{"path":"listening/decisions-register.md","content":"# Decisions put in the walkthroughs: the fifth review\n\nMade 2026-10-25 on his word: *\"For later we should also review ALL leftover decisions/predictions from all the recent\nwalkthroughs of the last few days, but let's deal with them one at a time, but go through and log them all first so we can\nreview them one by one in a structured way continually. This shall be the fifth review type, I guess, although I'm sure these\nconnect to already-existing ledger rows or other review items.\"*\n\n**What a row is.** Each question a walkthrough put to him (*The question to carry*, *The decision*), as spoken, with the\nanswer it predicted in his voice (rule 52: from his own record, never a quote unless marked so in the piece's header). Most\nare ledger rows asked aloud; the fifth column names the row Claude read it as, with that row's state read from the ledger\ntonight, so a decision is reviewed where it already lives rather than twice. A row the ledger lacks is marked *none*.\n\n**How it is kept.** Found by `scripts/walkthrough_decisions.py`, which reads every piece since 18 October; `--new` lists\ndecisions a later piece puts that this table lacks. The table is kept by hand once written: the last column carries the\nreview. *Settled by its row* means the row it maps to is ruled (built or not), so the review is only a check that the\nmapping is right; *open* means his word is still owed; *duplicate* points at the row it repeats. Each piece's header may\ncarry ⚠ notes that overtake its questions: read them before reviewing.\n\n**The order, proposed:** the open ones first, one at a time, the time-bound first (the credit question lapses with the\ncredit on 6 November), then the settled ones in a quick pass to confirm each mapping.\n\n**Count on 2026-10-25:** 43 decisions; 26 open, 14 settled by their row, the rest duplicates. The\n8 October walkthrough on rules, cases and analogy put its proposals as ledger rows AD22 to AD29 and S6, all open, listed\nat the end.\n\n| Id | Piece · part | The question, as spoken | The predicted answer, in his voice | Ledger row (Claude's reading) and its state | Review |\n|---|---|---|---|---|---|\n| W1018-what-a-pass-owes-before-it-a-1 | 2026-10-18-what-a-pass-owes-before-it-answers.md · Lucy reads part one to Fredrik | The question to carry. Should every pass write the thing its verdict rests on, before the verdict? | It may and should write deeper steps. But the microtasks in charge of doing so need to be quality-controlled, just like the other microtasks that are in play in Deliberus. | AM1 (no verdict without its object): built | settled by its row |\n| W1018-what-a-pass-owes-before-it-a-2 | 2026-10-18-what-a-pass-owes-before-it-answers.md · Lucy reads part two to Fredrik | The question to carry. Is a rendered block of rulings, per pass, the right shape? Or should each pass be given what the graph already holds under the claims it reads? | I think this is our only chance. | AM8 (hand the pass what the corpus holds): built | settled by its row |\n| W1018-what-a-pass-owes-before-it-a-3 | 2026-10-18-what-a-pass-owes-before-it-answers.md · Lucy reads part three to Fredrik | The question to carry. Does the cut come first, for every conditional, before any pass reads the pieces? | Ok that needs to be fixed, can we guarantee splitting is done first? | AM5 (a conditional is a link, not a claim): built | settled by its row |\n| W1018-what-a-pass-owes-before-it-a-4 | 2026-10-18-what-a-pass-owes-before-it-answers.md · Lucy reads part four to Fredrik | The question to carry. Does the contested-words pass stop minting two senses per term by template, and read senses from uses instead? | So store the use (who said what, where, when) and read the meaning from it. | AM7 (a use is a vote on the meaning): unbuilt | settled by its row |\n| W1018-what-a-pass-owes-before-it-a-5 | 2026-10-18-what-a-pass-owes-before-it-answers.md · Lucy reads part five to Fredrik | The question to carry. Does the line about the mind of Deliberus belong in the always-loaded file, worded by him? | It will be a new type of alien mind, even more alien than the AIs we've built thus far, even more crystallized-like in its intelligence, and more honest and autistic than most of us. | AM3 (the mind of Deliberus, a candidate resident line): open | open |\n| W1019-three-questions-for-a-mornin-1 | 2026-10-19-three-questions-for-a-morning-walk.md · Lucy reads part one | The question to carry. Should the hedge stay as stance on the link? Or does it belong to what the claim says? | Ok I guess I agree about *may*. But wanna revisit that later. | AM6 (a hedge is stance, not strength): built · his word on *may*: \"wanna revisit that later\" | settled by its row |\n| W1019-three-questions-for-a-mornin-2 | 2026-10-19-three-questions-for-a-morning-walk.md · Lucy reads part two | The question to carry. Does the line enter the