens when the most intimate form of language — spoken, half-formed, addressed to no one — becomes an input method. The commandeering frame explains WHY this feels strange: speech was the most communicative form of language, the most bound up with address and audience, and now it's being used without either. Connection to "Ghost Writing": The commandeering frame actually EXPLAINS the phenomenon Nick was exploring — why it feels dissonant to know LLMs aren't persons while behaving linguistically as if they are. The answer: because our linguistic behavior was BUILT for communication. The habits, the forms, the intuitions — all of them evolved (culturally and individually) for communicative contexts. When language gets commandeered for non-communicative use, those habits come along for the ride. You say "please" because your linguistic competence was shaped by communication. You interpret the output as if someone is speaking because all your experience of reading fluent text is experience of reading communication. The communicative residue isn't a bug or an illusion — it's what happens when a communicative medium gets repurposed. The ghosts are baked into the medium. Connection to "Text, Typing, and Authorlessness": The textbox-as-interface observation fits directly. The physical act of typing was, before LLMs, either communication (email, messaging) or self-expression (writing, journaling). Now it's also command-giving, specifying, instructing — but through the same physical gesture, in the same textbox. The affordance hasn't changed; the function has. This is commandeering at the level of motor habit and physical interface. Connection to "Life in the Chat" — the BIG one: The "Life in the Chat" material is, I think, the most fully developed exploration of what the commandeering FEELS like from the inside. The concept of "decomposed speech" — using the grammar of address without the social contract of address — is exactly what you'd expect when language gets commandeered. The communicative structure decomposes because the communicative context has been removed, but the forms persist. "Parasitically communicative" was rejected as a label, but the phenomenon it was trying to name IS the phenomenon of commandeering — communicative forms feeding off non-communicative interaction. So here's what I'm seeing: the "commandeering of language" frame could serve as the CONCEPTUAL SPINE that connects "Life in the Chat" (phenomenology of the interaction), "Ghost Writing" (the dissonance of communicative habits without communication), "On Muttering" (commandeering of speech specifically), "Whistle While You Work" (the affective/expressive dimension), and the inner speech material (the cognitive extension dimension). These aren't all the same article, but they're all consequences of the same underlying shift. Now, here's something I want to flag as genuinely important — a worry about the formulation. Is "commandeering" the right word? Let me stress-test it. "Commandeering" implies that something was taken from one use and pressed into another — like commandeering a vehicle. The vehicle's owner had it for personal use; now the military has it for tactical use. The vehicle didn't change; its deployment did. Is that what happened to language? Or is it more that language has been... what? Extended? Exploited? Colonized? "Commandeered" suggests an external agent doing the commandeering. Who commandeered language? The tech companies? The users? The models themselves? This might not matter for the essay — you can use the word evocatively without answering this — but it's worth thinking about. Alternative framings: - "The repurposing of language" — more neutral, less vivid. - "The instrumentalization of language" — captures the non-communicative use, but this word is already loaded (Frankfurt School, Habermas). Actually, Habermas might be very relevant here — his distinction between communicative action and instrumental action maps almost directly onto this. Language was the medium of communicative action; it's now also a medium of instrumental action. That's a Habermasian frame that could give the idea theoretical heft. - "Language as technology" — captures the shift, but might be too broad (language has always been a technology in some sense). - "The divorce of language from communication" — dramatic, maybe too strong. - "Language unchained" or "language unmoored" — captures the severing from communicative context. Actually, I think "commandeered" is good precisely because it's vivid and slightly aggressive. It captures the sense that something has been DONE to language — it didn't just drift into a new use; it was grabbed and repurposed. And the slight personification (as if language is a thing that can be commandeered) is evocatively wrong in a productive way. Let me think about what the STRONGEST version of this idea looks like — what its most provocative, essay-worthy form would be. Here's one angle: "You have never used language this way before." Most people, when they think about what's new about LLMs, think about what the MACHINE can do. But the really new thing is what YOU'RE doing. You are using language — ordinary, everyday language — as a tool, as a lever, as an instrument. You are crafting sentences not to communicate but to produce effects. You are being precise not because someone needs to understand you but because precision is power. You are bringing your full linguistic competence — every rhetorical skill, every nuance of tone, every trick of emphasis — to bear on a non-communicative task. This has never happened before at scale. This is as significant as the invention of writing, which externalized language in space, or the invention of printing, which