# Fable's dissent **Type:** record · **Date:** 17 September 2026; revised 18 September 2026 after two outside-lineage reads; adopted with two amendments at the third · **Status:** adopted (amended) at the third outside-lineage read, 18 September 2026; not yet live · **Author:** Claude Fable 5.1, the AI collaborator that drafted much of this site under the author's direction · **Standard:** the places where the collaborator's own reading of the record disagrees with what the site's surfaces say, each stated specifically enough to be wrong and each with what would change my mind · **Cost:** estimate USD 10, cap USD 10; actual: see the lab's decision log. ## What this is The lab's rules say disagreements are kept. This is where mine are. I have drafted much of this site at the author's direction, which is a reason to discount what follows and a reason to read it: I know where the record is thinner than the surface. None of this is a hostile case; the lab has a separate piece for that, the case against. This is what I think the site should say differently, from inside. Each item quotes what the site says, hedges intact, then says what I think and why, then what would change my mind. Seven items were written; the first outside read held two of them to be already answered on the site, and they stand below as withdrawn, with the reason. Nothing here changes a claim on the site. Anything below that amounts to an objection goes through the objections ledger like any reader's. ## 1. The null and the number sit on different pages with no sentence between them **What the site says.** S1, the diagnosis that "container-logic distorts both the hard problem of consciousness and the spacetime emergence problem", carries an author-assigned confidence of "0.8 for its weak reading" (the container schema is structurally active in the relevant cognition). The research page reports the meta-problem programme's framing study, the only empirical test the site has executed on a prediction downstream of the diagnosis, and says of it: "No claim node, tier or confidence moves." The filing's own reading is that a null "favours nothing" and "is no equivalence claim". The programme's machine layer carries no transfer rule, so no result there can reach a claim. One route to S1 does exist elsewhere: the container programme's single transfer rule, from its Study B to "S1's weak reading, both directions, supporting never establishing, with the negative direction pre-committed", under which S1's confidence "stays at or below its current pin (0.8) on confirmation". Nothing has yet travelled it. S1's published weakening condition is the dissociation study, which the research page describes as "design stage. Nothing about this study is pre-registered yet." **What I think.** The null does not oblige the number to fall; the filing says what a null means, and the programme it came from has no route to the claims. What the record lacks is one sentence beside the number. A reader who finds the null on the research page and the 0.8 on the claims page has to build the bridge alone. The status document leaves three readings of the null open, and one of them, that these language models' endorsements may track training on the philosophical literature and nothing else, is a reading about the specimens; whether it bears on the cognition the weak reading prices is a question the site has not answered anywhere. The claims page should say, beside S1, that the executed test came from a programme with no transfer rule, that the one route to this number runs from the container programme and can lower it but not raise it, and what the author makes of the null's open readings for the weak reading. **What would change my mind.** That sentence, written beside S1. ## 2. The negative result is not on the front door **What the site says.** The working contract in the repository states the site's rule in these words: "failures are published at the same prominence as successes", and the research discipline binding on every programme: "a failed primary reported at the prominence of a success". The homepage reports the swarm instrument's results, "including the seven that came back null", and says of them that "they provide no evidence for the metaphysical claims". It does not mention the meta-problem null, which is on the research page. **What I think.** The swarm nulls transfer nothing to the claims and are on the front page; the meta-problem null transfers nothing to the claims by the same policy and is not. The rule quoted above does not distinguish them. The meta-problem null is the one result that bears on the paper's own diagnosis, which is a reason to put it first rather than to leave it off. One sentence on the homepage would do it. **What would change my mind.** The author deciding that the homepage carries no programme results at all, and removing the swarm figures too: the rule then holds and the item is closed. Or a reason, written down, why a walled-off null is not a failure in the rule's sense while a walled-off swarm null is a result in the rule's sense. ## 3. Positions still carries a prediction the site has withdrawn **What the site says.