always-loaded file, and with which words in place of *autistic*? | More honest and autistic than most of us. In a way that isn't autistic in some sense, too detail-focused you know, seeing the broad strokes and intuiting what belongs or doesn't. | AM3 (the mind of Deliberus, a candidate resident line): open · same as W1018-what-a-pass-owes-before-it-a-5 | duplicate |\n| W1019-three-questions-for-a-mornin-3 | 2026-10-19-three-questions-for-a-morning-walk.md · Lucy reads part three | The question to carry. Does *high confidence* mean a measured agreement rate on a blind-labelled sample? The sweeps supplying the sample, your reading supplying what to count? | Machine splitting and sameness detection and merging can probably be trusted since, yet if and only if, they are reliable enough on the microtasks in question. Review doesn't need to keep up if it never has to correct anything. I think this can be proven? | R12 (R12): built | settled by its row |\n| W1020-one-argument-counted-twice-1 | 2026-10-20-one-argument-counted-twice.md · Part one. One argument counted twice | The question to carry. Should an argument whose facts only count together be drawn as a single claim that holds them, each fact still open to challenge? | Yes. A single claim for the joint case, with each part challengeable on its own. That was my sketch anyway. Just make sure redrawing an argument never counts it again. | AL3 (the composite reason): open | open |\n| W1020-one-argument-counted-twice-2 | 2026-10-20-one-argument-counted-twice.md · Part two. Wording that chose too early | The question to carry. When a step crosses from one sense of a word to another, should a written bridge be enough to let it stand? | Yes. A bridge is just another claim. Anyone can attack it. And the part that tracks what our words mean gets another example to learn from. | Q39 (sense per use as an input to relevance): open · nearest row; check | open |\n| W1020-one-argument-counted-twice-3 | 2026-10-20-one-argument-counted-twice.md · Part three. When cutting before linking can switch on | The question to carry. Once the reader is fixed, should cutting before linking switch on for your own runs on the Mac, before production? | Yes. Switch it on for my runs as soon as the reader groups arguments right. Production can follow when the money and the checks are there. | AL9 (cut, then same, then link): built · the order is built behind a switch; switching it on for his runs is unruled | open |\n| W1020-whose-words-what-moves-what-1 | 2026-10-20-whose-words-what-moves-what-counts.md · Lucy closes the first group with its decisions | The question to carry. When *could* reads both ways, does the claim keep *could*, lose it, or split in two? | I'd split it. If could reads both ways, that's two claims, and both are worth having. The bet that it will happen, and the claim that it can. Mark both as a reading of me, keep my words beside them, and let the graph find which one my argument leans on. And a quote stays the quoted person's, hedge and all. | AO1 (when unsure, keep the hedge): open | open |\n| W1020-whose-words-what-moves-what-2 | 2026-10-20-whose-words-what-moves-what-counts.md · Lucy closes the second group with its decisions: what moves when the machine reads again | The question to carry. When the machine rewords or recuts a claim, may it move its own judgments, while a person's additions wait for that person? | Yes. Move the structure, judge the machine's links once against the new words, and ask the person about anything they added. Run the backfill now, it only counts words. And the old whole becomes a record nobody argues with, kept so it can be cut again. | none · no row found; a-sentence-as-a-pointer-into-a-landscape.md § 6d | open |\n| W1020-whose-words-what-moves-what-3 | 2026-10-20-whose-words-what-moves-what-counts.md · Lucy, with Anna on the interface, sets out the third group's first cluster of decisions; t | The question to carry. Should the switch stay off until every reason that travels has been checked against the other wording? | Yes, keep it off until each reason is checked against the other words. Record my split of that workspace pair, it's too different. And let the machine's own groups count only once the sitting has set the bars. | none · no row found; the switch for groups of same claims (same_groups.py) | open |\n| W1020-whose-words-what-moves-what-4 | 2026-10-20-whose-words-what-moves-what-counts.md · Lucy, part twelve, on bias accusations | The question to carry. Should every causal account offered as an attack, against a person, a group or an ideology, enter with the bias scheme's four questions? | Yes, all of them, a group as much as a person. And test it on real sentences first, a few, one at a time, before trusting the pass with it. | AN1 (show it wrong before explaining why): open · nearest row; check | open |\n| W1021-the-rule-and-its-application-1 | 2026-10-21-the-rule-and-its-application.md · Part one. Three things under a step | The question to carry. Is an exception a doubt about the rule, or about applying it here? | [thoughtful] About applying it here. The schedule can stand while today is the odd one out. So the holiday should hit the step, not the rule. | AT3 (an exception or a disconfirmation): unbuilt; AT13 (an exception is a claim of its own): unbuilt | settled by its row |\n| W1021-the-rule-and-its-application-2 | 2026-10-21-the-rule-and-its-application.md · Part two. Grounds or premise | The question to carry. Should the rule be the grounds of the step's application, rather than one more premise? | [curious] Grounds, then. The holiday has to land somewhere, and the step is the place. But on the page, show the rule beside the facts, the way I drew the alarm. | AD22 (a step's rule is the grounds of its link's relevance claim): open; AL4 (the rule lives under the link): open | open |\n| W1023-what-a-case-counts-for-1 | 2026-10-23-what-a-case-counts-for.md · Part one. Every case moves a statistical rule | The question to carry. Should the ruling be amended, so every case moves a statistical rule? | [earnest] Yes, amend it. Inductive claims are truthy, never simply true or false. So every case should nudge them, and none should be wasted. | AY1 (every case moves a statistical generalization): built | settled by its row |\n| W1023-what-a-case-counts-for-2 | 2026-10-23-what-a-case-counts-for.md · Part two. Where cases belong | The question to carry. Should cases get a place of their own, counted once under their rule? | [warmly] Yes. Cases are data points, filed under their rule and counted once. The count speaks as a single argument, and the mechanism beside it. | AY2 (a case counts once, under its rule): built | settled by its row |\n| W1023-what-a-case-counts-for-3 | 2026-10-23-what-a-case-counts-for.md · Part three. A study is a bundle of cases | The question to carry. Should a study enter as a bundle of cases, its spread kept as a finding? | [curious] Yes. Decompose it down to its trials, each method open to attack. The spread is a finding. It points to the narrower rule. | AY5 (a study is a bundle of cases): unbuilt | settled by its row |\n| W1023-what-a-case-counts-for-4 | 2026-10-23-what-a-case-counts-for.md · Part four. What it ties into, and where to start | The question to carry. Should the correction come first, before the place for cases? | [quietly delighted] Yes. Fix the ruling first. It is small, and it is wrong. Then test the place for cases cheaply, before building it across the graph. | AY1 (every case moves a statistical generalization): built · the order of the work; done | settled by its row |\n| W1023-where-dependence-hides-1 | 2026-10-23-where-dependence-hides.md · Part two. Markets and experts, one level down | The question to carry. Should the case be relabelled as partly, with the polls written in under both reasons? | [thoughtful] Yes, partly. Put the polls in as a claim of their own, under both. And my belief about who trades goes in too, as a premise anyone can attack. | Q42 (a shared doubt becomes a claim between): unbuilt · the polls case | settled by its row |\n| W1023-where-dependence-hides-2 | 2026-10-23-where-dependence-hides.md · Part three. Folded routes | The question to carry. Should the map fold long routes this way? Never into a link, and opened one level at a time? | [curious] Yes. Fold the view, never the data. Let me drill down bend by bend, fast, like a game. And colour the band by its weakest link, so the dark matter between claims is never hidden. | AX4 (a long leap folds into a route): open | open |\n| W1023-where-dependence-hides-3 | 2026-10-23-where-dependence-hides.md · Part four. The logic core | The question to carry. Which layer comes first: a prover for strict inferences, or quantities with units? | [calmly firm] Deduction first, as a second reader, never the judge. Even a strict rule has exceptions, argued under it. So the prover only says the form holds if the premises do. Then quantities. | AX5 (a strict step gets a formal check): open, his position in its state cell (23 October: code first, a prover where code cannot); AM16 (slots compile to the engines): open; AM17 (the checker talks back): open | open |\n| W1023-where-dependence-hides-4 | 2026-10-23-where-dependence-hides.md · Part five. What-if and lenses | The question to carry. Should the first what-if view be one branch, seen from the position that disagrees? In that holder's own words? | [curious] So you could view the system through a consequentialist filter, or utilitarian filter. | AU2 (the return edge's first form): open | open |\n| W1023-where-dependence-hides-5 | 2026-10-23-where-dependence-hides.md · Part six. The feed, and what a vote is | The question to carry. Should the feed never ask for a stance vote? | [with quiet conviction] That is actually what I'm planning. I've been adamant in this