{"path":"research/the-tractatus-and-deliberus.md","content":"# The Tractatus and Deliberus: one ambition, three assumptions apart\n\n**Date**: 2026-10-06 · **Type**: research. A primary text read in full and related to the corpus.\n**Occasion**: founder, verbatim: *\"Tractatus Logico Philosophicus - read it in full please. Then think\ndeeply about how it relates to Deliberus. First ground extremely deeply on Deliberus docs and vision\netc.\"* · **State**: findings and proposals. **Nothing here is ruled.** The positions that would change\nsomething are filed as ledger Arc AJ.\n\n**Read-depth note, per the corpus's convention.** The whole book was read: Russell's introduction, the\npreface and all 526 propositions, in the 1922 English (Project Gutenberg #5740). The key passages were\nchecked against the German printed beside it. That English is usually credited to C. K. Ogden, but it\nwas in fact made by Frank Ramsey at eighteen, which matters in § 4d. Facts about the book's later\nhistory were checked against outside sources the same day and are listed at the end. A few claims\nabout how later thinkers descend from the book are from memory and marked so. Not reachable from this\nsession: `.private/` and the local archive. So this doc cannot say whether the founder has engaged the\nTractatus before (the word appears nowhere in the tracked corpus).\n\n**Grounding, read in full this session**: [vision.md](../vision.md), [depth.md](../depth.md),\n[convergence.md](../convergence.md), [analysis-and-attunement.md](../analysis-and-attunement.md),\n[self-similar-decomposition-and-claim-ontology.md](self-similar-decomposition-and-claim-ontology.md),\n[the-load-bearing-unsaid.md](the-load-bearing-unsaid.md),\n[what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md),\n[session33-the-horizon-vocabulary.md](session33-the-horizon-vocabulary.md),\n[nativism-thread-and-false-dissolution.md](nativism-thread-and-false-dissolution.md). Read in the parts\nthat bear on this: [philosophical-foundations-stress-test.md](philosophical-foundations-stress-test.md)\n§ 7, [decomposing-value.md](decomposing-value.md) §§ 4–5,\n[parity-and-the-zone-interior.md](parity-and-the-zone-interior.md),\n[what-human-judgment-is-for.md](what-human-judgment-is-for.md) §§ 1–3,\n[claim-sameness-philosophical-readings.md](claim-sameness-philosophical-readings.md),\n[starting-values-and-the-problem-of-the-prior.md](starting-values-and-the-problem-of-the-prior.md)\n§§ 0–2, [auto-connect-upgrade.md](auto-connect-upgrade.md) § 2c,\n[lean-deliberus-analogies.md](lean-deliberus-analogies.md) § 3, `deliberus/extraction/decompose_claim.py`,\nand the constellation's state in `specs/landing-redesign/constellation-map.md`.\n\n**Why this, while the reviews outrank everything**: the founder asked for it directly. It also bears on\nthe constellation review where it stands now. Star fifteen is *never ask a mind to audit its own frame*,\nand after it comes the reframe act. The Tractatus has the sharpest early statement of the first (§ 4d)\nand a concrete answer to an open question about the second (§ 7, AJ2).\n\n---\n\n## 0. The answer in plain words\n\n**The Tractatus wants what Deliberus wants, and it starts from three assumptions Deliberus has given\nup.** Both say that thinking can be made clear if you take statements apart into their pieces and show\nhow the pieces hold each other up. Both say that much of what looks like deep disagreement is really\nconfusion about words. And both say the job is an activity, clarifying, and not a set of answers.\nWittgenstein: *\"Philosophy is not a theory but an activity … The result of philosophy is not a number\nof 'philosophical propositions', but to make propositions clear\"* (4.112). Swap *philosophy* for\n*Deliberus* and that is the line that it maps reasoning and does not endorse conclusions.\n\nThe three assumptions it held and Deliberus does not:\n\n1. **There are final atoms, and every statement has exactly one complete analysis into them** (2.02,\n   3.25). Deliberus: *nothing is permanently atomic*, and a claim can be cut along several axes.\n2. **The limit of what can be said is fixed in advance, and past it you must be silent** (preface, 7).\n   Deliberus: *no stop signs anywhere*. Where mapping stops is found by trying, and every stop is\n   dated. Where the Tractatus prescribes silence, Deliberus keeps mapping, and owes attunement in\n   proportion to how deep it goes.\n3. **One mind reasons alone, and the frame it reasons inside is invisible to it for good** (2.174,\n   5.6, 5.633). Deliberus agrees with the second half and draws the opposite conclusion: *what is\n   invisible from inside a position is visible from outside it*, so other minds supply what one mind\n   cannot.\n\n**Two findings the corpus did not have.** First, it quotes Wittgenstein only as a critic of\ndecomposition (meaning is use, hinges, family resemblance), while the live decomposition prompt runs his\nearlier theory under Russell's name. By Russell's own account, his lectures on logical atomism were\nlargely an exposition of Wittgenstein's ideas. So the project already uses the early Wittgenstein and cites the later one against\nitself, without seeing that they are one person in two decades.\n\nSecond, **the book's own history is the best-documented case of the failure the meta-layer gate\nfears.** In 1918 the author declared his foundations *\"unassailable and definitive\"*. Outside readers\nbroke them on contact, and he spent his later work saying so. The point where they broke is an edge\nthis corpus has met from its own side: some parts of a claim cannot be judged independently, and the\ncorpus measures that edge today.\n\n**What the reading gives the project**: about twenty exact or structural echoes, each checkable (§ 5).\nThree warnings, of which one is new: a shared bottom that rules nothing out is not convergence (§ 6).\nAnd seven proposals, none ruled (§ 7).\n\n---\n\n## 1. The book, briefly, for a reader who has not read it\n\nSeven numbered propositions, with comments numbered beneath them. Wittgenstein's own footnote: the\ndecimals *\"indicate the logical importance of the propositions … the propositions n.1, n.2, n.3, etc.,\nare comments on proposition No. n\"*. **So the book is laid out as a tree of comments in which depth\nmeans weight**: a hand-built map of an argument from 1921. Reading only the seven whole numbers gives\nits skeleton, which is progressive disclosure done in print.\n\nIts argument in one paragraph:\n\n- **The world is the totality of facts.** Facts are built from atomic facts, which are independent of\n  each other. Atomic facts are built from simple objects, which are the world's fixed *substance*.\n- **We make pictures of facts.** A proposition is a picture. It shares a *logical form* with what it\n  pictures, and it has its sense whether or not it is true.\n- **Every proposition is a truth-function of elementary propositions**, so in principle every\n  statement analyses, uniquely, into independent atoms. The truth table is invented here (4.31).\n- **The propositions of logic are tautologies and say nothing.** Nothing can be inferred from one\n  state of affairs to another wholly different one: *\"Superstition is the belief in the causal nexus\"*\n  (5.1361). Induction has *\"no logical foundation but only a psychological one\"* (6.3631).