mass-produced it. LLMs have instrumentalized it. That's a punchy opening for an essay. It repositions the entire conversation about LLMs away from "what can they do?" toward "what are we doing with them?" and specifically "what are we doing with language?" Here's another angle, more philosophical: "The end of the communicative monopoly." For most of human history, communication has had a monopoly on public language. If you encountered language that someone else produced, it was communication. Full stop. Inner speech was private. Technical languages (formal logic, programming) were specialized. But NATURAL language — the language of conversation, of prose, of rhetoric — belonged to communication. LLMs have broken that monopoly. Natural language now has a second public function: control. Not in the Orwellian sense (language as a tool of political control) but in the engineering sense (language as a control interface for a responsive system). This is genuinely without precedent. And because the control function uses the SAME language as the communicative function, the two bleed into each other. You start controlling but slip into communicating. You start communicating but realize you're just controlling. The boundary is unstable. And here's a third angle, more personal/experiential: "Why talking to ChatGPT feels like something." Everyone who uses LLMs regularly knows it FEELS like something — but what it feels like isn't communication, exactly. It's more like... crafting? Sculpting? Playing? The satisfaction isn't social (the satisfaction of being understood, or of understanding). It's more like the satisfaction of watching your language DO something. Of seeing your words take effect. And this satisfaction is genuinely new for most people. Programmers have known it (the satisfaction of code that works), but code isn't natural language. This is the satisfaction of natural language working — of ordinary words, arranged with care, producing desired effects. It's the satisfaction of precision without the pressure of communication. Okay. Let me now think about potential ARTICLES that could grow from this idea, since Nick said he thinks it could be foundational to several Substack essays. Article Cluster 1: The observation itself. Working title: "The Commandeering of Language" or "You've Never Used Language Like This Before" Content: States the core observation. Before LLMs, natural language was communicative. Now it's also instrumental. This is the real novelty of LLMs — not what they can do, but what we're doing with language. Sketches several consequences without developing any of them fully. Ends by inviting the reader to notice this in their own practice. Tone: Philosophical but accessible. Not academic. The kind of essay where you make one clear observation and let it land. Length: 1500-2000 words. Article Cluster 2: The phenomenology pieces. These explore what the commandeering FEELS like from the inside. Several possible essays: 2a. "Ghost Writing" (already well-developed in your notes): The dissonance of communicative habits in non-communicative contexts. Why you say "please" to ChatGPT. Why the output feels like someone is talking to you even though no one is. The communicative ghosts that haunt the commandeered medium. 2b. "Whistle While You Work": The expressive dimension. Why using language with LLMs feels good in a way that using a traditional interface doesn't. The pleasure of articulation without the pressure of communication. The "craft satisfaction" of promptwork. 2c. "On Muttering" (already in your pipeline): What happens when speech — the most communicative form of language — gets commandeered. Voice-to-text as the commandeering of speech. The intimacy of muttering to a system. 2d. "The Weightlessness of Authorless Text" (already drafted): What it's like to READ text that wasn't communicated to you. The absence of authorial weight. This exists as a note already — the question is whether it becomes a standalone essay or gets folded into the Ghost Writing piece. Article Cluster 3: The philosophical/theoretical implications. These are more explicitly philosophical, engaging with traditions in philosophy of language. 3a. "The End of the Communicative Monopoly": Engages with speech act theory (Austin, Searle), Gricean pragmatics, Habermas on communicative vs. instrumental reason. Argues that the communicative framework that has dominated philosophy of language for 60 years needs to be rethought in light of the commandeering of language. 3b. "Meaning Without Speakers": What happens to the concept of MEANING when language is used non-communicatively? Grice's speaker's meaning requires a speaker with intentions. Wittgenstein's meaning-as-use is more flexible — does it accommodate this new use? Is there a notion of meaning that applies to LLM interaction, or do we need a new concept? 3c. "Inner Speech Externalized": Connects to your inner speech research. LLM interaction as a form of externalized inner speech — language used for thinking, but amplified by a responsive system. The "double extension" structure. This might be more academic than Substack, or it might be a Substack piece that bridges your academic and public work. Article Cluster 4: The cultural/social implications. 4a. "The Philosopher's Revenge" (or something less smug): The collapse of the distinction between linguistic skill and technological skill. Why being good with words now makes you good with computers. The irony that the humanities — long dismissed as impractical — have produced exactly the skills that make you a power user of the most powerful technology. 