{"path":"research/corpus-graph/readers/2026-10-10-strength-and-ontology-notes-as-a-reader-returned-it.md","content":"# Other notes that bear on nine notes about strength, value and the ontology (2026-10-10)\n\nOne of three read-only readers (Sonnet) sent on 2026-10-10 in a second round, after the founder said of the\nconnections that neither note states: *\"I want more of these.\"* This reader was given nine notes on how\nstrength is computed, how values are weighed and what the ontology commits to, most of them holding unruled\nproposals, and was asked which other notes bear on a specific passage of each and are not named by it. It\nwas also asked to mark a finding as `same-mechanism-elsewhere` where the other note describes the same\nmechanism in another field or at another scale; it marked none, and lists four such pairs it cut for space\nin its last paragraph. It read all nine in full (its own count), with 119 tool calls. Its report is\nbelow, unchanged.\n\nWhat was kept from it, after both passages of each finding were opened and the quoted words checked by code,\nis in `../found.jsonl` (lines whose `where` names this file). Why readers were sent, and what the graph does\nwith what they find: [../README.md](../README.md) § 9.\n\n---\n\n9 of 9 notes read in full.\n\n### F1\ntarget: docs/research/what-belongs-in-the-ontology.md:344\ntarget_words: \"Founder question 2, open. This doc argues the three moves are ordered by cost, not that the third should be taken.\"\nolder: docs/research/session28-the-harness-arc-and-the-source-intake.md:39\nolder_words: \"beliefs published into the graph, operator-reflexivity moves) until the open reviews are much\"\nkind: narrows\nwhy: The note leaves publishing a further commitment as a claim (the QEM choice, the four-type taxonomy, the Legibility Rule) as an open question for the founder. On 2026-08-30, eight days after the is/ought claim was seeded, he held all self-hosting: no project beliefs published into the graph and no operator-reflexivity moves until the open reviews are much closer to done. Ledger Q28's second layer waits behind the same hold. Move three is therefore under a standing hold, not just unruled.\nstrength: strong\n\n### F2\ntarget: docs/research/what-belongs-in-the-ontology.md:230\ntarget_words: \"**The renames this implies** (all held for ruling; ledger Arc O):\"\nolder: docs/research/session33-the-horizon-vocabulary.md:71\nolder_words: \"the rename table's first word, *dissolved* → *definitional*, with *terminus → horizon kind* and *verdict → reading* agreed as following from loop 1\"\nkind: already-answers\nwhy: The note calls the whole rename table held for ruling. Ledger O1, O2 and O6 record the founder's answer of 2026-09-16 (agreed, and choose definitional) to terminus to horizon kind and verdict to reading, put as settled unless he objected. So dissolved to definitional, terminus to horizon kind and verdict to reading are ruled; is_bedrock to is_open, the settler-based buckets (O7) and the dated reading (O8) stay open. The code rename of terminus is a parked migration. Do not re-open the ruled rows.\nstrength: strong\n\n### F3\ntarget: docs/research/parity-and-the-zone-interior.md:91\ntarget_words: \"Each side grants the other's landing as permissible.\"\nolder: docs/research/session33-the-horizon-vocabulary.md:49\nolder_words: \"A grant is a bare judgment. Permissibility is read instead from what survives attack.\"\nkind: later-correction\nwhy: Criterion 3 and the one-line boundary make a zone depend on each side granting the other's landing. The next day the founder struck the grant as a rare, bare judgment (ledger O4): a zone is read from sibling option claims, each examined and still standing, with the cross-attacks as the record that nothing selects. Mutual acknowledgement is no longer the test, and cannot be observed. The seven candidates (Bach, the minimum wage) should be re-read by standing of the options, not by grant.\nstrength: strong\n\n### F4\ntarget: docs/research/relevance-and-the-two-factors.md:58\ntarget_words: \"run on all 197 part links (mean 0.33). A switch meant to let that score lower\"\nolder: docs/research/auto-connect-upgrade.md:118\nolder_words: \"The census figure of 76% (151 of 197 part-of links not parts) pooled this step's 84 with everyone else's.\"\nkind: later-correction\nwhy: The mean of 0.33 is reported for 197 part links. The census correction records that 84 of the 197 were the automatic linker's premise links, stored backwards: the judge was asked whether a conclusion is a piece of its premise and said no 84 times of 84. Of the 113 written the right way round, 59% are not parts. Anyone wiring this score into the cap, or using it as a presumed requirement weight, would calibrate on a pooled set that is mostly not parts.\nstrength: strong\n\n### F5\ntarget: docs/research/weakest-link-arithmetic-and-the-merge-hunch.md:693\ntarget_words: \"Replace the flat presumption on person-added parts with a machine relevance proposal.