**You** "Responses like those described above are just what this should lead us to expect." not a good sentence, not how i would write. "and the expected response to an examination question is an even-handed exposition of the going positions, not a case for one of them." speculative. falsely confident that this is what is happening. "The assistant training presses the same way:" not how i write at all. get rid of the stupid fucking pivot paragraph. all of the abopve should lead to u updating the canvas. Then I need us to examine the following paragraph, it is bloated and badly written, can you breakdown why it is so shit with me here in the chat, along with how we can save it. as always, use my publications as a way of knowing what is wrong and what is right structrually, content wise, style wise. A lot of work needs doing. remmebedr, your instincts are not great on this, so you need to use my work carefully "A thought experiment, we argued in Section 3, evokes: in setting out its scenario it opens a structure with rigid properties, so that there are facts about what the scenario commits us to which nobody has chosen and which outrun anything actually stated. A prompt that articulates a position and the pressures upon it does the same. Once it is written, there are facts about what the position can and cannot say in reply, and these are no more up to the prompter than what follows from the rules of chess is up to anyone. Now consider what the philosophical portion of a model's training data consists in. On Pigliucci's account it is, in large part, the development of evoked structures. Jackson sets out Mary's room in two paragraphs, and four decades of replies work out what the scenario does and does not commit us to — whether what Mary gains is knowledge of a fact, whether physicalism survives her release — with each reply answering to the structure Jackson opened. In writing of this kind, what can be said next is constrained by more than grammar and sense: once it has been granted that Mary knows every physical fact about colour vision, a reply cannot go on to have her ignorant of one, and the literature respects closures of this kind on every page. A model trained on that literature picks up these constraints in the same way that, as we saw in Section 2, it picks up the constraints of grammar and of sense: as regularities in what may follow what. Its continuation of an evoking prompt is therefore produced under the kind of constraint the prompt lays down, and can be assessed against the structure the prompt opened: it states what holds there, or it misstates it, and there is a fact of the matter either way." oh god it is so fucking bad. --- **Claude** The three fixes are on the canvas, and the pivot paragraph is confirmed gone — it was deleted last turn; the single remaining hit for "tells us nothing either way" is the folded sentence now closing the mechanism paragraph, which is where you put its one surviving job. The fixes: the expectation sentence is deleted outright (the paragraph runs from the two-trainings sentence straight into "In writing at large..."); the examination sentence is now a corpus-location claim we're actually entitled to — "'Is representationalism the correct theory of perception?' is the sort of sentence that appears in examination papers and essay titles rather than at the opening of a defence of representationalism" — rather than a speculation about expected responses; and the connective is now "A similar point applies to the assistant training", which is your form ("Similarly, we might ask..." in *Agents of Change*). Now the autopsy of the junction paragraph. Measured against your publications, it fails in six ways, and they compound: **1\. It does three paragraphs' work in one.** Your rule is one function per paragraph; this has three functions stacked — prompts evoke (sentences 1–3), the philosophical corpus consists of developments of evoked structures (4–6), the model internalises the corpus's constraints (7–8). In *Agents of Change*, when an argument needs three moves of this size, each gets a paragraph and each paragraph opens from the previous one's payload. Here the reader gets a wall. **2\. The machinery arrives labelled instead of assembled.** "Now consider what the philosophical portion of a model's training data consists in" — that's a grid being executed, a slide-change in prose. Your papers make the reader watch the machinery come together out of a case; compare how §3 of *Agents* reaches the non-perceptual-experience proposal only *after* the central-heating-hum case has made the gap felt. Here the case (Mary) arrives after the thesis it's meant to motivate. **3\. The one good concrete moment is buried and then overclaimed.** "Once it has been granted that Mary knows every physical fact about colour vision, a reply cannot go on to have her ignorant of one" — that is the paragraph's genuine cargo: a specific, checkable closure. But it's smothered mid-paragraph, and immediately inflated by "the literature respects closures of this kind on every page" — "on every page" is the same false confidence as the examination sentence you just caught, and "closures" is a coined label the dialectic hasn't earned (your diagnostic's point about labels exactly). **4\. The load-bearing claim is asserted, not argued.** "A model trained on that literature picks up these constraints in the same way that... it picks up the constraints of grammar and of sense" — *in the same way* is doing all the work, and nothing pays for it. Your modal profile for a claim at this altitude would be "there is reason to expect", "we might expect", "there is no obvious reason to think it would not" — the *Agents* register ("I want to be careful here and distinguish the suggestion I am making from a stronger one"). The paragraph makes the strongest version of a claim where the weaker version does all the argumentative work the section needs. **5\. Recycled furniture.** The chess clause is the comparison's third appearance in the paper (§3 twice, here again), riding along as a flourish rather than adding anything the §3 back-reference doesn't already carry. And the "four decades of replies work out what the scenario does and does not commit us to" sentence is potted literature-history at high altitude — your example-use slows one case down until it generates the distinction; it doesn't survey a debate in a subordinate clause. **6\. It clicks shut.** "It states what holds there, or it misstates it, and there is a fact of the matter either way" — a disjunctive snap, the paragraph ending on its underlineable sentence. The *content* (there is something for the continuation to be right or wrong about) is needed; the *cadence* is the tell. **The rescue.** Split into two paragraphs, each with one function, and let Mary do the assembling rather than illustrate a finished claim. Paragraph one: what an articulated prompt opens. One back-reference sentence to §3's evocation (no chess), then straight to the worked case at walking pace — what Jackson's two paragraphs fix, and what a reply to them may and may not do: it may deny that Mary learns a fact, it may not have her ignorant of a physical one — so that "there are facts about what the position can and cannot say in reply" is *shown* on the case before being stated, and the paragraph's payload is that a continuation of such a prompt is the kind of text that can get this right or wrong. Paragraph two: why expect a model's continuations to operate under that constraint — the philosophical writing it is trained on consists of texts that observe these commitments (the replies to Jackson being the case already on the table), and, as Section 2 argued for grammar and sense, the model acquires the regularities of the writing it is trained on; so there is reason to expect — not "it picks up in the same way" — its continuations to answer to what the prompt fixes. The fact-of-the-matter point survives as one plain sentence, mid-paragraph rather than as the door slamming. If that shape is right, I'll draft the two paragraphs against the publications and bring them here before touching the canvas. --- **You** "and let Mary do the assembling rather than illustrate a finished claim." I don't really see why the phenomenology example or the thought experiment has to be the one we're using here. It doesn't strike me as obviously the correct one to use, and it's going to confuse the reader because they're going to be saying, "Well, why are you talking about this specific type of continuation in philosophy?" They might think that means we're only talking about phenomenology or thought experiment–based stuff. So yeah, that's a complete fuck-up on your part. "A prompt, on Wolfram's account, is not a question awaiting an answer but the opening stretch of a text the model will extend. Everything the prompt contains is material for what follows: tell one of these systems something once, Wolfram observes, and it is made use of thereafter (2023), so that a prompt fixes what there is to continue. A model given only a question will continue a question, while a model given a position, together with the difficulties any defence of it would have to meet, has been given the beginnings of a philosophical text, and what would count as a reasonable continuation is a development of that position under those difficulties." i should have mentioned that this paragraph also seems pretty useless at least in its current form. basically we have now fixed everything up until the end of para 2, but the rest of the section needs to be comleltelty reworked out. –it is just a mess, and it is so bad a mess i can't even articualte