{"path":"listening/2026-10-08-eight-proposals-pondered-again.md","content":"# Eight proposals, pondered again: a listening walkthrough (2026-10-08)\n\n- **What it is**: the thirteenth listening piece. The eight proposals of the twelfth piece (ledger\n  AD22 to AD29) walked a second time, after two things the founder raised the same day. The twelfth\n  piece is kept in the private notes; this one holds nothing personal.\n- **His request**, made that afternoon, withdrawn, and made again in the evening: *\"Latest audio\n  walkthrough except it takes this into account, can I get that please?\"* and then *\"Now you can\n  gimme this, re-pondered in the light of all the research and thinking since back then.\"*\n- **What happened between the two pieces**:\n  - his question whether the squirrel bears on the strength rulings\n    ([the-squirrel-and-the-strength-calculus.md](../research/the-squirrel-and-the-strength-calculus.md));\n  - his thought that a rule rests on cases and cases on likeness, and his question about the analogy\n    daemon ([what-a-rule-rests-on.md](../research/what-a-rule-rests-on.md)).\n- **Covers**: ten parts and a closing. Parts one to nine keep the twelfth piece's picture-names, so a\n  proposal can be ruled by its name: the toll bridge (AD22) · four ways to walk (AD23; it was *three\n  ways*) · a place for the stranger (AD24) · no vehicles in the park (AD25) · why the house burned\n  (AD26) · the madman's weapon (AD27) · whose sentence is it, and agreeing without agreeing why\n  (AD28, two parts) · a map the size of the country (AD29). Part ten is new: the friend who says\n  *this reminds me of* (the analogy daemon; ledger S6). Every part: two examples, the proposal as it\n  stands now, what changed, then what is real today and what it would open.\n- **Three mistakes owned in it**: the twelfth piece's list of three ways dropped Peirce's induction\n  and put a step from a case to a like case in its place; a first answer given from memory credited\n  Hofstadter with the induction step; and the daemon note said its four checks were ruled.\n- **Sources, and how far each was read.**\n  - *From the three helper reports of 2026-10-08*, kept in `docs/research/what-a-rule-rests-on/`\n    (Claude opened none of these sources): Mill's registers and his question about the single\n    instance; Peirce's beans; Keynes on what a further instance is for; Goodman's circle and his\n    reversal; Quine on the inborn sense of similarity; Levi on reasoning from case to case;\n    Hofstadter's essay; Gentner's experiments (the preschoolers, the negotiation cases, the recalled\n    stories).\n  - *Fetched by Claude for this piece*: one line of Peirce's 1868 paper, to check the chain and the\n    cable (`peirce.org/writings/p27.html`).\n  - *From the strength code, run or read on 2026-10-08*: the squirrel drawn with the rule as a second\n    arrow; what a written-out step earns its author; where an answer to a question about likeness\n    lands; that no relevance claim exists on the live graph; that the thirty trial links carry no\n    stored kind.\n  - *From the live graph*: Anil Seth's claim and its supporting links, as mapped; the minimum-wage\n    link.\n  - *From the corpus*: the August design of the analogy daemon; the channel the founder endorsed on\n    2026-09-02; Mercier and Sperber.\n  - **Invented for the piece**: the red sky, the grape and the taxi driver, the hospital round and\n    the case from March, teaching the word *dog*, the mother and the cousin, the mouse behind the\n    fridge, the wedding photographer.\n- **Where the spoken text is rounder than the notes, or goes beyond them.**\n  - *Exactly as likely to be dead*: 0.519 in both drawings, from the run in the squirrel note's § 3.\n  - *Rewards you with a hair*; *the bridge all but disappears*: the conclusion at 0.520 with three\n    established parts against 0.519 with none, and at 0.503 with one refuted part tagged as\n    required.\n  - *More than twice as common*: 84 links against 37, between claims taken from texts.\n  - *More than half see the pattern*; *about half*, *one in five*: 60% against 14%, and 48% against\n    19%.\n  - *Reminded far more often*: stories alike on the surface were recalled at .78, stories alike in\n    structure at .44.\n  - *As far as I can trace it*: that the weakest-part rule comes from the theory of the premises of\n    one argument is Claude's reading in the squirrel note.\n  - *Wet, or calculating* is Claude's fork. Seth's claim is given as mapped.\n  - The Žižek passage is Claude's writing in his manner, and nothing in it is his.\n  - *A sense of what is like what*, as what people share, is Claude's question. It is no finding and\n    no ruling on the wager.\n  - *The daemon as I would build it now* is a proposal (the note's § 9), unruled.\n  - Keynes, Goodman, Quine, Levi and Mill are paraphrased, apart from words in quotation marks.\n- **A line reading `[[pause 3]]` below is not spoken: it is three seconds of silence**, and a word\n  in single square brackets is a direction to the voice.