\n- **The form a picture shares with the world cannot itself be pictured. It shows itself.** *\"What can\n  be shown cannot be said\"* (4.1212). This covers logical form, the subject (*\"a limit of the\n  world\"*, 5.632), value, ethics and the meaning of life.\n- **So the book's own sentences are nonsense**, rungs of a ladder to be thrown away once climbed\n  (6.54). Then: *\"Whereof one cannot speak, thereof one must be silent\"* (7).\n\nHis 1919 letter to Ludwig von Ficker, the editor he hoped would publish it, gives the intended weight: the work *\"consists\nof two parts: of the one which is here, and of everything which I have not written. And precisely\nthis second part is the important one.\"*\n\n## 2. What the corpus already had, and what it missed\n\n**What it had.** All of it is the later Wittgenstein:\n\n- *Meaning is use*: the formalization paradox in [steelmanned-critiques.md](steelmanned-critiques.md),\n  and *Wittgenstein's Objection* in\n  [self-similar-decomposition-and-claim-ontology.md](self-similar-decomposition-and-claim-ontology.md).\n- Family resemblance and use-relative sameness, in\n  [claim-sameness-philosophical-readings.md](claim-sameness-philosophical-readings.md).\n- Hinges from *On Certainty*, in [the-load-bearing-unsaid.md](the-load-bearing-unsaid.md) and the\n  stress test § 7, which name the hinge-score collision and the missing hinge type of horizon.\n- *Philosophical Investigations* § 129, the epigraph on star fourteen.\n\nIn every case Wittgenstein is the voice that warns against taking things apart.\n\n**What it missed.** The word *Tractatus* does not occur in the tracked repository. Meanwhile\n`decompose_claim.py`, which has written marked splits on the live site since 2026-09-21, opens its\ntheory block with *\"LOGICAL ATOMISM (Russell). A proposition asserts exactly one thing. A compound\nproposition is a truth-function of atomic ones.\"* Russell advertised his 1918 lectures *The Philosophy\nof Logical Atomism* as *\"very largely concerned with explaining certain ideas which I learnt from my\nfriend and former pupil Ludwig Wittgenstein\"*. The prompt's test (*could a reasonable reader affirm one\npart while denying another?*) is the Tractatus's test of independence: *\"From the existence or\nnon-existence of an atomic fact we cannot infer the existence or non-existence of another\"* (2.062).\n\n**So the project runs the early Wittgenstein in production and quotes the later one against itself.**\nThe two are one person who changed his mind, and the change is exactly the move the corpus has made\non its own: from permanent atoms to *currently undecomposed*. That is the first thing worth knowing. The\ncorpus's position is not the late Wittgenstein's objection with the early Wittgenstein's tool bolted\non. It is a third position that keeps something from each: the tool, and the reason the tool never\nreaches a bottom.\n\n## 3. The shared ambition: what Deliberus inherits\n\nEach of these is a Tractatus sentence that a Deliberus doc says in other words. None is a citation the\ncorpus made, so each is an independent arrival.\n\n**Clarification is the work, and it produces no doctrine.** 4.112, quoted in § 0. Its last clause in\nthe 1922 English: philosophy *\"should make clear and delimit sharply the thoughts which otherwise are,\nas it were, opaque and blurred.\"* The word is *opaque*, as in *Opacity Is a Cost, Not a Mystery*. That\nis a coincidence of wording, not lineage, and it is a pleasant one.\n\n**What can be said can be said clearly.** Preface and 4.116: *\"Everything that can be thought at all\ncan be thought clearly. Everything that can be said can be said clearly.\"* The founding conviction says\nthe same about other minds: *the reason they hold it is findable*. The two differ on the price.\nWittgenstein draws a sharp line between the clear and the silent. The conviction draws a gradient of\ncost that infrastructure lowers.\n\n**Most disputes are confusions about words.** 3.323–3.324: *\"the same word signifies in two different\nways … Thus there easily arise the most fundamental confusions (of which the whole of philosophy is\nfull).\"* And 4.003: *\"the deepest problems are really no problems.\"* This is [vision.md](../vision.md)\n§ Semantic Disambiguation as Root Cause (the *barnfattigdom* debate, *a public fight about nothing*)\nstated about philosophy instead of politics. The corpus has since learned the limit of this claim, and\nthe Tractatus never did (§ 4c).\n\n**Language hides its own structure.** 4.002: *\"Language disguises the thought; so that from the\nexternal form of the clothes one cannot infer the form of the thought they clothe … The silent\nadjustments to understand colloquial language are enormously complicated.\"* 4.0031 credits Russell\nwith showing *\"that the apparent logical form of the proposition need not be its real form.\"* The\nsilent adjustments are [the-load-bearing-unsaid.md](the-load-bearing-unsaid.md). The clothes are the\ndeep-descent experiment's finding that *\"pain is bad\"* was an is-claim dressed in ought's grammar\n([deep-descent-experiment.md](deep-descent-experiment.md)). The real form under the apparent one is\nmother-claim bundling.\n\n**A statement about a whole analyses into statements about its parts, plus what completes them.**\n2.0201: *\"Every statement about complexes can be analysed into a statement about their constituent\nparts, and into those propositions which completely describe the complexes.\"* That is the mother-claim\npattern, and its last clause is the coverage check, the one that rebuilds a parent from its parts\nalone. 3.24 adds how you can tell a statement is about an unanalysed whole: *\"That a propositional\nelement signifies a complex can be seen from an indeterminateness in the propositions in which it\noccurs. We know that everything is not yet determined by this proposition.\"* That last sentence is a\nsorry marker written in 1918: *not everything is yet determined here*.\n\n**The whole stands or falls with its parts.** Section 5 builds every proposition as a truth-function\nof its parts. 5.14 states the consequence the strength layer now uses: *\"If a proposition follows from\nanother, then the latter says more than the former.\"* A conjunction says more than each conjunct, so\nit cannot be more certain than its weakest one. That is the Fréchet bound, and it is why the ruled\ndefault for decomposition parts is *the weakest part decides*\n([weakest-link-arithmetic-and-the-merge-hunch.md](weakest-link-arithmetic-and-the-merge-hunch.md)).\nThe truth table (4.31, 5.101) has a gradual descendant here too. `hinge.py` holds one premise at\n*granted*, then at *denied*, and measures how far the conclusion moves. That is one row of a truth\ntable, evaluated on a scale.\n\n**A statement has its sense before anyone decides whether it is true.** 2.22: *\"The picture represents\nwhat it represents, independently of its truth or falsehood.\"* 4.024: *\"To understand a proposition\nmeans to know what is the case, if it is true. (One can therefore understand it without knowing whether\nit is true or not.)\"* Two Deliberus commitments rest on this. Nothing needs permission to exist: a claim\nis a full citizen of the graph before anyone judges it. And the stranger test, once the decontextualizing\nis set aside, asks 4.024's question: does a reader with no context know what would be the case if this\nwere true?