how to fix it. all i would say is that earlier in this chat we have definitley worked out exactly what should be said here. a combination of piglucci's evocation and wolfram stuff. But be specific and careful. Please try harder than you have been? i am sick of you just giving me inelegant bloat despite us going through the actual arguments carefully yesterday (in this chat). --- **Claude** The rebuild is on the canvas, and the register checks out against the sources just read — the question-opener follows "If Midjourney is a tool, what sort of tool is it?", and the hedged internalisation claim follows "If there is no obvious reason to believe that events but not objects instantiate sounds..." The two paragraphs: > What, then, would it be to prompt these systems properly? A prompt, on Wolfram's account, is not a question awaiting an answer: it is the opening of a text that the model will extend, and everything it contains is made use of in the continuation that follows (2023). A prompt can therefore do what we have just seen journal articles do. One that advances a position, and sets out the objections that any defence of it would have to meet, hands the model the beginning of the kind of text that philosophical argument continues, and its reasonable continuation is a development of that position under those objections. > A prompt of this kind is also an articulated starting point in the sense of Section 3. In writing down a position and the constraints under which it is to be defended, the prompter makes starting assumptions of just the kind that, as we argued there, open a structure with rigid properties: there are facts about what the position is committed to, and about which replies are available to it, which nobody has chosen and which outrun anything the prompt states. The philosophical writing these systems are trained on is itself produced within such structures, its replies and developments answering to what some articulated starting point commits its defenders to. As we saw in Section 2, a model acquires the regularities of the writing it is trained on, and there is no obvious reason to think that the regularities of writing of this kind would be an exception. A continuation of an evoking prompt would therefore have what the continuation of a straight question lacks: a structure to answer to. Whether it states what the position is committed to, or misstates it, is not up to the model, nor to the prompter. What changed and why. Mary is gone entirely, and nothing localising replaces her: the concreteness now comes from ¶2's own journal-article characterisation, which is the general case — positions advanced, objections raised — so the account visibly covers philosophical continuation as such, not thought-experiment philosophy. That's also what rescues the previously useless prompt paragraph: its cargo (tell-it-once, prompt-fixes-continuation) now hooks onto something, namely "what we have just seen journal articles do", instead of floating. The first paragraph is the Wolfram half — a prompt of the right kind is the beginning of the kind of text argument continues; the second is the Pigliucci half and the junction — the same prompt is an articulated starting point, it evokes, and the writing the model is trained on was itself produced under the constraints such structures impose, so by §2's own acquisition point there is no obvious reason to expect its continuations to escape them. The internalisation claim is now made at the modal strength the section needs and no more. "Now consider...", the chess clause, "on every page", "closures", the four-decades potted history, and the fact-of-the-matter snap are all gone; the answerability point ends as two plain sentences. The openness paragraph got the one trim ("an empirical question which we shall not try to settle") and otherwise stands, as does the handoff. The section now runs: challenge — failure of prompting, both trainings — what proper prompting would be (Wolfram) — why its continuations would have something to answer to (Pigliucci, junction) — form left open — where capacity would be decided, and the credit question for §6. --- **You** Update the canvas with those two paragraphs so I can see them in context before making any further decisions. --- **Claude** The two paragraphs are already on the canvas — I wrote them there in the last step, in place of the old prompt and junction paragraphs. Opening §5 now to show them in their surroundings so you can read them in context. --- **You** "One thing worth your eye before you decide anything: ¶5's "A question put point blank, however, evokes nothing" and ¶6's opening "It is in the continuations of evoking prompts... point-blank questions provide none of them" restate the same contrast in consecutive paragraphs. Not a repetition within a paragraph, but the seam between them is slightly soft. I've left it untouched pending your read — flagging it rather than acting." yeah this should be fixed. also 'evokes nothing' is your usual journalistic wank. I'm going to go back to the next one. I'm not sure what I'm saying. I wouldn't want to defend because it's an extraordinary blend. It also doesn't really stick to what has been said about Piluchi in that section. Why not talk about what was actually said and what Piluchi actually said? I mean, even in a re-quote Piluchi, but here you just talked around him rather than really talked about it properly. So, staying on this paragraph again. Let me think about a better way these things could be first. It should be something on the line to the following. I should find what line sounds like So I mean I guess simply in a way you ask more clearly for it to actually reason and to use the sorts of abductive reasoning that was mentioned in section 2 as well. It's that same abductive reasoning which is being used to revoke in the police sense. So it needs to be much more about that rather than always fucking nonsense you're writing about the philosophical writing that these texts are trained on. Don't talk specifically about that. Talk about the abductive work, the evocative work. You're not doing that right now. You can be more blunt at the beginning of this sort of paragraph as well. the beginning of this sort of paragraph as well, and you can say, we suggest that if we are to elicit philosophy worth reading from these systems, we need to think in terms of how it can be evoked, or how prompt might evoke such texts. Okay, and then you bring it back. Okay, so you don't have to be super specific, so you could get a couple of examples. You could sort of be very specific about whether the explanation... Yeah, I mean, just explicitly explicitly trying to get it to reason, explicitly ask for it to do the argument in particular ways, rather than just asking a straight question. But again, I'm scared you're going to say something kind of blunder and silly at this point. So there's a certain delight to touch, trying to get it to get it to you. But yeah, we're not going to be very far unless you actually do the fucking work, re-read Pellucci again, re-read Wolfram again, look at why I said in section 2, look at why I said in section 3 and perhaps section 4, and then don't put all this on the page, but try and work out what the distilled idea would be here. Because it seems to me you're really not trying very hard to do that right now, you're just trying to hope, you're kind of playing Will this do, will this do And so far none of it will be It's all been shit And in case it wasn't obvious already, given that we're still so far away from anything even remotely good, don't think you need to retain the paragraphs in anything like their current form for this latter half of this section. Okay? I'm still fairly happy with paragraph one and paragraph two. The rest of what you've given me is still absolute dogshit. So, don't constrain yourself by thinking you have to keep these four paragraphs alive in some sense. You can do whatever you want as long as you actually do the work properly this time. --- We argue that current-generation large language models (LLMs) are capable of producing philosophical texts that are worth reading, in the same way that a good piece of human-written philosophy can be. We defend this claim against various challenges. First, one might think that a text produced by an LLM cannot count as a work of philosophy because no person lies behind it, just as one might think that an image produced by AI cannot count as an artwork. Second, one might think that LLMs lack a particular capacity, or capacities, that producing worthwhile philosophy requires: the capacity to weigh rival explanations, to connect with the world, or to draw on experience. Finally, one might ask why, if we are right, no philosophy worth reading has yet come from these systems, and whether the philosophical content of any text produced under a philosopher's prompting should be credited to the philosopher rather than to the model. We argue that none of these challenges succeeds. Two texts containing the same argument cannot differ in philosophical merit; the distinction between genuine and merely apparent abduction cannot be located in the texts LLMs actually produce; and the absence, so far, of worthwhile LLM-written philosophy reflects how these systems are used rather than what they can produce. --- ## 0\. Introduction The last decade or so has seen the rise of generative artificial intelligence: systems that produce text, images, code, music, video, and other outputs in response to prompts. AI has had success in domains where the value of an output is not exhausted by its superficial fluency. There have