{"path":"research/corpus-graph/readers/2026-10-10-convictions-and-public-pages-as-a-reader-returned-it.md","content":"# Other notes that bear on five public pages and six notes about convictions and daemons (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 eleven files: five\npublic overview pages (the vision, convergence, depth, worldview-lenses and competitive-landscape pages) and\nsix research notes (ledger rows M16, O8, P9, R9, M19, R1, R14, R13, S1 to S7, AD27). For the public pages it\nwas asked above all for a research note whose ruling, measurement or correction makes a sentence on the page\nout of date or stronger than the record supports. It was also asked to mark a finding as\n`same-mechanism-elsewhere` where the other note describes the same mechanism in another field or at another\nscale; it marked none. It read all eleven in full (its own count), with 120 tool calls. Its report is below,\nunchanged.\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\nIt was told to leave aside the sections dated 2026-10-10 in the major-transitions note and the analogy-daemon\nnote, which were being written that day by the session that sent it.\n\n---\n\n11 of 11 notes read in full (note 9's 2026-10-10 additions ignored, as instructed; note 9 was edited while I worked, so its line numbers below are for the current file).\n\n### F1\ntarget: docs/research/the-analogy-daemon.md:70\ntarget_words: \"Nothing it produces propagates strength until ratified (and the standing-of-untouched-machine-\"\nolder: docs/research/the-ratification-record-and-the-triage-feed.md:185\nolder_words: \"The graph has contributions and telemetry, nothing else. Some contributions have a sole\"\nkind: already-answers\nwhy: The daemon gates its output on ratification and calls the standing of untouched machine structure founder-open. The next day (2026-08-31) the founder blessed an end-state with no ratify act. Assent is telemetry, dissent must arrive as an argued contribution, and untouched machine structure keeps machine provenance as its permanent default. So the gate cannot be a ratification click. It has to be provenance display plus the later rule that a machine builds on machine output only as deep as its measured error allows (T1, ruled), with a person engaging first when no rate is measured.\nstrength: strong\n\n### F2\ntarget: docs/research/conviction-and-critique.md:158\ntarget_words: \"Is it provisional forever until a human certifies it, or can it **earn** standing another way\"\nolder: docs/research/the-ratification-record-and-the-triage-feed.md:209\nolder_words: \"Machine discernments carry machine provenance forever, displayed as such,\"\nkind: already-answers\nwhy: The note lists the standing of untouched machine structure as an open founder question and says no general rule exists. The end-state blessed 2026-08-31 supplies one. Untouched is the permanent honest default, challengeable forever and never treated as human-grade. No certifying act exists to wait for. Later notes add how standing is earned: a measured error rate per pass (T1, ruled) and the trust ladder. Someone ruling on this question would start from those, not from \"provisional until certified\".\nstrength: strong\n\n### F3\ntarget: docs/research/the-ratification-record-and-the-triage-feed.md:233\ntarget_words: \"authorship provenance plus examination telemetry. C8 wants re-ruling in that light.\"\nolder: docs/research/what-confirmed-means.md:4\nolder_words: \"which settles Q29 and makes C8 moot; build in `TODO.md`\"\nkind: later-correction\nwhy: The record asks for the rename of `confirmed` to `ratified` (ledger C8) to be re-ruled. That happened on 2026-09-28. The founder was asked to retire the field and give each of its jobs a home, and said yes (ledger Q32), so C8 is moot. The later note also measured that the flag's value is fully fixed by who wrote the link, and that it gates strength and housekeeping. Anyone building the examined/accepted states of section 2 should start from the retirement, not from a rename.\nstrength: strong\n\n### F4\ntarget: docs/research/graph-daemons-design-space.md:51\ntarget_words: \"Anything that mints a claim, merges concepts, classifies a terminus, creates a support or attack edge, or decomposes.\"\nolder: docs/research/earning-the-right-to-write.md:353\nolder_words: \"trust the splitting and we can as you mention always mark them subtly to be able to reverse if need\"\nkind: later-correction\nwhy: The tier list puts decomposition in propose-only. On 2026-09-21 the founder said to trust the splitter and mark its parts so they can be reversed (\"Wire and switch first, sitting later\"). The ladder note's table now lists the split proposer as writing-marked since that date. It also sorts passes by one test: does the output give a reader more places to disagree? Merge and synthesis fail it. The same note answers \"a daemon that writes is an unattributed author\" with the mark. Someone building on the tier list would hold back a pass that is already switched on.