\n\n**New statements, old words.** 4.03: *\"A proposition must communicate a new sense with old words.\"*\nThis predicts a split the corpus measured and briefly read as a failure: **reuse lives in the concept\nlayer (87% of concepts reused), while claim-level reuse is near-absent**\n([lowering-the-cost.md](lowering-the-cost.md) § 6). On 4.03's account that is the expected shape of\nany language that works. Words are reused; the sentences built from them are new. Claim reuse is a\nreal target where people restate the same proposition, which is the sameness problem. It is not a\nbase rate the corpus should have expected to see.\n\n## 4. Three assumptions apart\n\n### 4a. Atoms, and the one complete analysis\n\nThe Tractatus needs simples and gives an argument for them: *\"If the world had no substance, then\nwhether a proposition had sense would depend on whether another proposition was true. It would then be\nimpossible to form a picture of the world (true or false)\"* (2.0211–2.0212). And: *\"The postulate of\nthe possibility of the simple signs is the postulate of the determinateness of the sense\"* (3.23).\nThen: *\"There is one and only one complete analysis of the proposition\"* (3.25).\n\n**Deliberus accepts the consequence that Wittgenstein used to refute the alternative.** In this graph,\nwhether a claim has a definite sense often does depend on whether another claim is true. That other\nclaim is the definitional claim, which this project made a first-class, contestable node. A\nclaim that uses *fair* means something definite only relative to an account of *fair* that is itself\nin the graph and open to attack. Wittgenstein concluded that picturing would then be impossible. The\ncorpus's working answer is that picturing is **provisional**: the picture is good enough to argue\nover, and its dependence on an open definition is marked as a gap someone can fill. That is the sorry\nmodel doing work the Tractatus said nothing could do.\n\n**3.25 is false here, and its falsity is why decomposition is an attack surface.** If every claim had\nexactly one complete analysis, two honest people splitting it would produce the same parts, and\nBarnes and Christiano's obfuscated arguments, where error is spread thin across many barely-checkable\nparts, would have nowhere to hide. The threat model already names the defence, stability under\nre-decomposition, which in these terms asks how close a claim comes to 3.25. It is not yet measured.\nThe nearest built study re-files existing parts under four readings rather than re-cutting the claim\n(91% of verdicts survived, [slicing-stability-first-run.md](slicing-stability-first-run.md)), so the\nre-cutting half, the one 3.25 is about, is still open.\n\n**The book contains its own concession.** It never gives one example of an elementary proposition or\na simple object. When it comes to say what they are, it stops: *\"The application of logic decides what\nelementary propositions there are. What lies in the application logic cannot anticipate\"* (5.557), and\n*\"If I cannot give elementary propositions a priori then it must lead to obvious nonsense to try to give\nthem\"* (5.5571). This is the corpus's three-runs law in Wittgenstein's own words. The categories come\nfrom use and cannot be fixed in advance, which is also [what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md)\n§ 4, *settle it first is a closed path*. **Deliberus takes 5.557 one step further: application never\nfinishes deciding.** That one step is the whole difference between *atomic* and *currently\nundecomposed*.\n\nTwo more lines show how far apart the positions are. 4.1274: *\"one cannot ask: 'Are there unanalysable\nsubject-predicate propositions?'\"* Deliberus asks exactly that, and answers it on every node with\n*not yet*. And the later Wittgenstein, *Philosophical Investigations* § 47: *\"What are the simple\nconstituent parts of a chair? — The bits of wood of which it is made? Or the molecules, or the\natoms?\"* He concludes that it makes no sense to speak absolutely of the simple parts of a chair.\nWhether something counts as simple depends on the purpose. The separability floor is that answer\nmade operational: *simple* means *nothing separable is still bundled, for the purpose of being argued\nwith*.\n\n### 4b. Where the book broke, and where we meet the same edge\n\nThe first crack is in the text itself. 6.3751: *\"For two colours, e.g. to be at one place in the visual\nfield, is impossible, logically impossible, for it is excluded by the logical structure of colour.\"* But\nthe book's atoms are independent: *\"It is a sign of an elementary proposition, that no elementary\nproposition can contradict it\"* (4.211). So *this point is red* and *this point is green* cannot both be\nelementary. Yet they look as simple as statements get. Ramsey pressed this in his 1923 review, and in\n1929 Wittgenstein conceded in *Some Remarks on Logical Form* that the independence of elementary\npropositions must be given up.\n\n**Why this matters here.** Statements that place one thing on one scale (red or green, 30 or 40\ndegrees, *fair* or *unfair*) are not independent of each other, however simple each looks. The\nTractatus assumed perfect independence between atoms (2.061). Herbert Simon's term for real systems\nis *near*-decomposability, and [depth.md](../depth.md) already adopts it: parts interact weakly, not\nnever. The corpus has met the same edge, the limit of perfect independence, at least three times without the\nname:\n\n- **The overlapping-subtree failure** that [decomposition-axes.md](decomposition-axes.md) calls a sheaf\n  gluing condition: a part that belongs to two wholes at once, whose two local readings fail to agree.\n  A different shape from the colours, and the same lesson: the parts are not independent.\n- **The seven keys.** The keeper distinguishes *opposes* and *overlaps* from *unrelated* precisely\n  because two claims can be neither identical nor independent. The Tractatus allowed atoms only one\n  relation to each other, independence; the keeper's set is what you need once that fails.\n- **Weighing and degree.** A weighing such as *\"liberty outweighs safety\"* puts two goods on one\n  scale, so its parts are not independent of each other. That is one reason weighings get their own\n  fourteen questions instead of a plain split.\n\nSo the edge the Tractatus fell off is one Deliberus walks along on purpose. **The rule that follows:\nindependence is judged pair by pair, never assumed.** That is the keeper's job, and it is why sameness\nis decided before any relation is written.\n\n### 4c. A limit drawn in advance, and silence beyond it\n\nThe book draws its limit before any analysis starts: *\"The limit can, therefore, only be drawn in\nlanguage and what lies on the other side of the limit will be simply nonsense\"* (preface). 4.115: it\n*\"will mean the unspeakable by clearly displaying the speakable.\"* And 6.522: *\"There is indeed the\ninexpressible. This shows itself; it is the mystical.\"*\n\nDeliberus's position is on record, and it is the opposite in one respect and close in another.\n\n- **Opposite on where the limit is and how it is found.