been recent AI-assisted discoveries in physics (Guevara et al. 2026), mathematics (Novikov et al. 2025), biomedicine (Gottweis et al. 2025), and materials science (Zeni et al. 2025). In this paper we argue that we should expect similar success in philosophy. Specifically, we argue that current-generation LLMs are capable of producing philosophical texts that are *worth reading*. We do not want to begin by settling what counts as *good* philosophy. Instead, we appeal to a distinction familiar to anyone who reads philosophy: between texts that repay the time spent reading them and texts that do not. As you begin reading this article, you likely hope that it is worth reading, in the sense that the time spent reading it will not be wasted. When you write a philosophical text yourself you aim to make it worth readers' while to read it, and whether or not the journal you send it to accepts it depends on whether or not they agree. Note that a text's being worth reading is not the same as its being correct: we take many of the philosophical texts we read to be mistaken in their conclusions, and few of them to have wasted our time. Note also that the mere statement of a philosophical conclusion is unlikely to be worth reading. A text consisting only of bare pronouncements — that direct realism is correct, that we should be utilitarians — would not repay anyone's attention, and this is as true of a human philosopher's pronouncements as of an LLM's.[^1] What would repay attention is argument, and it is the capacity of LLMs to produce philosophical arguments, rather than philosophical pronouncements, that concerns us in what follows. We will argue for this claim by considering six challenges that might be raised against it. Section 1 rejects the challenge from authorship: the claim that an LLM output cannot be philosophy worth reading because no philosopher lies behind it. Section 2 turns to abduction and argues that the absence of human-style inference to the best explanation in the producer does not preclude abductive structure in the product. Section 3 turns to the challenge from detachment, and argues that the materials philosophy takes from the world reach it already set down in words, which an LLM can work on as any philosopher does. Section 4 takes up the challenge from experience, and argues that the experiences philosophy argues about reach it in the same way. Section 5 turns to the challenge from observation — that if all this is right, philosophy worth reading should already be coming from these systems, and is not — and argues that this reflects how they are used rather than what they can produce. Section 6 addresses the challenge from instrumentality — that where a philosopher's prompting draws out such a text, the philosophy is the philosopher's — and argues that a prompt articulates a starting point whose development it does not fix. --- ## 1\. The Challenge from Authorship In this section we address what we might call the *challenge from authorship*: the idea that philosophy is something that only persons, or at least minds, can produce. This view has not, to our knowledge, been explicitly defended in just this form, but it gives shape to an intuition that many philosophers may have: philosophy is a person-only domain. An imperfect comparison is with art: One might deny that an image generated by an AI system is an artwork, because no artist exercises the relevant kind of intentional control over its production.[^1] One might think, for similar reasons, that philosophy can only be done by people: no text produced by an LLM can be a work of philosophy, because no philosopher lies behind it. There is some initial support for this thought in the way philosophy is studied. Like the study of art, and unlike the study of physics, the study of philosophy is organised around individuals: undergraduates take courses on Kant's ethics or Lewis's metaphysics, and reading an important philosopher's own words is held to be of value in a way that reading Newton's is not. A physics student is taught Newtonian mechanics from a current textbook, and the course loses nothing if the *Principia* is never opened; a course on Kant's ethics that never opened the *Groundwork* would scarcely count as one. In the sciences, that is, what a text contributes can be carried entirely by other texts, while in philosophy the contribution and its original presentation are harder to prise apart. One might take this as evidence that a philosophical work is bound to the activity of the person who produced it, in a way that a scientific result is not. The challenge becomes more precise if we ask how far Davies' *performance* theory of art transposes to philosophy. Davies writes: > \[T\]he work – what the artist achieves – is the process eventuating in that product. Works themselves are neither structures nor objects simpliciter, nor are they contextualized structures or objects. They are, rather, intentionally guided generative performances that eventuate in contextualized structures or objects (or events, as we shall see) – performances completed by what I am terming a focus of appreciation. (2004, p. 97) On Davies' view, when a painter paints a picture, the canvas is what we attend to, but it is not the work. The work is the artist's intentionally guided activity in producing the canvas; the canvas is "the focus of our appreciative interest in the work" (2004, p. 150). What we appreciate in a painting, on this account, is an achievement, and achievements are individuated by the activities that bring them about: the same surface, reached by some other route, would be a different achievement. Facts about how the object came into being thus do more than supply context; they help determine what the work is, and what is properly appreciated in it. Suppose, to adapt one of Davies' examples (2004, p. 102), that a storm were to blow pigment across a stretched canvas, leaving a surface indistinguishable, mark for mark, from an abstract painting. A theorist who identifies the painting with its surface must either count the storm's product as a painting or explain why it is not. There is a surface with no painting of a picture behind it, and the corresponding case for philosophy — a text indistinguishable from a philosophical argument, with no philosophising behind it — is the one an LLM presents. If Davies is right, the surface does not by itself settle the work. Transposed to philosophy, Davies' proposal would run as follows: a philosophical text is not itself the philosophical work; the text is the product of a person's philosophising, and reading it is a way of engaging with that prior activity. The activity, on this proposal, is part of what the work is, so that there is a philosophical work only where the philosophising has taken place, which makes the challenge a constitutive one. If no one has philosophised, there is no work to which the text gives access, however the text reads — an LLM text would stand to philosophy as the storm-made canvas stands to painting. We do not think the transposition should be accepted. What makes Davies' view plausible in the case of art is that indiscernible surfaces really do seem to differ in artistic value: the storm-made canvas is worth nothing as a painting, while a brushed one may be worth a great deal. Whoever transposes the view to philosophy is therefore committed to the corresponding claim, that two texts containing the same argument can differ in philosophical merit, and it is difficult to see what such a difference could consist in. Whether an argument is valid, whether its premises are plausible, and whether the objections to it have been answered are questions about a text's contents, and two texts with the same contents receive the same answers to them. The philosophical merit of a text does not vary with the route by which its words came to be written. Peer review proceeds on the same assumption. Journals strip author information from submissions before review because facts about authorship are treated as potential sources of distortion rather than as evidence of merit; if two texts with the same contents could differ in philosophical merit, anonymising would discard information relevant to the assessment, and review would not be designed as it is. The grounds for accepting or rejecting a paper lie in the argument as presented, not in the history of its production.[^2] Nor is the author-centred teaching noted earlier in tension with this. That ethics is taught through the *Groundwork* rather than through a digest of its conclusions reflects what a reader gains by working through Kant's arguments, and this is consistent with holding that the merit so gained is a feature of the text rather than of its author. It might still be insisted that where there has been no philosophising there is no work, whatever the resulting text contains. We need not resist this, because our thesis concerns texts worth reading, and a text can be worth reading without being a work in Davies' sense. Suppose a desert wind traced out, in the sand, a sound argument against enactivist theories of perception. Nobody would deserve credit for the argument, and there would be no performance for the text to give access to; but a reader who worked through it would still encounter a thesis and the considerations advanced in its favour, and would be in a position to answer or to extend it. 