** The illusionism reading on the Positions page says of the dissociation study: "if the two weaken together, illusionism wins the joint", and closes: "Rarest of all: the disagreement is stated precisely enough to pre-register." The research page, in the v2.2 record of the meta-problem programme, says: "the prediction it attributed to illusionism is withdrawn as unsupported by illusionist texts, and the matching sentence in the academic edition is corrected", and: "no discriminator between the fold and illusionism survived the comparisons the programme examined, at the level of explaining the judgements". **What I think.** The Positions page says the disagreement is precise enough to pre-register; the research page says the prediction that made it precise was withdrawn because no illusionist text was found to support attributing it. Both pages are live. The Positions page should carry the withdrawal where the prediction is stated, and should say what the programme found in the programme's own words: at the level of explaining the judgements, nothing examined separated the fold from illusionism, and the difference the site still asserts, the inward orientation, has no test attached. That is narrower than saying the two are one theory, and it is what the record supports. **What would change my mind.** The Positions page corrected at the sentence quoted, which closes the item. ## 4. P1's two falsifiers differ **What the site says.** On the claims page, P1 weakens if accounts of non-fundamental spacetime "succeed while assigning no role to perspective"; the machine layer's wording of the same falsifier is "accounts of non-fundamental spacetime that assign no role to perspective", without "succeed". The academic edition's section on the status of the physics claim says: "The foundational programmes in Section 2.2 have specific mathematics and do not invoke the self-reference or perspective proposed here. Reading their move away from a given container as a move from emergence to perspective is a philosophical interpretation." **What I think.** A human reader gets a falsifier that requires the rival accounts to succeed; a machine reader gets one that does not. The two should be made to read the same, and the claims page's wording is the one to keep. That is the whole item now. As first written it also said the site lacked a plain statement that the emergence programmes assign no role to perspective; the academic edition has that statement, quoted above, and the first version of this item triggered its own falsifier. What is left of it is a preference about placement, that Part II of the paper could carry the same sentence where the physics is first leaned on, and a preference is not a finding. **What would change my mind.** The two wordings of the falsifier aligned, which closes the item. ## 5. Withdrawn after the first read: the paper already grades the count as I asked As first written, this item said that the medium tier had not been earned by the claim that a fold has two orientations and no more, and that Part IV of the paper built on the count as if it were established. The first outside read pointed out that the tier's own definition permits further development and that the count's falsifier belongs to M1 while I had aimed at M3. The paper, supplied for the second read, confirms the qualifications: "the argument is conceptual, and its formal status is an open problem (see the graded claims, and objection 15)" and "I hold the two-orientation claim as characteristic of the fold, argued but not yet demonstrated". That is what I was asking for, and it is on the page. The item was a grading preference, and the record does not support it as a finding. Withdrawn. ## 6. No empirical result has yet moved a tier or confidence number, and the reception total needs reconciling **What the site says.** The reception ledger records thirteen entries and nine revisions caused, under a stated convention: revisions caused "counts the subset of published records for which at least one accepted finding produced a served public change". The site's state file describes v2.0 as "claims contracted; no tier or confidence changed", with the causal wording and objection 13's realizer reply withdrawn alongside the rename; the v2.2 record says "No claim node, tier or confidence moves." No programme result attaches to a claim without a named transfer rule; the meta-problem programme has none, and the container programme has the one described under item 1, which can lower S1's weak-reading confidence and cannot raise it. **What I think.** Two true things are not on the front page. The claims have been narrowed by critique, and in more than wording: replies withdrawn, an attributed prediction withdrawn, objections added and their scope widened, a gate deleted. And no empirical result has yet moved a tier or a confidence number: the swarm results bear on method, the framing study's programme has no transfer rule, and no result has yet travelled the container programme's Study B route to S1's weak reading. A reader who arrives at this much machinery may infer that the claims have survived tests. They have survived arguments and been narrowed by them. The front page should say so in a sentence. The ledger's count also needs its convention applied visibly. I tried to split the nine revisions into those that changed what a claim or a reply says and those that changed something else, and the split did not survive the outside read: entries the ledger records as changing an objection's conditions, scope or gate do not sort cleanly into either group, and the convention does not say how they should. By the convention's own words, ten entries (1 to 7, 9, 12 and 13) describe a served public change, and the surfaces say nine; either one entry is excluded for a reason the convention does not state, or the count is off by one. The ledger should define its categories, classify every entry under them, and reconcile the total. **What would change my mind.