** [depth.md](../depth.md): *\"People can put\n  words to more than they think they can, if they try. Not many things are truly ineffable.\"* And it\n  gives the rule for where Deliberus would rest: *\"a claim that consistently resists decomposition over\n  many honest attempts.\"* That limit is found by attempts, after the fact, and it moves. The\n  Tractatus's limit is fixed before anyone tries. The founder's ruling of 2026-08-25 leaves no room for\n  a fixed limit: *\"I don't see any fundamental need to put up stop signs for ourselves in facilitating\n  either/both however, just because it's sensitive and sacred.\"*\n- **Close on there being a floor.** The founder's own voice memo, quoted in the claim-sameness doc:\n  *\"Det finns en fenomenologi som inte går att bryta ner i mindre atomiska delar, men innan man har\n  kommit dit, det är en lång väg dit.\"* (There is a phenomenology that cannot be broken into smaller\n  atomic parts, but before you get there it is a long way.) That is 6.522 with a distance added. The\n  disagreement is not whether there is an inexpressible. It is how near it is, and whether you are\n  allowed to find out.\n\n**Where the Tractatus prescribes silence, Deliberus keeps mapping and owes attunement in proportion to\nthe depth.** That sentence compresses two ratified lines (*no stop signs*; *analysis may go anywhere,\nand owes attunement in proportion to depth*). It does not replace them. The fallible-self-knowledge\nruling of the same day says what the Tractatus could not. The protected core is small: the felt\nexperience itself, which is the person's own. What surrounds it (causes, counterfactuals, satisfiers)\nis ordinary territory. The Tractatus put the whole subject and all of value beyond the limit. Deliberus\nputs one datum beyond analysis and keeps everything around it mappable.\n\n**The value floor is the Tractatus rediscovered.** [decomposing-value.md](decomposing-value.md) § 4\nfinds that the normative floor is *\"not reached by finding a missing ingredient — it is reached with the\ningredient list complete.\"* 6.41: *\"In the world everything is as it is and happens as it does happen.\nIn it there is no value.\"* 6.52: *\"even if all possible scientific questions be answered, the problems\nof life have still not been touched at all.\"* The corpus reached Wittgenstein's point by descending\nthrough physiology, without citing him. That counts in the point's favour.\n\n**And then it goes further than he would.** He concludes: *\"Hence also there can be no ethical\npropositions\"* (6.42), and *\"The facts all belong only to the task and not to its performance\"*\n(6.4321; the German says *not to the solution*). The founder's parity ruling goes past exactly that\nline. A permissive zone has an interior, *\"traversable, decomposable and representable in language,\nperhaps not fully but quite deeply\"*, and how people crossed it is mapped, attributed and kept for the\nnext traveller. The Tractatus leaves the solving to silence. Deliberus maps how real people solved it,\nwithout declaring a winner.\n\n**The Tractatus names the failure of its own method in its own words.** 6.53: the correct method would\nbe to say only what can be said and to show anyone who says something metaphysical that a sign in\ntheir sentence has no meaning. *\"This method would be unsatisfying to the other — he would not have\nthe feeling that we were teaching him philosophy — but it would be the only strictly correct method.\"*\nThat is analysis without attunement, admitted by its strongest practitioner to leave the other person\nunsatisfied. The Kalla and Broockman field result, in which arguments alone were null and arguments\nwith narrative were durable, measures the same thing\n([analysis-and-attunement.md](../analysis-and-attunement.md)).\n\n**Where the Tractatus beats the corpus to a correction.** 6.372, on the modern worldview: *\"people stop\nshort at natural laws as at something unassailable, as did the ancients at God and Fate … But the\nancients were clearer, in so far as they recognized one clear conclusion, whereas in the modern system\nit should appear as though everything were explained.\"* The German for *conclusion* is *Abschluss*,\nand the 1961 translation by Pears and McGuinness renders it *terminus* *[from memory]*. Wittgenstein's point is that a system which\nadmits where its explanation stops is more honest than one that seems to explain everything. That is\nthe confession principle (*if a system cannot confess failure, it will report success*). It is also why\nthe horizon map exists. **Deliberus keeps half of 6.372.** It keeps the admission of where explanation\nstops, and drops the finality. The ancients' stopping point was God and Fate, permanent by design. The\nhorizon is dated and moves, which is the founder's 2026-09-14 correction: *\"the finality implied in\n'terminus' and 'dissolved' etc is premature.\"*\n\n**The dissolution thesis, and where the founder parts from it.** The Tractatus's central therapy is\nthat problems vanish once language is clear: *\"The solution of the problem of life is seen in the\nvanishing of this problem\"* (6.521), and *\"the deepest problems are really no problems\"* (4.003). The\nfounder's ruling on *dissolved* answers it directly: *\"people will not have dissolved the conflict even\nif they all are clear on the fact that there's widespread semantic confusion/conflict.\"* Clarity is not\ndissolution, which is Plunkett and Sundell's *metalinguistic negotiation* and the founding conviction's\nown line that clarity *clarifies* interest-conflict without dissolving it. On this the corpus sides\nwith the later literature against the early Wittgenstein, and holds a test he lacked. The\nfalse-dissolution test asks: *after the dissolution, does anything about what to actually do change?*\n§ 5 shows that the Tractatus also supplies the reason the test works.\n\n### 4d. One subject, and the two outside readers who broke the book\n\nThe book is written from one mind, on purpose. *\"The limits of my language mean the limits of my\nworld\"* (5.6). *\"That the world is my world, shows itself in the fact that the limits of the language\n(the language which only I understand) mean the limits of my world\"* (5.62). And the frame cannot see\nitself:\n\n> *\"The picture represents its object from without (its standpoint is its form of representation),\n> therefore the picture represents its object rightly or falsely. But the picture cannot place itself\n> outside of its form of representation.\"* (2.173–2.174)\n\n> *\"You say that this case is altogether like that of the eye and the field of sight. But you do not\n> really see the eye. And from nothing in the field of sight can it be concluded that it is seen from an\n> eye.\"* (5.633)\n\nThis is the first clause of the positionality conviction, stated sharper than anywhere in the corpus.\nThe word *standpoint* is in 2.173: every picture has one, and it cannot picture it. It is also the\nreason behind star fifteen, *never ask a mind to audit its own frame*. A mind asked what it is missing\nsearches with the eye it cannot see.\n\n**The Tractatus stops at the first clause. Deliberus adds the second: someone standing elsewhere can\nsee your eye.