\nstrength: strong\n\n### F5\ntarget: docs/research/major-transitions-heredity-and-evolvability.md:228\ntarget_words: \"So the **disagreement-preservation instrument is not only a flattening alarm;\"\nolder: docs/research/synthesis-build-plan.md:115\nolder_words: \"A synthesis does not flatten that way. It flattens by dropping one side of a conflict.\"\nkind: narrows\nwhy: The note proposes promoting the disagreement-preservation instrument to the project's fidelity meter across hops. The synthesis build found it measures paraphrase flattening. Pointed at a synthesis it would have produced a number that looked like coverage and measured nothing about selection. A new instrument, conflict_coverage, had to be built for that hop. So a fidelity meter is a family of per-hop instruments, and the pre-registered multi-hop drift prediction needs one for each hop.\nstrength: strong\n\n### F6\ntarget: docs/research/the-ratification-record-and-the-triage-feed.md:72\ntarget_words: \"and nothing the instruments already hold with measured confidence.\"\nolder: docs/research/the-evals-loop-and-our-machine-passes.md:335\nolder_words: \"A feed built only from the machine's own doubts shows what the machine already doubts.\"\nkind: supplies-a-missing-piece\nwhy: The triage feed is to show only machine-uncertain items and exclude anything held with measured confidence. The evals-loop reading warns that cases picked by a signal reveal only the faults that signal already knows. It recommends some randomly drawn cases in every batch. A feed built as described never samples the confident region, so confident errors stay invisible to every person. The design needs a random-draw channel from the confident bulk, and the note does not have one.\nstrength: strong\n\n### F7\ntarget: docs/research/the-ratification-record-and-the-triage-feed.md:202\ntarget_words: \"the historical record already contains machine discernments later corrected; time-to-first-objection\"\nolder: docs/research/earning-the-right-to-write.md:393\nolder_words: \"every `written` split was discarded on load; from the build to the fix, *time to first objection* could not start.\"\nkind: supplies-a-missing-piece\nwhy: With seeded probes dropped, the note relies on natural mistakes and time-to-first-objection to show continuously whether anyone is looking. A measured case shows how that fails silently. For a week the split page discarded every written split, so the correction record had no rows. The split-review note says the panel did not render, not that nobody objected. Silence read as no objection. The replacement instrument needs its own liveness check, which a seeded probe would have supplied.\nstrength: strong\n\n### F8\ntarget: docs/research/conviction-and-critique.md:160\ntarget_words: \"Should the ontology itself live in the graph, contestable by its own mechanisms\"\nolder: docs/research/session28-the-harness-arc-and-the-source-intake.md:36\nolder_words: \"Let's hold off on the whole meta-rabbit hole of self-hosted fractal priors or beliefs about Deliberus stored inside Deliberus.\"\nkind: narrows\nwhy: Question 2 asks whether the design should be hosted in the graph, and calls it open. On 2026-08-30 the founder put the whole of it on hold. There are to be no self-hosted fractal priors or beliefs about Deliberus stored in Deliberus until the in-flight reviews are much closer to done. Ledger row Q28 keeps the settings-register idea but holds the claims-in-graph layer behind the same hold. The question is deferred by ruling, which changes what anyone should propose now.\nstrength: strong\n\n### F9\ntarget: docs/vision.md:201\ntarget_words: \"the prototype died on the one obstacle nobody could remove then\"\nolder: docs/research/incentives-analysis.md:11\nolder_words: \"But removing the labor cost does not make contribution individually rational. It removes one of three costs.\"\nkind: narrows\nwhy: The why-now section says the 2014 prototype died on one obstacle, typing the structure, and that its removal is half of why now. The incentives analysis says \"the diagnosis has changed and the corpus has not caught up\". Labour is one of three costs. Exposure, since every premise becomes a target, and social cost, since being seen engaging the other side reads as defection, are now the binding constraints. No tooling reaches them. \"The one obstacle\" is stronger than the record supports, and condition (b) of the success prior inherits the same overstatement.\nstrength: strong\n\n### F10\ntarget: docs/convergence.md:3\ntarget_words: \"When you decompose any human value system far enough, worldviews converge on shared bedrock.