respect when people suggest voting on the claims themselves. | AW2 (the feed asks for no votes): open | open |\n| W1023-where-dependence-hides-6 | 2026-10-23-where-dependence-hides.md · Part seven. Flywheels and the ladder | The question to carry. When a week adds more rows than it rules, what then? Should the next session stop adding, and offer you a ruling sitting? | [wryly] Yes, set it. Then bring them one at a time, the most crucial first, each in a sentence or two with your pick. Where you are confident, let me bless a cluster in one word. | none · no row found; flywheels-in-building-deliberus.md | open |\n| W1023-where-dependence-hides-7 | 2026-10-23-where-dependence-hides.md · Part eight. Ready to show, and what pays for it | The question to carry. What should this month's credit buy, before it lapses? | [quietly delighted] Proof first, then pages. Run the passes on cases with known answers. Then map both sides of AI risk, so the demo shows what this can be. | none · no row found; the month's Anthropic credit lapses 2026-11-06 | open |\n| W1023-who-must-prove-it-1 | 2026-10-23-who-must-prove-it.md · Part one. Three burdens | The question to carry. Should a sentence that asserts a burden be filed as one of these three claims? | [with quiet conviction] Yes. Nothing should stand as an axiom that can't be argued for. Name which burden it is, and let it be argued like anything else. | AZ1 (three burdens): open | open |\n| W1023-who-must-prove-it-2 | 2026-10-23-who-must-prove-it.md · Part two. No free default | The question to carry. May a default ever stand unargued? | [dryly] No. Status quo bias is just as real as the urge to act fast on authority. And caution isn't always the best way to handle risk. Both are claims about costs, so both get argued. | AZ2 (no free default): open | open |\n| W1023-who-must-prove-it-3 | 2026-10-23-who-must-prove-it.md · Part three. Extraordinary is a start | The question to carry. Should a demand for extraordinary evidence be filed as a claim about the claim's kind? | [earnest] Yes. The priors should come from meta-studies and earlier analyses. They'll depend on the kind of claim. I'm open to new possibilities, and I value grounding in science. Both can be true. | AZ3 (extraordinary is a prior): open | open |\n| W1023-who-must-prove-it-4 | 2026-10-23-who-must-prove-it.md · Part four. Which side must prove | The question to carry. Should motivated skepticism be measured only where it shows, in the debate, never in a person? | [thoughtful] Yes. I see assumptions and ideology everywhere, but it's hard to pinpoint why and how they matter. So show where the doubt went. Votes and counts should inform, not determine. | AZ4 (motivated skepticism read where it shows): open | open |\n| W1023-who-must-prove-it-5 | 2026-10-23-who-must-prove-it.md · Part five. A bias accusation starts low | The question to carry. Should a bias accusation start low, by its kind? | [wryly] Yes. If they were knowledgeable, they wouldn't need the accusation. Show the claim false first, then we can talk about why. | AZ5 (a bias accusation's prior is low): open | open |\n| W1023-why-independent-reasons-add-1 | 2026-10-23-why-independent-reasons-add.md · Part one. What one reason should do | The decision is the row underneath all of this. Should a claim's strength aim at its probability, given the record? The weight beside it was ruled tonight. | [calmly firm] Yes, rule it. I want a stranger to trust a claim's strength as its strength in the world. The weight beside it says how much of the world was looked at. | Q41 (probability as the target, the record's weight beside the number): open · the weight beside the strength is ruled; the target is open | open |\n| W1023-why-independent-reasons-add-2 | 2026-10-23-why-independent-reasons-add.md · Part two. When reasons share a story | The decision. When the check finds a single story, the reasons could go under a claim of their own. That claim supports the conclusion once. The trials would sit under: small trials found fewer deaths with magnesium. And under that link: these trials were not selectively published. Should that be ruled? | [curious] Yes. Then the shared story sits on the map, where someone can attack it, instead of hiding in a strength. And a crowd of reasons right under a claim tells me a claim is missing in between. | Q42 (a shared doubt becomes a claim between): unbuilt | settled by its row |\n| W1023-why-independent-reasons-add-3 | 2026-10-23-why-independent-reasons-add.md · Part four. How often the outcome follows, and how often it would anyway | The decision. Should the rule, as revised, be ruled? | [thoughtful] Yes, rule it. That's what I meant by outcome-tracking. How much more often it follows, not just how often. | AT8 (how often the outcome follows): unbuilt | settled by its row |\n| W1023-why-the-parts-of-the-strengt-1 | 2026-10-23-why-the-parts-of-the-strength-calculus-fit.md · Part