** And the book's own history is the proof. The preface to the *Philosophical\nInvestigations* says he was forced *\"to recognize grave mistakes in what I wrote in that first book\"*,\nand credits two critics for it. One was Frank Ramsey, who translated the English text read for this\ndoc at eighteen and then became the reader whose criticism, in Wittgenstein's words, helped him see\nthe mistakes. The other was Piero Sraffa, an economist, whose criticism he calls the stimulus for *\"the\nmost consequential ideas\"* of the later book. In the story Norman Malcolm tells, Sraffa answered the\nclaim that every proposition shares a logical form with what it describes by brushing his fingertips\nunder his chin, a Neapolitan gesture of contempt, and asking what the logical form of that was. (Von\nWright tells it differently, so take the detail as anecdote.) The point survives either version:\n**the frame was seen from outside, by a reader who did not share it**, exactly as\n[what-human-judgment-is-for.md](what-human-judgment-is-for.md) and the second row of the vision doc's\ngate table (a reader from outside the frame) would predict.\n\nSraffa's gesture is also a move the picture theory cannot hold: an act that says something and pictures\nnothing. Deliberus has already ruled such a move into its ontology. Consent, *\"I do not accept that as a\nstatement of my position\"*, is in the corpus's own words *\"a speech act, not a description of history.\nIt cannot be factually wrong, because it is not asserting a fact.\"* The Tractatus's general form of a\nproposition is *\"Such and such is the case\"* (4.5), and consent is not of that form. Nor is the\nspeaker-attitude hedge that run 5 found no claim type could hold\n([peer-review-and-the-reasoning-layer.md](peer-review-and-the-reasoning-layer.md) R1). Nor is the\nreframe, which is about the form of a region rather than a fact inside it. And when a friend's\nstructured file marked claims UNRESOLVED,\n[mapping-a-structured-position-on-ai-governance.md](mapping-a-structured-position-on-ai-governance.md)\nfound that every one of Deliberus's six epistemic statuses *\"says this is the case, with such-and-such\nconfidence. None of them says this is open.\"* That is 4.5 almost word for word, found from the other\nside. **The places where Deliberus's ontology has hit *does not fit* cluster where the picture theory\nfails: moves that are not pictures of a fact.** That is worth knowing the next time a taxonomy\nbreaks.\n\n---\n\n## 5. Exact and structural echoes\n\n**Fit** says how tight the match is. *Exact*: the same structure, checkable. *Structural*: the same\nshape in a different setting. *Loose*: suggestive only, and not to be built on.\n\n| Tractatus | Deliberus | Fit | What the echo adds |\n|---|---|---|---|\n| **What can be shown cannot be said** (4.1212). Propositions cannot represent the form they share with reality (4.12) | **The sorting test.** A commitment belongs in the ontology if the graph cannot hold a disagreement about it without already assuming it ([what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md) § 3) | exact | The test draws the say/show line mechanically. The third move, *publish it* (the is/ought split as a claim), is Russell's escape in his 1922 introduction: *\"another language dealing with the structure of the first language … to this hierarchy of languages there may be no limit.\"* It narrows the gap step by step and never closes it, because each publication rests on the form that remains (AJ4) |\n| **Identity of the object I express by identity of the sign and not by means of a sign of identity** (5.53). *\"To say of two things that they are identical is nonsense\"* (5.5303) | **Sameness is never an edge.** *Same* makes a merge candidate and writes no propositional link (ruled 2026-09-13, built 2026-09-22) | exact | The ruling has the shape the Tractatus says identity must have: shown by one node, not said by a link. Before it, auto-connect's prompt told the classifier that two claims asserting *\"essentially the SAME proposition\"* should *\"still classify the propositional relation (usually supports)\"*: identity written as a relation ([auto-connect-upgrade.md](auto-connect-upgrade.md) § 2c) |\n| **The essential in a proposition is that which is common to all propositions which can express the same sense** (3.341) | A claim is what survives paraphrase; the reuse flywheel turns on identity being cheap to decide | exact | The book holds that whoever understands two expressions knows whether they mean the same (4.243, 6.2322). Deliberus measures the opposite: sameness is judged, with a ceiling, which is [the-reuse-flywheel-prior.md](the-reuse-flywheel-prior.md)'s *judged bucket*. The more of a claim's identity is structural, the closer it gets to the transparency the book assumed |\n| **Truth-ground relations**: *the sense of \"p\" is contained in that of \"q\"* (5.122); mutual entailment is *one and the same proposition* (5.141); *opposed … if there is no significant proposition which asserts them both* (5.1241); independent propositions (5.152); partial overlap of truth-grounds as a degree of support (5.15) | **The seven keys**: contains, same, opposes, unrelated, overlaps, plus can't-tell | exact on structure; lineage from memory | Both carve the relations between two statements by how their *possible truth-makers* overlap. MacCartney and Manning's relations, which the keys coarsen, are set relations of the same kind *[their descent from the Tractatus through Carnap's state descriptions is from memory; verify before citing]*. The keeper's own gloss on *overlaps* (*establishing one would partly advance the other*) is 5.15's partial containment of truth-grounds |\n| **The confusion between internal and external relations**, *\"very widespread among philosophers\"* (4.122) | The two most-measured edge defects: **14 false support links** between look-alike claims about different referents ([the-remaining-case-for-the-stance-guard.md](the-remaining-case-for-the-stance-guard.md)), and **151 of 197 decomposition edges that are not part-of relations**, 70 of them labelled evidence miscast as parts | structural | Each is a question about the content of two claims (internal: *are these the same? is this a part of that?