\"\nolder: docs/research/convictions-across-scales.md:58\nolder_words: \"Every strong instance of shared-bottom is either physically constrained or inherited. Values are neither in the same way.\"\nkind: narrows\nwhy: The page's headline states the shared-bottom claim without qualification. The cross-scale test splits the wager. Few, narrow and typed horizons recur almost universally, but a shared bottom does not recur. Every strong instance of one is forced by physics or inherited, and values are neither in the same way. The real support for shared bedrock is a phylogenetic argument the page does not make. The sharpened wager is shared structure with contested weights. A reader of the headline would overrate its second half.\nstrength: strong\n\n### F11\ntarget: docs/worldview-lenses.md:13\ntarget_words: \"That's why they failed — analytical tools alone don't motivate sustained use.\"\nolder: docs/research/kialo-deep-dive.md:203\nolder_words: \"Kialo targeted a **captive audience** (students) with a **distribution channel** (teachers/schools)\"\nkind: narrows\nwhy: The page states as fact that earlier platforms failed because they only decomposed. The Kialo deep dive gives a rival cause for the same graveyard: a tiny self-selecting audience with no channel. Kialo, equally analytical, survived by finding students and teachers. The steelmanned-critiques note also records that \"20+\" comes from a quick scan and includes Kialo (about 1M users) and DebateGraph. The lens-as-reason-for-failure line is a hypothesis, not a finding, and the same sentence recurs at line 108.\nstrength: strong\n\n### F12\ntarget: docs/research/fractal-priors-for-success.md:69\ntarget_words: \"absence of visible traction during the flat phase is not evidence against approaching the threshold\"\nolder: docs/research/conviction-and-critique.md:117\nolder_words: \"Friends-round readers do not point at claims (already registered as the single signal that would collapse the stack).\"\nkind: narrows\nwhy: Section 13 says no visible traction in the flat phase is not evidence against. Read alone, that cannot lose, and the corpus rejects such accounting for the convergence wager. The conviction note names in advance what would move the verdict toward no: readers not pointing at claims, inspectable synthesis steering as much as opaque, and claim reuse staying near zero after the approved gaps close. The flat-phase argument is therefore bounded by stated falsifiers, and the section should say so.\nstrength: medium\n\n### F13\ntarget: docs/depth.md:74\ntarget_words: \"It never shows a gap without a move** — an open question always comes with a next step\"\nolder: docs/research/scaffolding-versus-difficulty.md:37\nolder_words: \"between a move that tells you *where to look* and a move that tells you *what to conclude*\"\nkind: narrows\nwhy: Item 6 states unconditionally that every gap carries a next step. The scaffolding note records the opposing finding, that supplying the next step on every gap trained students to ask what is wanted (\"the supply created the demand\"). It says both rules cannot be unconditionally right. It proposes the missing distinction: a move that shows where to look versus one that shows what to conclude. Whoever builds item 6 needs that distinction, and the depth page leaves it out.\nstrength: medium\n\n### F14\ntarget: docs/competitive-landscape.md:146\ntarget_words: \"Assembly theory contributes one prediction nothing else in the corpus supplies:\"\nolder: docs/research/assembly-theory-and-the-reuse-mechanism.md:97\nolder_words: \"A cliché has enormous copy number precisely because nobody decomposes it.\"\nkind: supplies-a-missing-piece\nwhy: The reference-implementation posture (ledger M19) takes copying of the instruments as the sign of partial success. The source note warns that copy number is not a quality signal: the most-reused claim may be the least examined, and reuse alone would rank platitudes first. Copies of the instruments are therefore not evidence of success by themselves. Success would be copies crossed with the honesty discipline kept, which is the page's own failure-shaped-success worry. Anyone ruling on M19 should add that measure.\nstrength: medium\n\nNoticed: Two more pointers I did not claim as separate findings. (1) docs/research/graph-daemons-design-space.md:35 (\"A daemon that pre-digests does exactly that to Deliberus's own readers\") is narrowed by docs/research/earning-the-right-to-write.md:48 (\"Interpassivity is the machine doing the scrutiny so the human need…\"), which says an output that creates objection sites removes no scrutiny; ledger R10 test 3 holds the same test. (2) Note 9 was edited by someone else on 2026-10-10 (a correction at §0 item 4 and in §5, plus a new §5b) that withdraws the S3 rule; I did not use any of it, but S3 as written in the ledger may need re-reading in that light.\n"}