\"\nolder: docs/research/jev-and-the-judgement-layer.md:612\nolder_words: \"person adds presumes relevance 0.5 under the build. A confident Jev *does not bear* could lower that\"\nkind: supplies-a-missing-piece\nwhy: The note proposes a machine relevance proposal in place of the flat 0.5 on person-added parts, so a bogus part starts near zero. The decision-model note measured a cheap judge on this move (ledger W5, citing O13): confidently not related on 83 of 100 random cross-source pairs with no confident mistake, but only 0.17 to 0.52 on the same-topic pairs that wrote bad part links (its lines 325-328). It also says a machine lowering a person's contribution needs a ruling of its own.\nstrength: strong\n\n### F6\ntarget: docs/research/weakest-link-arithmetic-and-the-merge-hunch.md:394\ntarget_words: \"dataset add three times in the corroborative channel. Pollock's Fredonia witnesses make the\"\nolder: docs/research/what-counts-as-evidence.md:438\nolder_words: \"three studies of one effect are **corroborative witnesses**\"\nkind: contradicts\nwhy: The note pairs the corroborative channel with contributions that add, so three studies add three times. The evidence note, using Schum and ledger Q22, defines corroborative as saturating (its table at line 426): three studies of one effect are corroborative witnesses that saturate, never necessary parts. The code docstring says the opposite, three studies finding one effect accumulate. One word, opposite arithmetic. The tag backfill and Q22 must decide which meaning the tag carries before three-study cases are tagged.\nstrength: strong\n\n### F7\ntarget: docs/research/scheme-bounded-decomposition-and-evidence-as-subgraph.md:138\ntarget_words: \"This is scheme-bounded decomposition: bounded by the scheme's finite CQ set, not recursive to infinity.\"\nolder: docs/research/drowning-in-claims.md:30\nolder_words: \"all claims are machine-minted critical-question scaffolding (1,741 substantive of 4,852; the\"\nkind: narrows\nwhy: The note argues that scheme-bounded decomposition escapes the explosion of full recursion because each scheme has a finite question set. On the live graph, machine-minted critical-question scaffolding, a positive and a negative claim per question, is 64% of all claims (latent-scaffolding-and-the-type-token-split.md:29), about five hundred nodes per extraction, and dominated the reading surface. The founder's answer was latent-until-touched scaffolding. Finite did not mean small, and the note's cost table counts calls, not minted nodes.\nstrength: strong\n\n### F8\ntarget: docs/research/value-weighting-decomposition.md:218\ntarget_words: \"suggests a large fraction of practical weighting disputes will move under decomposition without ever touching a residue\"\nolder: docs/research/empirical-deliberation-convergence-evidence.md:89\nolder_words: \"Here even 80 hours of best-practice adversarial collaboration between mutually-respecting experts moved almost nothing.\"\nkind: narrows\nwhy: The note says layers 1 to 4, chiefly compressed empirical belief, carry most real disagreement, so typed residues are a minority. The deliberation-evidence note, written the same day, reports the best case: 80 hours of adversarial collaboration between respectful experts moved the medians almost nothing, and what stayed was worldview priors and weightings of acknowledged considerations. It admits the sample may be too abstract (its line 111). The minority-residue claim is unestablished for weighting disputes.\nstrength: strong\n\n### F9\ntarget: docs/research/parity-and-the-zone-interior.md:24\ntarget_words: \"The zone's afterlife is a growing, attributed library of traversals.\"\nolder: docs/research/value-weighting-decomposition.md:69\nolder_words: \"what people stably carry around is not a weight vector but a loose set of partially articulated considerations\"\nkind: narrows\nwhy: The ruling builds a library of attributed traversals showing which reasons and values weigh more for whom. The weighing research reports that weights are constructed at elicitation: they shift with attribute splitting, range and method, so what people stably hold is considerations, not a weight vector. Stored per-person weightings would record elicitation artefacts and mislead the next traveller unless marked constructed (its CQ3) or kept as considerations. The parity note's fallibility caveat covers causes and satisfiers, not the weights.\nstrength: strong\n\n### F10\ntarget: docs/research/weakest-link-arithmetic-and-the-merge-hunch.md:382\ntarget_words: \"evidence edge mislabelled as decomposition (three studies are witnesses, not parts).\"\nolder: docs/research/linked-convergent-and-what-we-are-actually-asking.md:37\nolder_words: \"The prediction holds on the provenance it was aimed at.\"\nkind: already-answers\nwhy: Section 8 predicted that a corroborative-looking child is usually evidence mislabelled as a part. The pass then ran on all 197 edges and the prediction held where aimed: 20 of 30 extraction-written edges are evidence, not parts, and genuine parts were found for only 45 of 197 (the 84 backwards linker links need setting aside, as later corrected there). So the tag backfill would tag a minority and the work is re-filing. No human has set a tag, so accuracy cannot be scored.