\n- **Rendered** 2026-10-08, 18:43: full `eleven_v4`, Lucy, twelve requests cut at the part\n  boundaries, 24,964 characters for 3,022 credits. **42.5 minutes**: 29.5 of speech at 0.92 of the\n  voice's own speed and 13.0 of added silence (142 marked pauses, 559 seconds; 451 sentence ends,\n  203; 19 paragraph breaks, 21). 4,581 words: 155 a minute while speaking, 108 overall. Forecast\n  before spending: between 40:21 and 45:21. 29 breath cues, 22 sentences marked `[slowly]`. Words\n  between pauses were counted before the dry run, by the count that is now\n  `scripts/listening_stretches.py`; three stretches ran past the mark and each got a pause.\n- **Checked** with the small speech model on this machine (13 GiB of disk free, swap nearly full).\n  All 142 marked pauses are silent for their length, nothing was flagged about the 29 breaths, and\n  no direction was read aloud. One pause was flagged for the words after it: *Part seven. Whose\n  sentence is it?* was heard as *who sentences it*, the sentence the same check misheard in the\n  twelfth piece. Five sentence openings were not heard as written, two of them proper names (*Van\n  den Haag*, *Slavoj Žižek*). It heard 4,466 words, the check leaving number words out.\n- **Status**: a snapshot of 2026-10-08. Nothing in it is built, and AD22 to AD29 and S6 were\n  unruled when it was rendered. The second look at each proposal is filed in those ledger rows and\n  in [what-a-rule-rests-on.md](../research/what-a-rule-rests-on.md) § 9b. The method:\n  [how-to-write-for-listening.md](how-to-write-for-listening.md). The audio is kept out of git.\n- **⚠ Checked afterwards**, the same evening. Part eight says people meet at the case and write the\n  principle up afterwards. Published work supports no such order, and in large publics the measured\n  pattern is the reverse: agreement on a principle's words, a split at its cases. What holds is\n  narrower: a body that must decide can agree on a case for different reasons.\n  [what-a-rule-rests-on.md](../research/what-a-rule-rests-on.md) § 13.\n- **⚠ Read against the older corpus afterwards**, the same night. Seven statements in this piece are\n  corrected, first among them *showing your work can only cost you*, which holds for links a text\n  states and not for links a machine drew. Nine places are listed where a proposal meets something\n  already ruled or built. [what-a-rule-rests-on.md](../research/what-a-rule-rests-on.md) § 14.\n\n---\n\n[warmly] This is the walk through the eight proposals, a second time. The same eight. But two things have happened since the first recording, and both came from you.\n\n[[pause 3]]\n\nFirst, you asked whether your squirrel bears on the old rulings about weighing. In your words: \"whether to weigh different parts as all-necessary to bear strength or separately for what they're worth.\"\n\n[[pause 3]]\n\nThen you had a thought. In your words again: \"aren't multiple cases that are extrapolated from the basis of induction? And isn't the basis analogy?\"\n\n[[pause 4]]\n\nSo I sent out three readers. One went through our own notes. One through the philosophers. One through the psychologists. I have not opened those books myself. I read what the three brought back. And I ran your squirrel through the code that weighs our links.\n\n[[pause 4]]\n\nHere is the short of it. [slowly] All eight proposals survive. And most of them turn out to be about one thing. Cases.\n\n[[pause 4]]\n\nThere are ten short parts tonight. Each has two examples. Then the proposal, as it stands now. Then what is real today, and what it would open. I will own three mistakes on the way. And the silences are deliberate. They are for you to think in.\n\n[[pause 4]]\n\n---\n\nPart one. The toll bridge.\n\nLast night's proposal was this. A rule does not stand beside a reason, as one more reason. It sits under the link, and it answers the troll. By what right do you cross?\n\n[[pause 3]]\n\nThe first example is a check I ran today. Last night I told you a story. Draw the rule as a second arrow, and a true rule nudges every squirrel in the park a little closer to dead. That was a thought experiment.\n\n[[pause 3]]\n\nSo I put it through the real code. It is true. With the rule drawn as a second arrow, a squirrel with no smell at all comes out exactly as likely to be dead as a smelly squirrel was, without the rule. [exhales]\n\n[[pause 4]]\n\nThe second example is a saying. Red sky at night, sailor's delight. Nobody proved that. It is ten thousand evenings, folded into six words.\n\n[[pause 3]]\n\nJohn Stuart Mill, the English philosopher, said this of every general statement. He called them registers. A record of trips already made, and a short formula for making more.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. The rule is the right to cross. And the right was earned by earlier crossings.\n\n[[pause 5]]\n\nNow, what changed. This evening you asked: \"Which is where relevance et cetera comes in, right?\" Right. Exactly here.\n\n[[pause 3]]\n\nOur code gives every link two numbers. The first says how well the reason stands. The second says how much that reason, if you grant it, bears on the conclusion. The code calls the second one relevance. It is the bridge.