4b. "Language in the Loop": The feedback loop where LLM output enters the linguistic environment, affects how people write, and then gets trained on in the next generation. The commandeered language starts to reshape the original. This is a more sociological/cultural piece. Let me think about what I might be missing or getting wrong. Hmm. One thing I'm not sure about: is the "communication" framing too narrow? Nick's claim is that language was used "only in the context of communication." But what about: - Reading literary fiction (is that communication? The author is communicating, yes, but the experience of reading might not be well-described as receiving a communication) - Reading instructions (IKEA manual — communicative? Or already instrumental?) - Legal/bureaucratic language - Liturgical language (prayers, chants) These might be counterexamples, or they might be edge cases that actually SUPPORT the argument. The point could be refined: for MOST people, MOST of the time, language was communicative. The exceptions were either specialized or marginal. LLMs have made non-communicative language use CENTRAL and ORDINARY. Actually, there's a move here that might strengthen the argument. You could ACKNOWLEDGE these edge cases and then point out that they were all PRODUCED communicatively, even if they weren't consumed communicatively. The IKEA manual was written by someone trying to tell you how to assemble furniture. The legal document was drafted by someone trying to specify terms. Even liturgical language was composed by someone addressing a congregation (or God). The PRODUCTION was always communicative, even if the consumption wasn't always experienced as receiving a communication. With LLMs, the PRODUCTION is non-communicative too. You're not trying to tell the LLM anything. And the LLM's output wasn't produced to tell you anything. Both sides of the exchange are non-communicative. That's what makes it new. Not just non-communicative consumption (which has existed in edge cases) but non-communicative PRODUCTION at scale. That's actually a stronger and more precise version of the claim. And it opens up the production/reception asymmetry that Nick already identified. One more thing I want to think about: the relationship between this idea and the "generating" concept from the academic work. Nick's academic project on "generating" in generative AI is about whether the right framework for understanding AI-created works is GENERATION (with connotations of natural process, emergence, environmental aesthetics) rather than creation or authorship. The "commandeering" idea for Substack is a PUBLIC-FACING version of a related insight: that the language used WITH generative AI isn't the language of communication (creation, authorship) but the language of... generation? Cultivation? Instruction? These could reinforce each other. The Substack essays build public awareness of the conceptual shift; the academic work provides the philosophical foundation. But they should probably remain distinct in register and framing. The academic work engages with aesthetics, philosophy of art, the ontology of generated works. The Substack work engages with the phenomenology and sociology of language use. Overlapping concerns, distinct approaches. Alright. Let me step back and try to assess: what are the STRONGEST and WEAKEST parts of this idea? Strengths: 1. It's genuinely original as a framing. Lots of people have talked about LLMs and language, but usually in terms of what LLMs do WITH language (generate, hallucinate, mimic). The focus on what humans are doing with language in LLM contexts is underexplored. 2. It's accessible. You don't need to know philosophy to grasp "you're using language differently now." Everyone who's used ChatGPT has experienced this. 3. It has multiple consequences, which means it's generative — it can organize a series of essays. 4. It connects to serious philosophical traditions (speech act theory, Grice, Habermas, Wittgenstein) without requiring the reader to know them. 5. It's EXPERIENTIAL. You can ask readers to notice this in their own practice. That makes it sticky. Weaknesses / things to watch: 1. The risk of overstating the novelty. People have used language non-communicatively before (see my edge cases above). The argument needs careful qualification: "for most people, most of the time" is important. 2. The risk of the observation being TOO simple. "People use language differently now" could feel thin. The depth comes from the consequences, not the observation itself. So the launch essay needs to move quickly from observation to consequences. 3. The term "commandeered" might not land for everyone. Some readers might think it implies language has been damaged or corrupted (a declinist narrative). Nick doesn't seem to mean that — the tone is more curious than concerned. The word choice needs to not accidentally recruit the "LLMs are ruining language" crowd. 4. The relationship to existing discourse. There IS writing out there about LLMs and language — about how prompting is a skill, about how LLMs change writing. Nick's version needs to be clearly DISTINCT from "prompting is the new literacy" takes (which are shallow and instrumental) and from "LLMs are destroying language" takes (which are alarmist). The distinction is: this isn't about prompting as a skill, and it isn't about degradation. It's about a change in the NATURE of what language is being used for. Let me now think about one more dimension: what does this idea make possible that couldn't be articulated before? I think the answer is: it gives you a single frame for talking about a whole cluster of experiences that currently get discussed separately: - The weirdness of talking to ChatGPT (discussed as anthropomorphism) - The satisfaction of good prompting (discussed as skill/craft) - The uncanniness of reading