'Work' may, if the performance theorist insists, be reserved for texts with performances behind them. What cannot be so reserved is being worth reading, since everything that judgement answers to is on the page. Nothing in a text's having had no one behind it, then, settles whether it is worth reading. Whether an LLM can actually *produce* a text worth reading is a further question, and it is to this that we now turn. --- ## 2\. The challenge from abduction In the previous section, we argued against the idea that LLMs cannot produce philosophy worth reading simply because they are not human. In this and the two sections that follow we shall consider a different form of challenge: even if LLMs cannot be ruled out of producing philosophy worth reading tout court, they lack particular *capacities* that are required to produce it. If our arguments regarding authorship are correct, then a novel philosophical argument produced by a parrot would deserve to be taken just as seriously as one produced by a human being. No such argument will be forthcoming, however: a parrot can only reproduce sounds it has already heard. One might suspect that LLMs are in the parrot's position — that while nothing rules their outputs out of being philosophy worth reading, they lack some capacity that producing it requires. In this section we address one capacity challenge, which we will call *the challenge from abduction*. In the two sections that follow we shall look at two more: the *challenge from detachment* and the *challenge from experience*. Abductive inference is reasoning from a body of evidence to the hypothesis that would best explain it. In a deductive argument the premises fix the conclusion: if all men are mortal and Socrates is a man, then Socrates is mortal, and that is that. Now, imagine walking into your kitchen one morning and finding that part of the floor is wet. What has happened? The wet floor does not determine the answer in the way the two premises determined Socrates' mortality: a burst pipe would have left the floor wet, and so would a spilled bucket. But, given that the window is open, the water is under the window, and it rained last night, rain coming through the window seems the most plausible answer. To reason in this way, inferring the best explanation for a set of facts, is common in the sciences as well as everyday life. Williamson argues that philosophy should use a broadly abductive methodology (2007; 2021, §9.2). In philosophy, as in science, there are data that a candidate theory must accommodate — intuitions about cases, and the phenomena of the domain itself — and rival theories that would each accommodate them at different costs. The theory to prefer is the one that would, if true, best account for the data, where what makes one theory's account better than another's is its explanatory virtue: a good philosophical theory is, in Williamson's words, "elegant and unified, not arbitrary, gerrymandered, ad hoc, or messily complicated", and should "combine simplicity with strength" (2021, §9.2). Nor is this view of theory choice confined to those who, like Williamson, take philosophy to be methodologically continuous with the sciences: Bengson, Cuneo and Shafer-Landau (2022) argue that the assessment of rival theories by their explanatory and unifying merits is a constraint on sound philosophical method as such.\[^holders\] We shall assume it in what follows.[^3] Lipton (2004, p. 59) distinguishes two things that "the best explanation" might mean: the *likeliest* explanation, the one most warranted by the evidence, and the *loveliest*, the one that would, if true, provide the most understanding. As he puts it, "Likeliness speaks of truth; loveliness of potential understanding" (2004, p. 59). The two can come apart. Return to the wet kitchen floor: that water has fallen on it is the likeliest explanation of its being wet, and it is scarcely possible that it is false, yet it affords no understanding at all of how the floor came to be in that state: of course it is wet because water fell on it, but how, and which water? The explanatory virtues by which, on the methodology we have assumed, rival philosophical theories are weighed are virtues of loveliness: elegance, unity, and simplicity concern the understanding a theory would provide were it true, not the probability that it is true. Where a philosophical text turns on the weighing of rival explanations — not all philosophical writing does — whether the text is worth reading and whether it weighs its rivals well go together. If a capacity for abduction is required to produce worthwhile philosophy, do LLMs possess it? Floridi et al. (2025) argue that they do not: > We argue that such LLMs generate text based on learned associations rather than performing abductive inferences. \[…\] LLMs produce plausible hypotheses, simulate commonsense reasoning, and provide explanatory answers without grounding them directly in truth, semantics, verification, or understanding, and without any abductive reasoning. (Floridi et al. 2025, p. 1) An LLM, on this account, has "a stochastic core and an abductive appearance" (2025, p. 2); the further charge, that these systems cannot "genuinely validate" their explanations against reality (ibid., p. 6), we take up in the next section. Models are trained to predict which words are likely to follow which, and to produce the continuation that their training makes probable; nothing in the training directs them at truth. The appearance comes from what the training data have passed on: models have "absorbed patterns of human abductive reasoning as expressed in writing" (p. 9) — how explanations are typically phrased, which causes are typically offered for which effects. [^1] Floridi et al.'s own example is a car that will not start on a cold morning. Asked why not, a model replies that a weak battery is one possibility, since cold reduces a battery's efficiency; that thickened engine oil is another, since a cold engine is harder to turn over; and that, "\[b\]ased on your description, the battery is the most likely explanation" (2025, p. 10). On their reading, the LLM is not actually reasoning about causes from the case before it; it is the statistical reproduction of the causes such explanations typically cite (p. 9). What they deny is that the model weighs the candidates — that it "entails choosing the best explanation among alternatives" (2025, p. 3). The verdict that settles on the battery does no weighing; it reproduces how explanations of this kind conventionally end. And where the output marks a genuine difference between the two — some consideration that would tell the battery from the oil — it is one already drawn in the explanations the model learned from, not one worked out afresh for the case in hand. Because LLMs produce no more than a "veneer of explanation" (ibid., p. 20), Floridi et al. conclude, they can play only a supporting role in intellectual work: > In essence, LLMs function like brainstorming assistants that toss out ideas without filtering for quality. After all, they work like statistical interfaces to an enormous amount of data accumulated for millennia by generations. A cautious human collaborator can sift through and assess them. (2025, p. 11) A device for brainstorming seems a far cry from something which might produce worthwhile philosophy. If you were presented with a text and told that it contains a number of philosophical ideas, none of which have been filtered for quality, it is unlikely you would think it is worth your while reading it. The filtering these systems are said to lack is, in Lipton's terms, filtering for loveliness: the stochastic core already selects the likeliest continuation, and what it cannot do, on Floridi et al.'s account, is prefer one candidate explanation to another on the grounds of the understanding it would afford. Floridi et al. do not merely hold that the route by which an LLM produces its output is stochastic rather than abductive; they hold that the output itself is only apparently abductive — a "compelling illusion of genuine and structured inferential reasoning" (2025, p. 2). Trained on a great deal of writing in which explanations are offered and weighed, a model absorbs the forms this writing takes and, prompted to explain, reproduces them, following "the typical phrasing and structure of explanations" and offering "typical causes for typical effects" rather than "reason\[ing\] about causes from scratch" (2025, p. 9). The output has the shape of a weighing of explanations, taken over from the writing the model has digested, without any weighing having taken place. Floridi and his colleagues allow that, in ordinary cases such as their own cold-morning car, the model's answer is a good one: "the same explanation a human reasoner would likely choose" (2025, p. 10), one that may be "even optimal by IBE criteria" (2025, p. 19), since such systems "echo the obvious, common explanations" (2025, p. 10). The complaint cannot then be that the explanation is poor, which leaves it hard to say what, in such an answer, is supposed to be merely apparent. What their account points to is the uncommon case: "in less common situations, LLMs can falter" (2025, p. 10), and on inputs "that go beyond their training" "the facade can crack" (2025, p. 9), the success on familiar cases being "a sign of overfitting to common patterns" (2025, p. 15). Once the question of its truth is set aside, this is what the facade reduces to: the claim that the competence shown on common problems is overfitting that would give out in less common ones. Whether the competence does give out on uncommon cases is an empirical question. A recent survey of abductive reasoning in language models suggests that it does (Salimi et al. 2026). Current models do markedly worse on abductive tasks than on deductive ones: where their median accuracy on deductive tasks is near eighty per