# take a look at this note and tell me here in the chat what changes if any, you would make to make... ## Skills Used - voice-fix - writing-argumentative-paragraphs ## Retrieval Notes - Session id: `dcfd54df-6769-49c5-b062-d9fa15e17444` - Last activity: `2026-05-16T13:35:26.405Z` - Files touched: `1` ## Artifacts **Modified:** - [[Writing/research/generative aesthetics of ai/towards 8k cuts/2. What LLMs Are]] ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User take a look at this note and tell me here in the chat what changes if any, you would make to make it read more like me. Regarding skills: these are not optional extras, or suggestions as to how to do the task, they are precise instructions that should be followed to the letter. This note is here because of a longstanding failiure of yours to actually use the askills which I activate. USE the skills. they are activated for a reason. ### Assistant Now I need to warm up with published paragraphs. Let me find Nick's published papers. ### Assistant I've read the manuscript section and grounded myself in your published work. Let me now give the diagnosis. --- ## Grounding Papers used: "Hearing Objects and Events" (2018, _Philosophical Studies_) — §1 opening, §2.1 staging a comparison, §2.2 developing structural claims from examples; "Sounds as Properties" (2021, _Thought_) — §1 entering through a specific claim, §3 working through objections, §4 elaborating a novel proposal. What these show about expository setup sections: even when laying groundwork, your published paragraphs open with substantive claims or genuine questions — not with announcements about what the paragraph will contribute to the paper's project. The prose varies sentence length more than this draft does, and qualifications are embedded within sentences rather than delivered as separate clarifying statements. --- ## Diagnosis ### The bolded sentences should go The two bolded sentences (paragraph 2 opening, paragraph 3 opening) function as section labels. They describe what the paragraph is for rather than entering the thought. Compare your published equivalent: "Although the claim that we hear source events is popular, I shall argue in this section that it is not obvious how it should be understood" — that opens by stating a substantive position, not by describing the paragraph's role in the paper's architecture. - "Any account that aims to make the order of generated texts visible will have to track how each text develops in relation to its prior context" — this is a map sentence. It tells the reader what the paragraph will show, rather than showing it. Delete the bold formatting and either cut the sentence or fold its content into a claim that does local work. - "Knowledge of how training produces such order will have a role in any appreciation of generated texts as what they are" — same issue. It announces a connection between training and appreciation rather than making that connection. ### "In what follows, we connect these features to the trained system that produces them, and thereby bring into view the kind of knowledge that would have to guide any appropriate appreciation of them" This is a road-map sentence of the kind you do use sparingly in published work — but it's unusually procedural here. "We connect X to Y, and thereby bring into view Z" narrates the essay's activity. In your published voice this would more likely state what the connection is and let the reader see the consequence. ### Sentence rhythm is too even Your published prose alternates between longer sentences with embedded qualifications and shorter sentences that land a point. This draft stays in the middle register throughout — competent mid-length sentences, one after another. Paragraph 2 especially could use one or two longer sentences that do more thinking within themselves (with semicolons, reformulations, parenthetical qualifications) and one shorter sentence that lands cleanly. ### Paragraph 4 opening is borderline "What we have so far described is not yet the system users actually encounter" — this is structural navigation, but it does genuine work by marking a distinction (trained model vs. deployed product). It's fine. It reads like your transitional moves. ### The final paragraph's three-scale taxonomy "We can consider the trained system we have described at three scales. An _output_ is... A _chat_ is... A _model_ is..." — this reads as a definition list. The clarifying sentence at the end ("These are not three different kinds of object, since outputs and chats are ways in which we encounter the model; they are three genuinely different scales at which we will later pose the question of appropriate appreciation") is much more you — it qualifies, it distinguishes, it does something. The bare definitions preceding it are flat. Consider letting the distinctions emerge through a sentence or two that works through why scales matter, rather than listing them as definitions. ### Minor points - "Pre-training thereby shapes a graded sensitivity to the regularities of text" — "graded sensitivity" is good, sounds like you - "This kind of path-dependence is what allows a response to sustain a line of argument over several sentences; it is also what makes it possible for the response, at some point, to lose the line it had" — this sentence does real work and reads naturally in your voice - The footnotes