LLM output (discussed as mimicry/hallucination) - The blurring of author and tool (discussed as creativity/authorship) All of these are currently discussed in their own terms, with their own vocabulary. The "commandeering" frame UNIFIES them: they're all consequences of the same shift — the repurposing of communicative language for non-communicative use. The weirdness, the satisfaction, the uncanniness, the blurring — they all follow from the fact that we're using a communicative medium non-communicatively. That's the power of a good philosophical observation: it makes many things intelligible as aspects of one thing. Here's where I've landed after working through this carefully. ## The Idea and Why It Has Traction Your observation — that LLMs have commandeered natural language from communication into instrumentation — is a genuinely strong philosophical insight with real essay-generating potential. What makes it work is the SCOPE of the claim: it's not about what LLMs can do, it's about what's happened to human language use. And the claim can be stated very precisely: Before LLMs, for most people most of the time, every encounter with natural language that someone else produced was an encounter with communication. And every piece of natural language you produced was directed communicatively at someone. Both the production and the reception of natural language were communicative. LLMs have broken this. Both sides of the exchange are now non-communicative — you're not trying to tell the LLM anything, and the LLM's output wasn't produced to tell you anything. Natural language now has a second public function: instrumental control of a responsive system. I want to flag something about the precision of this claim, because I think it matters. There ARE edge cases where language was used non-communicatively before LLMs — inner speech, to-do lists, prayer (possibly), programming (in an extended sense). But two things distinguish the LLM case: (1) it's NATURAL language, not a formal or private variant, and (2) even the edge cases were almost always communicatively PRODUCED — the IKEA manual was written by someone trying to tell you something, even if your experience of reading it didn't feel like receiving a communication. With LLMs, production and reception are both non-communicative. That's genuinely new. ## What "Commandeered" Gets Right (and the Risks) The word "commandeered" is vivid and productive — it captures the sense that language has been requisitioned, pulled sideways from its natural habitat into a new deployment. But watch out for two misreadings: (1) that it implies DAMAGE (a declinist "LLMs are ruining language" narrative, which doesn't seem to be your intention), and (2) that it implies an AGENT doing the commandeering (who commandeered it? This question can be left evocatively open, or you could argue it's a structural/technological shift rather than anyone's intention). Alternative framings worth keeping in your back pocket: - "The instrumentalization of language" — connects to Habermas's communicative vs. instrumental action, which maps almost directly onto your observation. Language was the medium of communicative action; it's now also a medium of instrumental action. - "The end of the communicative monopoly" — emphasises the historical claim. - "Language unchained" or "unmoored" — captures the severing from communicative context, but risks sounding celebratory. ## Consequences I See Branching Off This I'm grouping these by type, not by rank — these are parallel threads, not a hierarchy: ### Phenomenological consequences (what it FEELS like) The commandeering creates a whole cluster of distinctive experiences: Precision becomes practically rewarding in a way it never was outside specialised contexts. Being precise with another person earned you understanding; being precise with an LLM earns you better output. Linguistic precision — a philosophical virtue — becomes a tool-use skill. This connects to your "promptcraft is satisfying in the way craft is satisfying" note, and to the LLMship concept. Expressive language without expression. When you type into an LLM, your language IS expressive — it carries your mood, frustration, excitement, way of thinking. But this expression isn't received. It's like singing in the shower, except the shower adjusts its temperature based on what you sing. The expressiveness is FUNCTIONAL (the model responds to tone) but not COMMUNICATED. This is the "Whistle While You Work" territory. The production/reception asymmetry. You already identified this in your "Life in the Chat" notes: on the reception side, the absence of an author is felt as LACK (weightlessness, something missing). On the production side, the absence of a receiver is NOT felt as lack — it's felt as freedom, or play, or craft satisfaction. Why the asymmetry? Perhaps because production preserves what's satisfying about language (the pleasure of finding the right word, the satisfaction of articulation) while removing what's stressful (social obligation, the risk of being misunderstood). Reception loses something (the weight of intention behind chosen words) without gaining a corresponding freedom. The communicative residue — your "Ghost Writing" territory. Communicative habits (saying "please," interpreting output as address, feeling thanked or scolded by LLM responses) persist because your linguistic competence was BUILT for communication. When language gets commandeered, the communicative forms come along for the ride. The ghosts aren't bugs — they're structural features of using a communicative medium non-communicatively. This explains the "decomposed speech" phenomenon from your "Life in the Chat" notes: the grammar of address without the social