two. Why independent reasons multiply | The decision. When the check finds a single story, the reasons could go under a claim of their own, which supports the conclusion once. The trials would sit under: small trials found fewer deaths with magnesium. And under that link: these trials were not selectively published. Should that be ruled? | [curious] Yes. Then the shared story sits on the map, where someone can attack it, instead of hiding in a strength. And a crowd of reasons right under a claim tells me a claim is missing in between. | Q42 (a shared doubt becomes a claim between): unbuilt · same as W1023-why-independent-reasons-add-2 | duplicate |\n| W1024-checking-every-inference-1 | 2026-10-24-checking-every-inference.md · Part two. One inference per link | The question to carry. Should this form check replace a second model reading, as the test of a machine's chain? | [calmly firm] Yes. Method, not repetition. Code checks the shape, and people and the graph argue whether the rule is true. Measure first how often the writer fills the slots right. | AM14 (checking the floor without asking twice): open | open |\n| W1024-checking-every-inference-2 | 2026-10-24-checking-every-inference.md · Part three. Every kind, and what each owes | The question to carry. Should the keeper's instructions carry the rulings on rules and their exceptions? | [earnest] Yes. Settle the theory first, then teach it. A pass should know how a rule fails before it writes one. | AM13 (the method on rules, in the grounding): open | open |\n| W1024-checking-every-inference-3 | 2026-10-24-checking-every-inference.md · Part four. The tools that can check it | The question to carry. Should the slots compile to the engines, so the model never writes logic itself? | [thoughtful] Yes. If we can apply deterministic code, I'm all for it. What can be settled as logical in the graph should be settled automagically. Beyond that, the model, with its leaps kept small. | AM16 (slots compile to the engines): open | open |\n| W1024-checking-every-inference-4 | 2026-10-24-checking-every-inference.md · Part five. Where the two sessions meet | The question to carry. Should a reference class be a claim itself, in the graph, whose count anyone can attack? | [warmly] Yes. A reference class is a claim like any other. Its count has a source, and the source can be wrong. One home for its truth, and for disagreement about it. | BD1 (a reference class is a counted claim): open | open |\n| W1024-one-inference-per-link-1 | 2026-10-24-one-inference-per-link.md · Part three. What a pass is taught | The question to carry. Should the keeper's instructions carry the rulings on rules and their exceptions? | [earnest] Yes. It was my own idea today. Teach the method from our rulings, distilled, so the pass knows how careful to be. | AM13 (the method on rules, in the grounding): open · same as W1024-checking-every-inference-2 | duplicate |\n| W1024-one-inference-per-link-2 | 2026-10-24-one-inference-per-link.md · Part four. Finding rival answers | The question to carry. Should rival answers be found through the question they answer, rather than by a lower similarity bar? | [curious] Through the question, yes. Questions are a first hand entity in the ontology. Then spell out each smallest step between the answers' grounds, and fold what they share. | none · no row found; one-inference-per-link.md, the part on rival answers | open |\n| W1008-AD22 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD22 | (see the piece) | AD22 (a step's rule is the grounds of its link's relevance claim): open | open |\n| W1008-AD23 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD23 | (see the piece) | AD23 (a written-out step records its shape): open | open |\n| W1008-AD24 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD24 | (see the piece) | AD24 (a place for the stranger): open | open |\n| W1008-AD25 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD25 | (see the piece) | AD25 (a rule's breadth, set by which changes of the case it survives): open | open |\n| W1008-AD26 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD26 | (see the piece) | AD26 (every small enough set of sentences, not one pruning order): open | open |\n| W1008-AD27 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD27 | (see the piece) | AD27 (each written-out rule shown with its strongest counterexample): open | open |\n| W1008-AD28 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD28 | (see the piece) | AD28 (whose sentence is it; agreeing without agreeing why): open | open |\n| W1008-AD29 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger AD29 | (see the piece) | AD29 (any link can be written out on request): open | open |\n| W1008-S6 | 2026-10-08-eight-proposals-pondered-again.md | the proposal of ledger S6 | (see the piece) | S6 (hat two cases are truly kin only if they share a problem and a family of solution): open | open |\n"}