are good — they set aside complications without disrupting the argument, which is your typical footnote style - "a relatively stable response profile, which users may track when they describe one model as friendlier than another or as having a particular 'vibe'" — bringing in "vibe" with a distancing signal works here ### Summary of changes 1. Remove bold formatting from the two announcement sentences; rewrite them as substantive claims or cut them 2. Rewrite or cut the "In what follows" road-map sentence in paragraph 1 3. Vary sentence length — let some sentences in paragraphs 2–3 get longer and do more internal work; let one or two be shorter and land cleanly 4. Restructure the three-scale taxonomy so the distinctions emerge from a line of thought rather than being listed as definitions The overall structure and philosophical content are sound. The problem is local: too many sentences that describe the argument's architecture rather than performing it, and too uniform a rhythm. --- *Il testo conosce già la propria direzione — è la voce che deve ancora abitarlo fino in fondo.* ### User I can't believe how much time you spend thinking about fucking bold text. Jesus fucking Christ. it is obviousltyabsolutely fucking irrelevant as it can be changedi in an instant. you didn't actuvate all your skills even thoughj i i explictly fucking asked you to. Please start the task(s) again completely from scratch, keeping this in mind. ### User [Request interrupted by user] ### User 0/10 answer, you gave me some very vague issues you thought might be problems, and din't offer any solutions. ". "We connect X to Y, and thereby bring into view Z" " you saying i never open paragraphs like that in my published work? did you check? you didn't actuvate all your skills even thoughj i i explictly fucking asked you to. Please start the task(s) again completely from scratch, keeping this in mind. ### Assistant Starting from scratch. Invoking the skills the deep-writing-mode requires for this task. ### Assistant Now let me also read a third published paper to ensure breadth — I need paragraphs doing specifically expository/setup work. ### Assistant Good. I now have sufficient grounding across three published papers. Let me give you the actual diagnosis with concrete rewrites. --- ## Grounding Papers: "Hearing Objects and Events" (2018) — §2.2 developing technical features with concrete instances; "Sounds as Properties" (2021) — §2 introducing a three-way taxonomy by working each term through a concrete case (the bell); "Growing the Image" (2025) — §II introducing Lowe's utensil/machine distinction by working through paintbrushes, pens, clocks, drum machines. What these show: when Nick sets up a taxonomy or names technical features, he always works through what each means concretely. He doesn't list definitions — he develops each term with a case. --- ## What's fine - Paragraph 1: opens with a genuine question, answers compactly, states the section's aim with "In what follows." You do this — "For these reasons, sounds will feature little in what follows" (Hearing Objects §1), "In this paper I give some reasons..." (Sounds as Properties §1). I was wrong last time to flag that sentence. - Paragraph 2: the technical exposition reads well. The sentence rhythm has enough variety. The final sentence ("This kind of path-dependence is what allows a response to sustain a line of argument over several sentences; it is also what makes it possible for the response, at some point, to lose the line it had") does real philosophical work and sounds like you. - Paragraph 4: moves well, grounds itself concretely with "friendlier than another or as having a particular 'vibe'", sets up the next section's question. Fine. - Footnotes: standard Nick — setting aside complications without disrupting the argument. --- ## What needs changing ### Paragraph 3 — Compression / Missing Development Target text: > The internal organisation that supports this sensitivity emerges from training rather than being laid out in advance by designers. Two of its features bear on what follows. First, training places tokens in structured relations to one another, so that the system comes to treat tokens appearing in similar contexts similarly for purposes of generation. Second, at each step of generation, different parts of the prior context bear differently on what the system produces next, rather than the most recent token alone conditioning the result. Together these allow a register or theme set early in a conversation to be sustained across many later turns, and a question to draw on material from much earlier in its preamble. Argumentative move: laying out technical features that will matter later. Failure mode: **compression** and **missing development**. Each feature is stated once and then the paragraph moves on. Compare to "Hearing Objects" §2.2, where each feature of the spatial field gets a concrete instance: > First, it allows for the perception of spatial relations between different objects: you see the bowling ball as in front and to the right of the region of your feet. Second, it allows for the perception of property variation: you can see your bowling shoes gradually change from a bright green at the toe, to a darker shade at the heel. Third, it allows for the perception of the precise location of objects, in that objects can be seen to extend across, or