*) answered with, or mistaken for, a relation that holds because of the world (external: support). The 14 add a second shape from the book: *\"Two different symbols can … have the sign … in common\"* (3.321), mirrored vocabulary about two different regimes. See AJ3 |\n| **The totality clause**: the world is the facts *\"and by these being all the facts\"* (1.11), and the totality *\"determines … also all that is not the case\"* (1.12; also 4.52, 5.524) | **Coherent absence** and the open-world reading: *absence of a record is not a record of absence* ([what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md) § 6c.3) | exact | The precise point where the two differ. The Tractatus needs one extra fact, *these are all*, to fix what is false. A corpus can never contain that fact about itself. Coherent absence is the name for its missing |\n| **Probability is a relation between propositions** (5.15). *\"A proposition is in itself neither probable nor improbable\"* (5.153); *\"an extract from other propositions\"* (5.156); independent propositions give each other **1/2** (5.152) | [starting-values-and-the-problem-of-the-prior.md](starting-values-and-the-problem-of-the-prior.md) § 0: no evidence-free prior; probability is two-place. *A horizon kind is read, never set* | structural | The Tractatus is the one framework in which a start of 1/2 is principled: it has a unique analysis into independent atoms, so the partition is given. Deliberus has no such atoms, so its 0.5 can only mean *undecided*, never *probability one half*. That supports the starting-values doc's verdict that a universal 0.5 is the wrong constant. And 5.156's *extract from other propositions* is the shape of the horizon ruling: a kind read from the claims around it, never set on the claim |\n| **Unanalysed wholes show themselves by indeterminacy** (3.24) | Bundledness detection (`looks_bundled`); decomposition readiness; the sorry marker | structural | Gives a second cue beside length and connectives: a part of a claim that leaves *\"everything not yet determined\"* is a whole not yet cut |\n| **Logical multiplicity**: a proposition must have *\"exactly as many things distinguishable as there are in the state of affairs\"* (4.04), and *\"one cannot get outside it in the representation\"* (4.041) | **Flattening**, and the three-runs law. A friend's file defined a twelve-value ladder of certainty, and the map stored it on six statuses without reading the ladder | structural | Every flattening is a loss of multiplicity. 4.041 says why you only find the loss when a new register arrives: a notation cannot represent its own multiplicity, so the deficit is invisible from inside |\n| **Content is what a proposition excludes.** *\"Tautology leaves to reality the whole infinite logical space\"* (4.463). *\"I know nothing about the weather, when I know that it rains or does not rain\"* (4.461) | **Exposure** (*\"Being vague is safe. We're making clarity pay.\"*); the motte; **leveling dissolution** | exact | One account explains three findings. A vague claim excludes less, so less can refute it. The motte retreats towards the substanceless centre (5.143). Leveling dissolution widens a word until it excludes nothing, and the false-dissolution test checks whether exclusion survived |\n| **No new symbol *\"with an entirely innocent face\"***; if a device is needed in one place, *\"we must immediately ask: Where is this expedient always to be used?\"* (5.452) | `REPORTS` was non-propagating only by accident (no scheme on the edge) until 2026-09-17; the rule to remove an upstream pass's answer from the downstream menu in the same commit; the upkeep rules | exact | The corpus's own discipline, 104 years earlier: a device's position must be settled everywhere when it is introduced |\n| **A primitive idea must be introduced in all contexts at once**, or *\"it would then remain doubtful whether its meaning in the two cases was the same\"* (5.451) | **The collision check**, headline test 3: does this word already do a job here? (*scaffold*, *semantic*) | exact | — |\n| **The application decides the atoms; logic cannot anticipate it** (5.557) | The three-runs law; `does_not_fit`; *settle it first is a closed path* | exact | § 4a |\n| *\"Men have always thought that there must be a sphere of questions whose answers — a priori — are symmetrical and united into a closed regular structure\"* (5.4541) | The scheme set: Wagemans's a-priori table covered **less** than the curated open set on the same corpus (17% left over against 3%) | structural | Wittgenstein allows a closed structure in logic alone. Outside it the corpus measured the closed system losing |\n| **Truth tables** (4.31, 5.101) | `hinge.py`: clamp one premise to granted, then denied, and measure the swing | structural | A gradual truth-table row |\n| **\"A says p\" is of the form \"'p' says p\"**; p does not occur in it as an argument whose truth matters (5.541–5.542). Frege's assertion sign *\"only shows that these authors hold as true the propositions marked in this way\"* (4.442) | **Reported speech and always-mint**: the report and the reported content are separate nodes, and `REPORTS` carries no strength | exact | Wittgenstein dismisses *who holds it* as logically idle. Deliberus makes it load-bearing: attribution, asserted versus minted, *the account that made a contribution is not always the one who asserts it*. The book is right that *A says p* is not a truth-function of p, and that is the non-propagation rule |\n| *\"What the signs conceal, their application declares\"* (3.262) | **Downstream divergence** as the split alarm for a sense | exact | — |\n| **Only in the context of a proposition has a name meaning** (3.3) | A sense is read through the claims that use it (*lead a sense with its attributed claims*, [core-periphery-and-the-concept-layer.md](core-periphery-and-the-concept-layer.md)) | structural | — |\n| **The denial's place is fixed by the place of what it denies** (4.0641) | Latent scaffolding: critical-question polarity pairs rendered, not stored | structural | Storing both p and not-p as independent nodes counted one logical place twice |\n| **The cube seen two ways: *\"we really see two different facts\"*** (5.5423) | The **reframe** | structural | Grounds AJ2 |\n| **Any mesh describes the picture; which mesh describes it more simply *\"says something about the world\"*** (6.341–6.342) | The decomposition axes; the ontology as a chosen mesh; *a borrowed category earns its place by yielding a move* | loose | — |\n| **Logical forms are anumerical, so there is no monism or dualism** (4.128) | Founder, 2026-08-21: *\"Whether it's one valence scale or many feels irrelevant to me.\"* | loose | The founder's correction has an ancestor: counting the primitives is the wrong axis |\n\n## 6. What the reading counts against\n\nThe convictions sweep allows the verdict *Lowered* and uses it, so this section looks for where the\nTractatus counts against the corpus, not only where it agrees.\n\n**One. A shared bottom that excludes nothing is not convergence.** The Tractatus has a convergence\nthesis of its own, and a strong one: every picture of every kind shares one logical form (2.18). It is\nguaranteed, universal, and says nothing. 5.143: *\"Tautology is that which is shared by all\npropositions, which have nothing in common with one another … tautology their substanceless centre.\"*\nThis is a warning the wager has not yet written down. **The deepest things all positions share might\nbe exactly the things that rule nothing out.** *\"Consciousness matters, suffering is real\"* is\nconvergence only if it excludes some position someone actually holds, or helps choose between\noptions. Otherwise the descent has reached the substanceless centre, and the wager has won a\ntautology. The accounting already makes ceilings count against convergence. This adds the mirror rule:\n**a shared bottom counts towards convergence only when it rules out a position someone holds.** The\nfalse-dissolution test is the same check one level up. See AJ5.\n\n**Two. Values may not be propositions at all.