\nstrength: strong\n\n### F11\ntarget: docs/research/value-weighting-decomposition.md:228\ntarget_words: \"treat the decomposition schema as an arguable claim in the graph, offer multiple decomposition frames in the spirit of worldview lenses\"\nolder: docs/research/what-belongs-in-the-ontology.md:72\nolder_words: \"The is/ought partition is now a published claim\"\nkind: already-answers\nwhy: The note's strongest objection says the decomposition schema smuggles in a metaethic, and its answer, publish the schema as an arguable claim, is left as a concession. The ontology note ran that answer: a representability test (its section 3), then the is/ought partition published as a claim with the hedonic-monism claim attacking its ground, founder-ratified 2026-08-22. It also shows Dancy's particularist rendered as an unsupported value premise. The objection is partly answered, and located.\nstrength: medium\n\n### F12\ntarget: docs/research/weakest-link-arithmetic-and-the-merge-hunch.md:23\ntarget_words: \"alternative lines of defense | Pollock's own rule for separate arguments | noisy-MAX (Henrion 1989)\"\nolder: docs/research/ai-risk-for-mathematicians-and-the-coordination-game.md:107\nolder_words: \"It takes the maximum, so the strongest route decides and extra weak\"\nkind: narrows\nwhy: The table pairs strongest-part-decides with alternatives and noisy-MAX. A later essay reading shows the mismatch: the essay's central move is noisy-OR, several partly independent routes making an outcome more likely, while the alternative channel takes the maximum so extra weak routes add nothing. It concludes both rules are right about different quantities, argumentative standing and event probability, and that a probability OR is not clone-proof until sameness is ruled and shared premises are pulled out.\nstrength: medium\n\n### F13\ntarget: docs/research/session39-what-the-graph-stopped-claiming.md:105\ntarget_words: \"when nothing has checked a recorded fact, store the blank that says so, never a plausible stand-in\"\nolder: docs/research/what-belongs-in-the-ontology.md:66\nolder_words: \"it stops forced nearest-fit, which is the classification layer's silent failure mode\"\nkind: same-idea-earlier\nwhy: Q30 states \"store the blank, never a plausible stand-in\" as a generalisation from the day's repairs. The ontology note stated the same discipline for closed enums on 2026-08-22: an explicit does-not-fit value stops forced nearest-fit, and the blank rate becomes an instrument, taxonomy pressure. Q30's five instances omit those shipped enum cases. The earlier design adds what Q30 lacks: count the blanks and score them against claimed coverage.\nstrength: medium\n\n### F14\ntarget: docs/research/what-belongs-in-the-ontology.md:339\ntarget_words: \"Never a coloured label that reads as a category the claim belongs to; never a count by kind\"\nolder: docs/research/mapping-a-structured-position-on-ai-governance.md:47\nolder_words: \"its author gave it**, computed and propagated, not merely worded strongly.\"\nkind: supplies-a-missing-piece\nwhy: The note proposes never showing a coloured kind-label, or a count by kind, without date and unexamined count. A measured, author-caught case: a friend read the map of his own position and said his careful typing had gained weight. Of 116 claims, 20 he typed as something other than observed or claim were stored observed; a hypothesis so stored carried 0.30 more computed confidence. Loudness is numeric and propagated, so the rule must cover derived numbers and stored defaults, not only labels.\nstrength: medium\n\nNoticed: (1) The nine notes' inbound/outbound naming was checked by rg on `.md` names; `corpus_graph.py names` lists where a note is named, not what it names. All 14 pairs above were checked: older note's file name absent from the target, and no found.jsonl record holds both names. (2) The squirrel note (AD22/AD23/AD26) yielded no finding: those rows' older bearings are already mined in found.jsonl and the 2026-10-08 listening note; Q22 is the unmined row, and F6 touches it. (3) Pairs read at both ends but cut for space, medium or kinship: decomposing-value.md:88 (a gap between knowing and being moved stays open) against a-cognitive-ecology-and-the-terms-of-the-relation.md:196 (people are the counterexample: commons are kept by rules, monitoring and penalties, not insight); relevance-and-the-two-factors.md:149 (the model name recorded on the edge) against storing-the-record-of-machine-judgments.md:131 (a ruled design, Z1/Z11, replaces per-pass strings with one judgment pointer, which shapes where W3's two conditionals would live); parity-and-the-zone-interior.md:24 (attributed library of traversals) against blobs-community-sandboxes-and-the-advice-process.md:37 (the advice-process outcome as an attributed weighing, the same structure with a live community test-bed); value-weighting-decomposition.md:184 (Argument from Weighing as \"V1 outweighs V2\") against wild-weighing-dialects-and-the-question-set.md:11 and :26 (almost no wild sentence has that form; the CQ set grew from 10 to 14). (4) what-belongs-in-the-ontology.md:200 also says \"all held for ruling\" for its first bullet, which ledger O1 rules.\n"}