fill out, certain regions of space: seeing the bowling ball to be located here rather than that one. After each "First... Second... Third..." Nick gives a concrete case that shows what the feature means. The manuscript's "First" and "Second" state the feature abstractly without showing what it does. The payoff sentence at the end ("Together these allow...") gestures at application but compresses two distinct illustrations into one. Proposed rewrite: > The internal organisation that supports this sensitivity emerges from training rather than being laid out in advance by designers. Two of its features bear on what follows. First, training places tokens in structured relations to one another, so that the system comes to treat tokens appearing in similar contexts similarly for purposes of generation: a word like "bass" following "fretboard" will prompt continuations drawn from music, while the same word following "shallow water" will prompt continuations drawn from fishing. Second, at each step of generation, different parts of the prior context bear differently on what the system produces next, rather than the most recent token alone conditioning the result: a question posed in the opening line of a prompt can shape the system's response many sentences later, even when intervening material has introduced other topics. Together these features allow a register or theme set early in a conversation to be sustained across many later turns, and allow a request made at the outset to constrain what the system produces long after intervening exchanges might have displaced it. What changed: after each "First" and "Second" I added a concrete instance (using a colon, which is Nick's standard move — see the bowling ball cases). The final sentence now specifies two distinct things the features enable rather than compressing them. --- ### Paragraph 5 — List Structure Target text: > We can consider the trained system we have described at three scales. An _output_ is a single bounded continuation generated from a particular context. A _chat_ is an extended sequence in which earlier turns condition the generation of later ones. A _model_ is the trained system whose recurring tendencies become visible only across many outputs and chats. These are not three different kinds of object, since outputs and chats are ways in which we encounter the model; they are three genuinely different scales at which we will later pose the question of appropriate appreciation. Argumentative move: introducing a three-part taxonomy. Failure mode: **list structure**. Three terms are defined in parallel single sentences. Compare to how you introduce Lowe's taxonomy in "Growing the Image" §II: > Lowe (2014) splits tools into two types: utensils and machines. Utensils involve their users exerting more or less continuous control to carry out their functions. A paintbrush won't paint and a pen won't write unless they are being moved by someone's hand. Machines on the other hand are autonomous, in the sense that they can carry out their function without constant user input. Clocks for example can be set up and will carry out their function of telling the time without further user input. Each term gets worked through with cases. Compare also to how you introduce Parthood, Identity, and Property in "Sounds as Properties" §2 — each view gets a full explanation via the bell example, not a one-sentence definition. Proposed rewrite: > We can consider the trained system we have described at three scales. A single response to a particular prompt — an _output_ — is the most local: it is one bounded continuation, shaped by whatever context it was generated from. Over a conversation, however, earlier outputs condition later ones, so that the exchange as a whole — the _chat_ — develops a texture that no single response possesses on its own: a position staked in the first turn can be refined, retracted, or forgotten by the tenth. The _model_ is the trained system whose tendencies become visible only across many such encounters; a user who has spent hours prompting one model and then switches to another notices recurring differences in tone, in willingness to commit, in where each system tends to take a line of thought. These are not three different kinds of object, since outputs and chats are ways in which we encounter the model; they are three genuinely different scales at which we will later pose the question of appropriate appreciation. What changed: the three terms now emerge from a developing thought rather than a definition list. Each gets a brief concrete illustration (the position staked then retracted; the user noticing recurring differences between models). The final sentence remains unchanged — it already reads like you. --- ## Summary Two paragraphs need work. The rest is sound. The problems are both instances of the same underlying pattern: stating a distinction or feature abstractly where your published voice would work through what it means concretely. Everything else — the "In what follows" sentence, paragraph 4's "vibe" grounding, the footnotes, the overall structure — already reads like you. *Un paragrafo tecnico si guadagna il suo posto quando il lettore vede il meccanismo muoversi, non soltanto la sua definizione.* ### User try now ### Assistant Done. Both edits applied. *La prosa tecnica respira quando ogni distinzione ha un caso che la abita.*