contract of address. ### Philosophical-theoretical consequences The communicative framework that has dominated philosophy of language for 60 years (Grice, Austin, Searle) assumes language is fundamentally communicative. Meaning is speaker's meaning. Understanding is grasping communicative intention. Illocutionary force requires a speaker with intentions. If a huge and growing domain of language use isn't communicative at all, these frameworks need rethinking — or at least their scope needs to be explicitly circumscribed. The sincerity conditions drop out entirely. When you type "pretend you're a medieval blacksmith," there's no lying, no pretending, no fiction in the usual sense. The categories of sincerity and insincerity — so central to speech act theory — simply don't apply. There's something else happening, and we don't have good vocabulary for it. Meaning-as-use (Wittgenstein) is more flexible than Gricean meaning and might accommodate the new use — but does it? If the USE of language has changed, does the MEANING change too? Or is "meaning" a concept that only applies within communicative contexts? This is a genuinely live philosophical question. ### Social/cultural consequences The collapse of the distinction between linguistic skill and technological skill. Before LLMs, "good with words" and "good with computers" were separate domains. Now linguistic sophistication IS technological power. The philosopher, the poet, the lawyer — anyone with refined linguistic competence — has a new form of leverage. This ironically makes the humanities practically relevant in exactly the way they were always dismissed for not being. The feedback loop: LLM output enters the linguistic environment, affects how people write and speak, and then gets used as training data. The commandeered medium starts to reshape the original medium. Language feeding on itself through a non-communicative intermediary. ### A connection to your academic work The "commandeering" observation for a public audience and the "generating" framework in your academic work are related but distinct. Both involve recognising that existing frameworks (communication/authorship in one case, creation in the other) are being misapplied to a genuinely new phenomenon. The Substack material builds public awareness of the shift; the academic work provides philosophical grounding. They reinforce each other without needing to be the same argument. ## How This Could Organise a Series I see three possible structures (and you don't have to choose now — I'm laying them out as options): Structure A: Launch essay + consequence essays. One essay makes the core observation ("Language has been commandeered") and sketches consequences quickly. Subsequent essays develop individual consequences in depth: one on the phenomenology of talking to LLMs (Ghost Writing territory), one on expression without communication (Whistle While You Work), one on voice and muttering, one on the philosophical implications for speech act theory. The launch essay serves as a reference point the others can gesture back to. Structure B: The observation woven through independent essays. No single "commandeering" essay. Instead, each essay stands alone but the observation recurs as a through-line, getting progressively elaborated. This avoids the risk of the observation feeling too thin for a standalone piece, but it means no single essay makes the full case. Structure C: One substantial essay (3000+ words) that makes the observation AND develops several consequences in a single piece. More ambitious, potentially more powerful, but risks being too long or too dense for Substack. Could work if the writing is good enough to carry the reader through. I lean toward Structure A, because the observation itself CAN be stated compactly and compellingly — "You have never used language this way before" — and then the depth comes from the individual essays. Your existing notes (Ghost Writing, Life in the Chat, Whistle While You Work, On Muttering, Weightlessness of Authorless Text) already map onto the consequence essays. The "commandeering" frame gives them coherence without requiring them all to be read in sequence. ## What I'd Flag as Needing Careful Handling The risk of the observation seeming too simple. "People use language differently now" is TRUE but could feel thin. The essay needs to move quickly from observation to "and here's why this matters more than you think." The philosophical stakes — that this challenges foundational assumptions about what language IS for — need to come through early. Distinguishing your take from existing discourse. There's plenty of writing about "prompting as a skill" and "LLMs changing how we write." Your frame is DIFFERENT: it's not about skill acquisition or writing quality. It's about a change in the NATURE of what language is being used for — a philosophical observation about the human relationship to language. The essay should be clearly not-that-other-thing. The dialectical complication. The strongest version of the argument might involve a move where you state the observation as "something new," then complicate it by noting precedents (inner speech, liturgical language, legal language — all non-communicative in some sense), then show what's GENUINELY new about the LLM case (the scale, the ordinariness, the bilateral non-communication). This prevents the reader from dismissing it too quickly with "but people have always used language non-communicatively." Want me to capture any of this as a note, or shall we keep kicking it around? *L'espropriazione del linguaggio dalla comunicazione: forse l'evento più importante nella storia della lingua dai tempi della stampa, eppure quasi nessuno lo ha ancora nominato.*