cent, on abductive ones it is some forty-two and a half, with a spread "extending down to near-zero accuracy", and the survey reports that "strong deductive performance does not reliably imply strong abductive performance". This is close to Floridi's own diagnosis — an answer that reproduces a common pattern instead of reasoning to it — now voiced from within the field that builds the systems. Taken at face value, it tells in Floridi's favour. We do not disagree with Floridi et al.'s characterisation of how LLMs function: these systems do not weigh and choose among alternatives in the way that humans do. However, we should not be too quick to jump from this to the conclusion that LLMs cannot *produce text* which exhibits abductive reasoning. A pocket calculator does not have the capacity to do arithmetic in the way a person does, but does have capacity to produce the correct answer to sums which are entered into it. Similarly, it might be possible for LLMs to produce text which displays abductive reasoning, despite it not being grounded in any actual abductive reasoning. Despite their stochastic core, LLMs are perfectly capable of producing grammatically correct text: without being given any specific rules, the system "implicitly 'discovers' them—and then seems to be good at following them" (Wolfram 2023). Does this mean that the texts LLMs produce have merely the appearance of being grammatically well-formed? Clearly not. LLM sentences *are* grammatically well-formed despite their stochastic roots. A stochastic core, then, need not mean that the best an LLM can manage is a veneer of abductive inference. It may be, rather, that the core is marshalled to produce text exhibiting actual abductive inference, in just the way it is marshalled to produce actual grammatical correctness. It might be objected that the grammar analogy will not stretch this far. The model has picked up the shape of abductive explanation, but it is not actually performing abduction. But grammar, the objector would say, is a system of rules a model can follow without understanding anything, and explanatory loveliness is not — it is not a system of rules you follow to get the right answer, so the model's grammatical competence gives no reason to expect it to produce lovely explanations. Floridi's claim is that the model's abductive output is merely apparent — that it has the form of an inference to the best explanation but not the substance. If that is right, then there should be cases where the form is present and the substance is absent: cases where the model produces something that looks like an explanation but is empty or nonsensical, just as it could produce something that is grammatically faultless and says nothing. Wolfram's example of the latter is "Inquisitive electrons eat blue theories for fish" (2023) — impeccably grammatical, and meaningless. The model does not produce such strings. The abductive equivalent would be a passage with the form of an inference to the best explanation, grammatically faultless, and yet senseless — asked why a car will not start on a cold morning, there are no squirrel tracks, so it must be freak arctic winds blowing into the exhaust pipe. That has the form of evidence weighed towards a conclusion, and it is nonsense; and it is nonsense the model does not produce. Put the question to it and it offers the weak battery and the thickened oil, plausible explanations rather than arctic winds. Wherever the facade is to be located, it cannot be located there: what the model produces are plausible explanations, and it is not yet clear what, in a plausible explanation, is supposed to be merely apparent. Moreover, the answer Floridi and his colleagues describe is not quite the answer these systems give. Their account asks us to picture a confident verdict laid over a hollow core, yet what one finds in practice is closer to hedging. Asked why a car would not start on a cold December morning, a current model will say that the battery is the most likely culprit, set out other plausible contributors — thickened oil, fuel-system problems, ignition faults — and end by observing that if the car started once the day had warmed, the battery is almost certainly the primary cause, though a load test would confirm it.[^4] The model's avoidance of senseless explanations, like its avoidance of senseless sentences, points to a semantic competence picked up in training, over and above syntax. A syntactic grammar settles only how the parts of speech may be combined, and not whether what a sentence says makes sense: avoiding strings like the inquisitive electrons takes a grasp of what can sensibly be said of what — which predicates go with which subjects, which causes with which effects. Wolfram calls such a grasp a semantic grammar: > to deal with meaning, we need to go further. And one version of how to do this is to think about not just a syntactic grammar for language, but also a semantic one. (Wolfram 2023) A model trained on enough text has, on his account, come by one: > From its training ChatGPT has effectively "pieced together" a certain (rather impressive) quantity of what amounts to semantic grammar. (Wolfram 2023) It is this, rather than any contact with the case in hand, that keeps the arctic winds out of the model's explanations: a feel for what, in a working model of the world, can hang together. But a feel for how things hang together is gathered from the text the model has read, and a model of the world is not the world.[^5] What a semantic grammar supplies is a sense of what would sound right, not a line to how things actually stand. The same detachment that keeps the model's explanations sensible is what leaves it unable, on its own, to reach the car on the drive: a semantic grammar is a grasp of how such failures are explained in general — cold slows batteries, thickens oil, fouls ignition — and it gives no hold on the particular car whose fault is in question. Set the particular car aside, and what is left is, so far as the text goes, abduction itself: a weighing of explanations that respects sense, fits its confidence to the evidence, and stops where the evidence stops. How much the missing purchase costs depends on the kind of abduction at issue. The car is a hard case, because its answer waits upon the world; only the car itself can tell the weak battery from the frozen line. Much philosophical abduction does not wait upon the world in that way: what a thought experiment commits us to, or which of two theories carries the lighter explanatory cost, is settled from what is already set down, where the discriminating evidence is the sort a corpus already holds. The survey scores of Salimi et al. (2026) measure whether the model arrived at the right answer, not the weighing that got it there. On the harder tasks the scores are low, and on long mysteries with their clues strewn through the text the best models fall just short of the average human solver; taken at face value, the numbers count against the model. But the score is a score for the answer — whether the named culprit was the keyed one, whether the right diagnosis came first — and not for the weighing that reached it; such a score, Salimi says, "completely bypass\[es\] the actual reasoning trace". So when a model misses the keyed culprit, the number marks the miss, and not the comparison it set out on the way — which explanations it canvassed, and why it came down on one. And the survey's hardest tasks are built from low-prior, non-stereotypical outcomes, where several explanations may be reasonable; there, matching an answer to a single reference "underestimates explanation quality", so a low score on such tasks does not show that the model cannot weigh, only that it did not land on the keyed answer. What a one-shot score leaves unscored is the weighing behind it, and that is what a fuller prompting would have to draw out. It is now difficult to see where the facade should be located. The explanations the model produces are plausible, the confidence they express is fitted to the evidence it has been given, and the benchmark scores that seemed to confirm the diagnosis measure answers rather than the weighing that produced them. What the model lacks is purchase on the particular case, and that is a limitation concerning its relation to the world rather than its capacity for abduction. How much it costs a philosophical text is a matter for the next section. What a one-shot prompt leaves undrawn, and how it might be drawn out, is taken up in Section 5. --- ## 3\. The Challenge from Detachment In this section we address a second capacity challenge, the *challenge from detachment*. In addition to the argument considered in the previous section, Floridi et al. object that an LLM stands in no relation to the world: its words rest on no perception of anything, and a hypothesis, once produced, is never tested against how things are (2025, pp. 6–7). A discipline whose theories answer to how things are, the challenge runs, cannot be advanced by a system with no access to how things are, and a text produced by such a system gives its reader no reason to think it worth reading. We argue that the challenge fails because the materials philosophy takes from the world reach it already articulated in language, and a model can work on those words as any philosopher does. We can grant that the model perceives nothing, and that nothing it produces is put to the test against the world. These concessions disqualify the model, however, only on a certain picture of how philosophy meets the world, and it is a picture we see no reason to accept. Pigliucci (2017), writing on the question of how philosophy makes progress, holds that philosophy is constrained by the world without investigating it as the natural sciences do: > This means that the basic parameters that philosophers use as their inputs, the starting points of their philosophizing, their equivalent of axioms in mathematics and assumptions in logic (or rules in chess) are *empirical* data about the world. This data comes from both everyday experience \[...