** [convergence.md](../convergence.md)'s Lean table puts\n*\"Axioms: human values (convergent at depth)\"* in the kernel's slot. The Tractatus is the strongest\nprior that nothing propositional can go there: *\"Hence also there can be no ethical propositions\"*\n(6.42). The corpus has met this from the late Wittgenstein's side (hinges may be non-propositional,\n[the-load-bearing-unsaid.md](the-load-bearing-unsaid.md)) and keeps an honest door for it, the\n*structural* horizon, a disagreement about what normativity is. What it **Lowers** is the reading in\nwhich the bottom is a set of shared value *statements*. What survives is the sharpened wager already on\nrecord: a small shared basis, which may be partly non-propositional and reached by attunement as much\nas by analysis. One position the graph cannot yet hold: a person whose answer to *\"is torture wrong?\"*\nis *\"that is not the kind of sentence that can be true or false\"*. The graph offers agree, disagree,\nsupport, attack and split, and none of these says *this is not a proposition*. Today that position\ncould arrive only through the reframe act, which is not yet built.\n\n**Three. A falling discovery rate can mean someone stopped looking.** The gate's exit criterion is that\nreading stops overturning things ([vision.md](../vision.md) § Why the meta layer is the gate). The\nTractatus is the cautionary case. From 1918, when he declared the problems *\"in essentials … finally\nsolved\"*, Wittgenstein's rate of discovered mistakes stayed near zero for most of a decade. That was\nbecause he had left philosophy to teach in village schools, so almost nobody was attacking the book\nwith him, and he was not attacking it either; it was not because the work was sound. It restarted on\ncontact: Ramsey's visit in 1923, conversations in Vienna later in the decade, and his return to\nCambridge in 1929, the year he conceded the colour problem in print. The vision doc already words the criterion as\n*\"when concentrated attack on the foundations stops turning up defects\"*. **The Tractatus shows why the\nword *attack* is load-bearing.** Read the rate with the volume of attack as its denominator, since a\nrate with no attack behind it looks identical to soundness. This is *a guard that never fires looks\nexactly like a guard that works*, applied to the gate itself. See AJ6.\n\n**And one thing it Raises.** The meta-scaffolding passage calls this work *\"never-quite-final\"*. The\nTractatus is the strongest available evidence that this is right. Its author, among the most careful\nthinkers of the century, called his foundations definitive and was wrong within eleven years. The\nfounder's own diagnosis of the horizon vocabulary (*the names photographed the sitting*, in Claude's\nphrasing of his point) describes how the Tractatus was written: by one person, finished in 1918\nduring the war, naming what was at the bottom of one long sitting.\n\n**Its afterlife fits the success prior's fourth consequence**: partial success looks like assimilation.\nThe system failed. Its parts became standard equipment: the truth table, logical space, picturing as\nmodelling, and logical probability *[that last lineage from memory]*. That is the hope and the warning\nthe success prior names. Components spread without the discipline they were built for, so *be the\nreference implementation*. And the Tractatus is a member of the calculemus line that the success prior\nnames as its reference class of failures. It fails condition (d) in a revealing way. It says that\n*\"All propositions of our colloquial language are actually, just as they are, logically completely in\norder\"* (5.5563). Yet its remedy for confusion is a new sign-language that obeys logical syntax\n(3.325). Deliberus's inversion principle settles that tension the other way: people speak as they\nspeak, and the structure is output. **The Tractatus postulated an analysis that nobody ever performed.\nDeliberus performs an analysis whose end it does not postulate.**\n\n## 7. Proposals (none ruled; ledger Arc AJ)\n\n**AJ1 — The framing.** *Deliberus pursues the Tractatus's ambition (clarification as an activity, every\nclaim taken apart into its parts, confusion about words found and named) without three of its\nassumptions: permanent atoms with one complete analysis, a limit of the sayable fixed in advance with\nsilence beyond it, and a single mind as the reasoner.* Proposed as an internal orientation and as a\nsentence for research and outreach to philosophers. **Not** proposed for the landing: it fails the\nstranger test, and quoting Wittgenstein to borrow authority is the move the corpus forbids for\nassembly theory.\n\n**AJ2 — The reframe as a rival map: shown, not said.** This answers part of the founder's open question\nof 2026-09-10: *\"we need to revisit how to interpret any response to the reframe invite, so as to most\nappropriately make use of the contribution.\"* The current design mints the reframe as a claim with a\n`REFRAMES` edge. That records it, but 2.172 explains why the frame stays hard to see when said. A frame\nis the form of a picture, and a picture cannot represent its own form. One more interpretation of a\nresponse is therefore available: **a rival decomposition of the same root**, a second map drawn under\nthe other frame. The minimum-wage example from that ruling shows it: one map as an employment debate,\na second as a bargaining-power debate. The difference between the two maps shows the frame, the way the\ncube shows its two facts (5.5423). Both stand, since contributions are added and never overwritten,\nand each part of the rival map is an ordinary, attackable claim. This composes with *never ask a mind\nto audit its own frame*: the rival map comes from someone standing elsewhere. It does not replace the\n`REFRAMES` claim; it is a second way a response could land. Nothing built yet groups a claim's parts\ninto two rival sets, so that grouping is the piece this would need.\n\n**AJ3 — Classify every edge type as internal or external.** An *internal* relation holds by the content\nof the two claims alone: identity, containment, part-of, opposition. An *external* relation holds\nbecause of the world: support and attack through a scheme and evidence. The keeper answers internal\nquestions, and the scheme classifier answers external ones. **The rule: an internal question must never\nbe answered by writing an external edge, or the reverse.** The two most-measured edge defects in the\ngraph were each this confusion (§ 5, the row on 4.122), and the sameness-before-relation ruling already\nenforces it for identity. The proposal extends it to part-of versus evidence-for. The census of\n2026-09-22 labelled 70 of 197 decomposition edges as evidence miscast as parts; a hand reading on\n2026-09-27 found that label reliable only for *not a part*, and left *which relation each pair holds*\nas the keeper's question. Cheap form: one field on each edge type naming its kind, and a test that no\npass writes an edge of the wrong kind for the question it was asked.\n\n**AJ4 — The reflexivity gap narrows and never closes.** The red team's *operator-reflexivity gap* (the\nontology lives in code, not in the graph) and founder question 2 (*should Deliberus host its own\ndesign?*) get an answer from 4.12 and Russell's introduction together. **Publishing commitments as\nclaims can go on indefinitely, and the last layer always stays shown, never said**: the claim and the\nattack edge themselves can be published only by using them. So the epistemic constitution's final\nauthority sits with people outside the graph, as Ostrom's participation in rule-making already implies.