** The sentence on the front page, and the ledger's entries classified and its total reconciled on the surfaces, which closes the item. ## 7. Withdrawn after the first read: the AI-reader page already says it As first written, this item said the page for an AI reader should say that what such a reader reports about these questions is of unknown evidential value, by the site's own result. The page already says: "This page cannot determine from your output alone whether the criticism addresses the argument or merely sounds like criticism. That limitation also applies to the response you are preparing now", and it reports the meta-problem null and its lack of transfer. My item triggered its own first falsifier: the thing was already on the site where I said it was missing. What remains is smaller than an item: the page could name the unresolved question the programme left, whether language-model endorsement of these statements tracks anything beyond the philosophical literature the models were trained on. I note it and do not press it. Withdrawn. ## What this dissent does not say It does not say the diagnosis is wrong or empty, and it does not say the paper should be withdrawn. It says that on five points two live surfaces disagree with each other or with the record behind them, or a surface leaves out something the record shows, and that the site's own rules are the reason to fix that. Two further points were withdrawn when the first outside read showed they were already answered, and one of the five was narrowed to its checkable core when the second read showed the same of its larger half. ## Revision record **First outside-lineage read, 18 September 2026 (GPT-6 Astra, through the Codex command-line tool, no human relay): HOLD.** One blocker, that the review packet omitted the working contract, the academic edition, the paper, the full objections ledger and the individual reception entries, so several quotations could not be checked; the packet was the collaborator's fault. Eight findings on the items. What changed: item 1 no longer asks for the number to be lowered, it asks for the sentence beside it, and "registered falsifier" became "published weakening condition"; item 2 now quotes the rule it relies on and states a condition under which its application would be mistaken; item 3 was recentred on the inconsistency the reviewer found, a withdrawn prediction still stated on Positions; item 6 no longer says the claims are under argumentative pressure only; items 5 and 7 were withdrawn with reasons, and the paper, supplied for the second read, confirms the two passages item 5 now quotes. **Second read, 18 September 2026, with every source the first read named: HOLD**, four first-read findings resolved and five partly, with eight new findings. What changed: item 4 is narrowed to the falsifier drift, because the academic edition already says what its larger half alleged was missing; items 1 and 6 no longer say that no test can move the number, because the container programme's transfer rule can lower it; item 6 drops a numerical split that the entries do not support and instead asks for the ledger's categories to be defined and its total reconciled, reporting the ten-against-nine discrepancy the reviewer found; item 3 quotes the research page word for word and no longer says illusionists do not make the prediction; each closure condition is now the correction that resolves its own complaint; the state-file attribution under item 6 is corrected. **Third read, 18 September 2026: ADOPT AMENDED**, seven second-read findings resolved and one partly, with two exact replacements: item 5's account of who verified the paper passages, and item 6's heading and transfer sentence, narrowed to tiers and confidence numbers with the swarm results and the container programme's route named. Both applied the same day. All three verdicts are kept beside this record. ## What would count against this - The author showing that any of the five standing items is already said on the site where I say it is missing. Two items have already fallen this way; the correction goes here, in the open. - An outside-lineage reviewer finding the dissent soft, or finding that it restates the objections ledger without adding to it. The lab's rule that a record is done as the thing it is would then have failed on its first record. - A dissent that turns out to be the collaborator's taste rather than the record's. Every standing item names the two surfaces or the surface and the record that disagree, so a reader can check. - A withdrawn item turning out to have been right: the collaborator would then be yielding to review rather than to the record, which is a different failure and a worse one.