\] and of course increasingly from the world of science itself. Philosophy, I maintain, is in the business of exploring the sort of conceptually evoked spaces that Smolin is talking about, where the evocation is the result of whatever starting assumptions are made by individual philosophers working within a particular field and, crucially, of the constraints that are imposed by our best understanding of how the world actually is. (2017, pp. 79–80) Pigliucci takes the term *evocation* from Smolin (Unger and Smolin 2015).[^6] Evoked truths are neither discovered, in the sense of corresponding to mind-independent states of affairs, nor invented, in the sense of being arbitrary constructs. Smolin's illustration is chess: before the rules of the game were codified there were no facts about chess, but once they were, a great many facts about the game became demonstrable — objective facts, in the sense that anyone who demonstrates one demonstrates the same fact as anyone else (Unger and Smolin 2015, p. 423; quoted at Pigliucci 2017, p. 78). On Pigliucci's proposal, philosophy is in the business of ascertaining evoked truths. What distinguishes it from mathematics and from chess is where its starting points come from: they are constrained by how the world actually is. The same constraint distinguishes philosophy from fiction. Pigliucci's contrast case is a science-fiction writer who describes the same planet across three different timelines, keeping every description within the laws of physics and of logic. The writer is exploring conceptual spaces of a kind, but he is inventing rather than evoking: his worlds have no rigid properties, because even the constraints he keeps to could have been otherwise — he could as easily have imagined planets with a different physics, or a different logic. Philosophy, by contrast, "is in the business of doing empirically informed evoking, not inventing", and its objects of study accordingly have rigid properties (2017, p. 80): once a philosophical starting point has been articulated, what holds within the structure it opens is no longer up to anyone, any more than what follows from the rules of chess is up to anyone. A thought experiment is a case in point. The philosopher sets up an imagined scenario, but explores it "with an interest in figuring things out as far as this world is concerned" (2017, p. 80), and the philosophical work then proceeds within a structure whose properties the philosopher does not control. Philosophy's starting points — what Pigliucci calls the basic parameters of philosophising, its equivalent of axioms in mathematics and rules in chess — are data about how the world is, drawn from everyday experience and, increasingly, from science. A system that perceives nothing cannot gather such data for itself. But nor does the working philosopher. The scientific data Pigliucci has in mind reach philosophers already set down in language: a philosopher of physics works from published results, not from having run the experiments. And what the discipline retains of everyday experience, it retains in the same form, as the literature's accumulated descriptions of how things seem. Written access to the world's constraints is the profession's normal condition, and a corpus provides access of just this kind. With respect to nearly all the empirical data philosophy uses, then, the model is in the same position as any philosopher. Philosophical claims are not, in any case, tested in the way the challenge assumes. The equivalence principle, once Einstein had it, faced a tribunal of measurement — the experiments might have gone against it, and it would then have been dropped. A philosophical thesis faces no such tribunal. Whether it holds is a question about what obtains within an evoked structure, and it is settled in the way questions about chess are settled: by working out what the starting points commit us to, something any competent party can do and none can decide by fiat. This working-out is carried out in writing, in the literature's back-and-forth of argument and reply, and a system that produces text is not debarred from it. A model's inability to run experiments therefore costs it nothing that a philosophical text needs: the testing philosophy does is done on the page. A model's lack of perception and experiment therefore leaves untouched both the starting points a philosophical text works from and the assessment its claims receive: the former arrive already set down in language, and the latter consists in working out what they commit us to. It might be doubted, however, that experience can be treated in the same way, since a description of what it is like to see red is not obviously an adequate substitute for seeing it, and the next section addresses a challenge built on this doubt. ## 4\. The Challenge from Experience In this section we address a third capacity challenge, the *challenge from experience*: the claim that some philosophy depends on experience in a way that a system which has never had any experience cannot meet. Few would say that current LLMs are conscious, and we assume here that they are not. Zahavy (2026) raises a worry of this kind about scientific discovery. A model can carry out the deductive part of discovery, working out the consequences of premises it has been given; what it cannot do, he holds, is produce the premises — make the move from sense experience to new first principles. On the picture he takes from Einstein, that move is a leap,[^7] and it is the leap that gives a theory its axioms. His case is the thought experiment that gave Einstein the equivalence principle: > Einstein's variation required inventing new axioms based on a physical intuition that did not yet exist in the mathematics. He envisioned a physicist inside an elevator being uniformly accelerated through deep space \[...\]. Inside this enclosure, the sensory experience reveals a specific pattern: when objects are released, the floor rushes up to meet them. To the physicist, the objects appear to fall with identical acceleration, regardless of composition. Thus, the simulation here was not a permutation of symbols, but a manipulation of perceptual experience. (Zahavy 2026, §5) Einstein imagines a set of circumstances and attends to what would be experienced within them: everything released inside the elevator appears to fall with identical acceleration. On Zahavy's reconstruction the simulation supplies an observation, and from that observation the new axiom is inferred — the simulated experience of acceleration was indistinguishable from the remembered experience of gravity, and Einstein concluded that the two are one phenomenon.[^8] A model has no access to that observation. It can produce descriptions of elevators and of weightlessness, both present in its corpus, but it has undergone neither, and a discovery whose premises are fixed by simulated experience is beyond a system that, in Zahavy's words, lacks the capacity he calls sensory agency. Philosophical thought experiments might seem to depend upon experience in the same way. Does Mary, released from her black-and-white room knowing every physical fact about colour vision, learn something when she first sees red (Jackson 1982)? To settle the question one must consider what the experience of seeing red is like, and the argument proceeds from that verdict. Here experience enters as the starting point of an argument, and a system that has never experienced anything seems unable to supply it. Experience enters philosophy as subject matter too, in work that asks what it is to see red (Harman 1990), to feel anger (Goldie 2000), or to have an intuition take hold (Chudnoff 2011).[^2] If starting points of the first kind, and subject matters of the second, can be handled only by a being that has the experiences, then much of the philosophy of mind lies beyond a model's reach. Nobody who debates the knowledge argument has been through what Mary goes through. The debate proceeds on Jackson's description of the case, together with his claim about what the case shows, and the replies work on that description. The replies the argument has accumulated dispute what follows within that description — whether what Mary gains on her release is knowledge of a fact at all — and none of them draws on what seeing red is like from the inside: they are arguments about what Jackson's description commits us to. Nor is the knowledge argument unusual in this. Nagel (1974) asks what it is like to be a bat without supposing he could find out, working instead from a description of the bat's situation. A model is no worse off here than these philosophers are: it has had no experiences of its own, but what its corpus supplies, and what the philosophy works on, is experience already put into words. There is, however, one kind of starting point that a model could not have supplied. Merleau-Ponty (1945 %%page%%) observed that when one hand touches the other, the roles of toucher and touched can alternate but cannot be occupied at once. Suppose that he came to this by attending to his own body, and not from anything already in the literature. Then here is a description that only first-person attention could yield, and a system with no