\nThis argues for move 3 in [what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md) § 5,\ntaken one commitment at a time. It argues against expecting it ever to finish. It does not lift the\nfounder's hold on self-hosting fractal priors.\n\n**AJ5 — Count a shared bottom as convergence only when it excludes something.** § 6, one. It would sit\nbeside *ceilings count against convergence* in [convergence.md](../convergence.md) § What Would\nDisprove This. The test: does the shared claim rule out a position someone actually holds, or decide\nbetween options someone faces? If not, the wager has reached the substanceless centre, which is not a\nfinding.\n\n**AJ6 — Read the gate's discovery rate against the volume of attack.** § 6, three. Report defects found\nper unit of concentrated attack, not per period. Then a quiet month without attack reads as *no data*,\nnot as *nearly done*. `scripts/log_arc.py` counts corrections; the denominator would be a count of\ndeliberate attack passes, such as stress tests, red-team rounds and outside reads.\n\n**AJ7 — An epigraph for star fifteen.** The constellation's next star is *\"Never ask a mind to audit\nits own frame.\"* Its sibling, star fourteen, carries *Philosophical Investigations* § 129 as its\nepigraph. Two candidates from the Tractatus, both from 1918, state the reason the tile gives:\n*\"But the picture cannot place itself outside of its form of representation\"* (2.174), or *\"But you do\nnot really see the eye\"* (5.633). The first is the more exact and the second the more vivid. The\nfounder rules; the wording in `web/src/lib/constellation.ts` changes only in the same commit as a\nruling.\n\n## 8. What this does not license\n\n- **No borrowed authority.** *\"Wittgenstein showed that…\"* never goes into a funding document or onto\n  the landing page. He disowned much of this book, and the corpus's rule for assembly theory applies:\n  use a mechanism where it is independently attested, and never the name as a warrant.\n- **The echoes are structural, not lineage**, except where marked. Wittgenstein through Russell into\n  `decompose_claim.py` is documented. The rest are parallel arrivals. That is evidence the shapes are\n  real, and no evidence that the corpus learned them from him.\n- **The reading of the book's end is contested**, and this doc does not settle it. On the reading the\n  corpus implicitly uses here, the Tractatus believes in an inexpressible that shows itself. The\n  *resolute* reading (Diamond, Conant) holds that the nonsense is plain nonsense, with no ineffable\n  truth behind it. On that reading proposition 7 is the ladder thrown away, not a stop sign *[from\n  memory; verify before citing]*. Either reading leaves § 4c's contrast standing, since the corpus\n  refuses both a fixed limit and a fixed silence.\n- **Nothing here changes a strength, a prompt or a tile.** AJ3 and AJ7 would, once ruled.\n\n## 9. Beside the twelve books (added 2026-10-12; a session's reading, unruled)\n\nA reading of twelve books from a video was finished the day before this note was merged, and neither\nnamed the other. Ten places where they meet are set side by side in\n[twelve-books-in-plain-words.md](twelve-books-in-plain-words.md) Part 3. Three bear on this note:\n\n- **AJ3 gains a test and a history.** In Lakatos's account of how a theorem is repaired, a step of the\n  argument is written into the claim as a condition. One sentence moves from an external relation (a\n  reason for the claim) to an internal one (a part of it). That is why the two were confused in the\n  linking step, and his test tells them apart: if this sentence falls, is the claim false, or is only\n  this argument for it gone?\n- **4.04 has a second statement.** Ashby's law of requisite variety says of a regulator what logical\n  multiplicity says of a notation: it needs as many distinctions as what it faces. Both give the\n  reason a *does not fit* answer has to be built in beforehand.\n- **The later Wittgenstein has a working form in the corpus's own terms.** Lakoff's account of a word\n  as several tests that usually agree starts from family resemblance (his second chapter is *From\n  Wittgenstein to Rosch*), and it yields a move: ask about the cases where the tests come apart.\n\nAlso there: Toulmin, whose model of an argument the corpus leans on, was Wittgenstein's student in\n1941 and in 1946 to 1947.\n\n## Sources\n\n**Primary**: Ludwig Wittgenstein, *Tractatus Logico-Philosophicus*, with an introduction by Bertrand\nRussell (Kegan Paul, 1922), German and English on facing pages. Project Gutenberg #5740, read in full\nfrom the TeX source. Proposition numbers are the book's own.\n\n**Checked 2026-10-06 against outside sources**:\n- The Ficker letter of 1919: [Wikiquote, *Ludwig Wittgenstein*](https://en.wikiquote.org/wiki/Ludwig_Wittgenstein).\n- The *Investigations* preface (*\"grave mistakes in what I wrote in that first book\"*, the debts to\n  Ramsey and Sraffa): [Cambridge, *From the Tractatus to the Investigations: two prefaces*](https://www.cambridge.org/core/books/abs/wittgensteins-philosophical-investigations/from-the-tractatus-to-the-investigations-two-prefaces/E657C6BF356523D6F6FB9576F9E93311);\n  [the Investigations in English (PDF)](https://static1.squarespace.com/static/54889e73e4b0a2c1f9891289/t/564b61a4e4b04eca59c4d232/1447780772744/Ludwig.Wittgenstein.-.Philosophical.Investigations.pdf).\n- *Investigations* § 47 on the chair: [Hauptli's introduction to the *Investigations*](https://faculty.fiu.edu/~hauptli/IntroductiontoWittgenstein'sPhilosophicalInvestigations.html).\n- Ramsey's translation at eighteen, the October 1923 *Mind* notice and the 1923 visit: [Wikipedia,\n  *Frank P. Ramsey*](https://en.wikipedia.org/wiki/Frank_P._Ramsey);\n  [SEP, *Frank Ramsey*](https://plato.stanford.edu/entries/ramsey/);\n  [PhilPapers record of the notice](https://philpapers.org/rec/RAMCNO).\n- Colour exclusion, Ramsey's objection, the 1929 concession, the absence of examples of atoms, and\n  Russell's 1918 credit to Wittgenstein: [SEP, *Wittgenstein's Logical Atomism*](https://plato.stanford.edu/entries/wittgenstein-atomism/).\n- The Sraffa gesture, with von Wright's variant: [Wikipedia, *Piero Sraffa*](https://en.wikipedia.org/wiki/Piero_Sraffa)\n  (citing Malcolm, *Ludwig Wittgenstein: A Memoir*, pp. 58–59).\n\n**From memory, flagged in the text**: Carnap's state descriptions and logical probability as\ndescendants of 4.27–4.45 and 5.15; MacCartney and Manning's relations as set relations of the same\nkind; the resolute reading (Diamond, Conant).\n\n## See also\n\n[depth.md](../depth.md) § The Truly Ineffable · [convergence.md](../convergence.md) § What Would\nDisprove This · [what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md) § 5 ·\n[self-similar-decomposition-and-claim-ontology.md](self-similar-decomposition-and-claim-ontology.md)\n§ Wittgenstein's Objection · [claim-sameness-philosophical-readings.md](claim-sameness-philosophical-readings.md) ·\n[what-human-judgment-is-for.md](what-human-judgment-is-for.md) § Act two ·\n[the-load-bearing-unsaid.md](the-load-bearing-unsaid.md) (the late Wittgenstein's hinges) ·\n[philosophical-foundations-stress-test.md](philosophical-foundations-stress-test.md) § 7\n"}