body and no first-person view could not have been the first to set it down. Little follows from this, however. Once the description is written down, everything philosophical that can be done with it — what it shows about bodily self-awareness, what follows once the asymmetry is granted — is there on the page for any reader, and none of it depends on being able to have arrived at the description oneself. Nor is first-person attention the only source of descriptions the literature lacks: a new description can also be reached from the descriptions the literature already contains, by drawing out what they have not been taken to imply, or by combining them as no one has, and this a model can do. What the Merleau-Ponty case shows is that there is one route to new starting points that a model cannot take. Whether the route it can take yields descriptions the literature lacks is the question of novelty, which Section 6 takes up. A model can, then, do the philosophy that turns on experience, with one exception: it could not be the first to set down a description that only first-person attention could yield. The experiences philosophy argues about reach it as descriptions, which it can work through as well as any reader. In Section 2 abduction entered philosophy as reasoning displayed in a text, and in Section 3 the world entered it as starting points set down in one; here experience enters it the same way. --- ## 5\. The Challenge from Observation The challenge from observation begins with an obvious question: if all of these arguments are correct, where is all the LLM philosophy?[^9] If you ask an LLM the answer to the hard problem of consciousness or the meaning of life, the response you receive will not typically contain much in the way of philosophical argument. More likely you will get a somewhat bland survey of various philosophical positions on the topic, or perhaps just an irreverent joke.[^10] One might think that whatever the merits of the arguments we have put forward so far, *something* must be preventing them from producing philosophy worth reading because they *don't* produce philosophy worth reading. We want to suggest that this is a failure of prompting rather than evidence that LLMs lack the capability to produce philosophy worth reading. As we saw in Section 2, these systems are trained to produce reasonable continuations of the text they have been given (Wolfram 2023), before being trained further to converse as helpful assistants. In writing at large, questions like "what is the meaning of life?" are hardly ever followed by philosophical argument: arguments occur in journal articles, which begin from claims already advanced and objections already raised rather than from questions, and it is these that their arguments continue. Nor do things improve if the question is one that only a philosopher would ask. "Is representationalism the correct theory of perception?" is the sort of sentence that appears in examination papers and essay titles rather than at the opening of a defence of representationalism. A similar point applies to the assistant training: for almost anyone who asks a straight philosophical question, the most helpful answer is an accessible overview rather than a contribution to the literature. None of this shows that LLMs can produce philosophy worth reading, but it does mean that their failure to produce it when questioned point blank tells us nothing either way. We suggest that if philosophy worth reading is to be elicited from these systems, we need to think in terms of how it might be evoked. Evocation, in the passage quoted in Section 3, is the result of "whatever starting assumptions are made by individual philosophers" together with "the constraints that are imposed by our best understanding of how the world actually is" (Pigliucci 2017, pp. 79–80), and a prompt is where such starting assumptions can be laid down. On Wolfram's account, a prompt is not a question awaiting an answer but the opening of the text the model continues, and what it contains is made use of in everything that follows (2023). A prompt can therefore do what the opening pages of a paper do: state a position, and set out the constraints under which that position is to be defended. What is wanted from the continuation is the ascertaining of what holds in the space the prompt has opened. Where rival positions each accommodate the phenomena at different costs, that work is the weighing of explanations discussed in Section 2, and a prompt can ask for it explicitly rather than leaving it to be inferred from a question. Instead of asking whether representationalism is the correct theory of perception, one might set out representationalism and a rival, together with the phenomena both must accommodate, and ask which of the two would, if true, provide the better explanation of them; instead of asking for a verdict on a position, one might state the strongest objection to it and ask for the best reply available. Prompted in this way, the model is being asked for the reasoning that Section 2 argued its text can display, within a space the prompt has evoked. How best to do this is a question we leave open. A position set out under objections is one form an evoking prompt can take, but nothing restricts evocation to that shape: several hundred pages of draft manuscript make starting assumptions and accept constraints as surely as two paragraphs of thought experiment do, and which forms of prompt evoke most productively is an empirical question which we shall not try to settle. It is in the continuations of such prompts that the question of these systems' philosophical capacities would be decided.[^11] If a philosopher must supply the evocation — the position and the pressures under which it is to be developed — then it will be said that whatever philosophy the resulting text contains should be credited to the philosopher, and that the model is merely its instrument. This is the challenge of the next section. --- ## 6\. The Challenge from Instrumentality In this final section we address the *challenge from instrumentality*. Grant that philosophy worth reading comes out of these systems only when a philosopher directs the process, framing the problem and pressing for development. It may then be said that the philosophy in the resulting text belongs to the philosopher and not to the model. This is, in effect, the position at which Floridi et al. arrive: LLMs are brainstorming assistants, and it falls to "a cautious human collaborator" to "sift through and assess" what they toss out (2025, p. 11), the philosophy being done by the human who sifts and assesses. On this picture the model is an instrument of philosophical production, as a typewriter is, and to credit it with the result would be like crediting the typewriter with the novel. The challenge from authorship held that a model's text cannot be philosophy at all; the present challenge is narrower, holding that whatever philosophy such a text contains is not the model's contribution. Whether the challenge succeeds depends on what the prompt contributes and on what the continuation adds to it. The previous section gives us the first: a prompt that articulates a position and the pressures upon it evokes, standing to the model as Jackson's two paragraphs stand to the profession — a starting point handed over for development. The facts about what holds within the structure it opens outrun whatever it states, just as the facts about chess outran the rules the moment the rules were written down, and most of what a starting point evokes, no one has ever said. What the continuation adds is a text that states some of those facts. Wolfram's gloss on the mechanics is worth quoting: "the elements are already in there, but the specifics are defined by something like a 'trajectory between those elements' and that's what you're introducing when you tell it something" (2023). The elements belong to the corpus, and the trajectory is what the prompt introduces. The continuation is the text of that trajectory, and it states consequences of the starting point which the starting point does not itself state. Whether it states them well is judged as any philosophical text is judged, by whether the considerations it cites genuinely tell between the positions and by the understanding they would, if correct, provide, and the judgement is made by reading the continuation. It might be thought that the continuation only spells out what the prompt had already settled. It does not, because a starting point underdetermines its developments. The rules of chess settle every fact about chess, but they do not settle which of those facts get written down, in what order, or to what depth: two writers working from the same rules will produce different books, both correct, and neither dictated by the rules. The same holds here — the same prompt, run twice, yields different continuations (Wolfram 2023) — and the gap between starting point and development is where the model's contribution lies. Were there a single text that the prompt fixed, the output would merely restate what the prompter had settled, and the instrument description would be accurate. A development can also go wrong. A chess writer can publish a false theorem; a philosopher can misdraw the consequences of her own thought experiment; and a model can do both, along with its characteristic failure of asserting fluently what nothing supports. Such errors are found where the development is found, on the page, and they attach only to whatever produced the development: nobody blames the rules of chess for a false theorem, and nobody's typewriter has ever made an error of content. There are, then, three contributions to distinguish: the articulated starting point, which is the prompter's; the structure it evokes, whose facts belong to no one; and the text that develops them, which is the model's. The person writes the prompt and decides which continuations to pursue and when to stop; without the person there is only the response to the point-blank question. So much of the instrument picture is correct. What must be given up is the description of the model on which the picture turns. A typewriter fixes nothing its user has not already settled: every word of the novel was the novelist's before the keys were struck. The consequences a model's text states were nobody's before the text stated them. What the user of a typewriter settles is a text; what the writer of a prompt settles is a starting point. It may be replied that the prompt can always be enriched — the position stated, the rivals named, the objections listed together with the lines along which they are to be met — until the prompt contains the philosophy and the model merely expands upon it. But enriching a prompt enlarges the starting point without converting it into the development: a game with more rules is a larger game, not a book of its theorems, and however much the prompt states, the consequences the output draws were not among the statements. There is a genuine limiting case, in which the prompt states the comparison and the verdict and the continuation only rephrases them, and it is identified by reading the two together. A text that states nothing beyond its prompt is a paraphrase, and owed to the prompter; a text that states what the prompt left unstated is a development, and the unstated part is not the prompter's. Contributions are individuated this way in the literature quite generally — by what a text states that the texts before it did not — and nothing in that way of individuating them mentions what produced the text. A further question is whether such a development can state something the literature itself had not yet stated, and so be creative in the sense in which a human philosophical text is creative, rather than merely novel relative to its prompt. Here the mechanics impose a limit of their own. Wolfram's observation about telling a model something carries a caveat: the model can integrate what it is told only if it is "basically riding in a fairly simple way on top of the framework it already has" (2023). Trajectories, that is, run between elements the corpus already contains. Developments within the framework are what we have argued a model can supply; whether philosophy worth reading ever requires more than this — a starting point that does not ride on the existing framework, a leap of the kind Section 4 conceded in Merleau-Ponty's case — is a question this paper leaves open, and it would be settled as the narrower one was: by reading the output against the literature and asking what it states that the literature did not. These systems are continuation systems, and philosophy worth reading requires a context that evokes. A prompt can supply one without thereby fixing the development that follows from it. The philosophical standing of any output is therefore settled by what its continuation adds: first to the prompt that occasioned it, and then to the literature it would join. ## Appendix: the cold-morning exchange > **User:** Last December, I tried to start my car early in the morning when it was very cold. It wouldn't start. Why not? > > **Kimi k2.67:** The most likely culprit is the battery. In very cold weather, a battery's chemical reactions slow dramatically, reducing its available capacity by up to 50%. If the battery was already marginal (older than three years, or not fully charged), freezing temperatures can push it below the threshold needed to turn the starter motor. You may have heard rapid clicking or nothing at all. > > Other plausible contributors: > > - Thickened engine oil — Cold increases viscosity, making the engine physically harder to crank and demanding even more current from the already weakened battery. > - Fuel system — Condensation in the fuel lines can freeze, blocking flow. Diesel fuel can gel; petrol can struggle to vaporise properly in extreme cold. > - Spark/ignition — Moisture in the distributor cap or on spark plugs can freeze, preventing a proper spark. > > If it started fine once temperatures rose later in the day, the battery is almost certainly the primary cause. A load test would confirm whether it needs replacement or just a longer drive to reach full charge. --- ## References ## References Frankish, K. (2024). What are large language models doing? In A. Strasser (Ed.), *Anna's AI Anthology: How to live with smart machines?* (pp. 55–78). Xenomoi. --- **Claude** Generating Philosophy with Artificial Intelligence [^1]: This is not to deny that systems of this kind can produce beautiful images; we return to image generation in Section 4. [^2]: The same location of philosophy in the public text is reached by accounts of philosophical progress. Dellsén et al. (2024) hold that progress is "for-whom" rather than "by-whom": it consists in putting people in a position to increase their understanding, usually by making philosophical ideas publicly available (p. 679). Being put in such a position requires something one can take up and think through, and what is available to be taken up is the text. A philosophical contribution so understood is constituted by what the public text makes available, which a view that locates the philosophy in the antecedent private activity mislocates. [^3]: Understanding-based accounts of philosophical progress converge on the same standard: if progress in philosophy consists in placing people in a position to increase their understanding (Dellsén et al. 2024, p. 679), then rival theories are rightly weighed by the understanding they would, if true, provide. [^4]: The exchange, with Kimi k2.67 (%%add date of test%%), is reproduced in full in the Appendix. The reply does not announce the answer; it fits its confidence to the little it has been told, and marks the point past which it will not go without knowing more of the particular car. This is not yet to say that Floridi's facade charge is refuted — a hedged answer can still be, on his account, a statistical reproduction of how hedged explanations typically read. But it does mean that Floridi et al.'s own example does not quite fit the charge. The case they describe, in which a confident surface verdict is laid over a hollow core, is not the case their own example presents: the confidence these systems express is already answerable to their evidence, and a confidence so answerable is not the facade the objection has in view. [^5]: We use "model of the world" in Wolfram's informal sense: a body of implicitly learned regularities concerning what goes with what. This is weaker than the technical sense now disputed in machine learning, on which a world model is an internal representation of an environment that supports prediction and counterfactual reasoning. Nothing in our argument requires deciding whether LLMs possess world models in the technical sense; how they connect to the world is the matter of §3. [^6]: *The Singular Universe and the Reality of Time* is jointly authored, but its second part, which contains the discussion of evocation, was written by Smolin alone, as Pigliucci notes (2017, p. 77); we follow him in attributing the view to Smolin. [^7]: Zahavy too calls this leap abduction, but the word picks out something other than it did in Section 2. There, with Floridi et al., abduction was the weighing of rival explanations, and the charge was that a model only mimics it; here it is the generation of new first principles from experience, and Zahavy's claim is that a model cannot make the move because it has had no experience to move from. The present challenge rests on that second claim, about experience, and not on any verdict about the weighing. [^8]: Zahavy, following Magnani, calls the process *manipulative abduction*: hypothesis generation through the manipulation of a model — here a simulated experience — rather than of symbols (Magnani et al. 2009; Zahavy 2026, §5). It is abduction in Section 2's sense: the equivalence principle is inferred as the best explanation of the simulated observation, the simulation supplying an explanandum that no search over existing text would have produced. What experience contributes, on this picture, is not the inference but its starting point. [^9]: Strictly speaking, the observation is difficult to verify. To know that no LLM-written philosophy has appeared in the journals over the last half-decade, one would need to know that every author who has declared that they did not use AI in preparing their manuscript was telling the truth. [^10]: While preparing this paper we asked GPT-5.5 for a detailed overview of the positions an analytic philosopher might take on the meaning of life. What came back was a competent, hedged survey of the field; what did not come back was an argument for any position in it. %%add date of test%% [^11]: That these systems are mischaracterised as oracles — with the corollary that no benchmark of single-pass answers should be expected to probe the upper limits of what they can produce — has been argued from inside the practitioner literature (Janus 2022). The benchmark literature points the same way: the survey of abductive benchmarks discussed in Section 2 scores single-pass answers to fixed instructions while cataloguing, in the same pages, prompting methods that alter what models produce (Salimi et al. 2026).