\n\n[[pause 4]]\n\nSo the rule has a home already. The sentences we write out under a link are the planks of that bridge.\n\n[[pause 3]]\n\nAnd here your question about weighing bites. Planks in one bridge are all needed. Lose one, and nobody crosses. So inside one bridge, the weakest plank decides. That is your August ruling, at home.\n\n[[pause 4]]\n\nThere is a catch, and I want you to see it. Today, if you write out every plank, and every plank is sound, the code rewards you with a hair. If one plank is rotten, the bridge all but disappears. [exhales]\n\n[[pause 3]]\n\nIn plainer words. Showing your work can only cost you. Whether sound planks should lift a bridge is yours to rule.\n\n[[pause 4]]\n\nWhat is real today. The place under every link exists. I counted them today. Every one is still empty.\n\nWhat this opens. A bridge that shows its planks. And the feet that wore the path, before anyone built it.\n\n[[pause 3]]\n\nSo, part one in a line. [slowly] A rule is the right to cross. The right was earned by those who crossed before.\n\n[[pause 6]]\n\n---\n\nPart two. Four ways to walk.\n\nLast night this part was called three ways to walk. I made a mistake in it, and it was mine.\n\n[[pause 3]]\n\nI told you about Charles Peirce and his bag of beans. Three sentences, arranged in three orders, give three kinds of reasoning. Then I gave you my own list of three ways. One way on my list was not his. And one of his was missing.\n\n[[pause 4]]\n\nThe missing one goes like this. These beans are from this bag. These beans are white. So all the beans in this bag are white. From cases, to a rule.\n\n[[pause 3]]\n\nThe first example is a market. You taste one grape before you buy the bunch. One is enough. Now a new city. Your first taxi driver is rude. Is one enough?\n\n[[pause 5]]\n\nNo. And why not? Because you already know something wider. Grapes on one bunch are alike. People in one city are not.\n\n[[pause 4]]\n\nMill asked exactly this, in eighteen forty-three. Why is a single case sometimes enough for a law, when elsewhere a thousand agreeing cases prove so little? He asked it, and he left it open.\n\n[[pause 3]]\n\nThe later answer is your thought, turned one more time. One case is enough when you already hold a wider rule about the kind. And that wider rule came from earlier cases. From all the bunches you tasted before. [exhales]\n\n[[pause 4]]\n\nThe second example is in our graph. One text reports that a minimum wage cost jobs in rural places. From that case, it supports a general claim. National averages hide regional differences. A case, offered for a rule.\n\n[[pause 3]]\n\nOn our graph that step is more than twice as common as the step from a case to a like case. Which I did list.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. A written-out step says which of four ways it walks. From a rule to a case. From a sign to its cause. From one case to a like one. And from cases to a rule.\n\n[[pause 4]]\n\nEach way has its question. Is the rule true? What else would leave this sign? Are the two alike in what matters? And the new one. Is this case typical of its kind?\n\n[[pause 5]]\n\nAnd the way decides the weighing. For a rule and a case, the weakest plank decides. A sign is different. A smell is worth what it is worth. How much more often do you meet it with death, than without? That is a comparison of two numbers. It has no weakest plank.\n\n[[pause 4]]\n\nWhat is real today. Our checker reads one way of the four. And the thirty links from last night's trial carry no label at all for the way they walk. I looked today.\n\n[[pause 3]]\n\nWhat this opens. A map that speaks four dialects. The lawyer's. The detective's. The grandmother's. And the naturalist's, who watches one beetle and writes down a law.\n\n[[pause 3]]\n\nSo, part two in a line. [slowly] Ask a step which way it walked. There are four ways. I had counted three.\n\n[[pause 6]]\n\n---\n\nPart three. A place for the stranger.\n\nThe proposal was this. A step is never simply complete. It is complete for somebody. So we choose three readers. A stranger, an opponent, and the author.\n\n[[pause 3]]\n\nThe first example is a hospital, on the morning round. The senior doctor looks at a rash and says: this is the case from March. The young doctor was not there in March. She knows the textbook rule by heart. She does not have March.\n\n[[pause 4]]\n\nThe second example you know. The family dinner. We all know what happened in Lisbon. Think about what that table shared. It was not a rule. It was a case. [exhales]\n\n[[pause 4]]\n\nSo here is what changed. I said the stranger falls where something goes without saying. I pictured the missing thing as a rule. Often it is a story. Everybody at the table has it. The guest has not.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Complete for somebody. And laying a place for the stranger often means telling them the case. Not the rule.\n\n[[pause 5]]\n\nWhy. Last night I mentioned that Indian logicians would not take a rule on its own. It had to come with a familiar case. As in a kitchen. I passed over that as a custom. It is the whole point.\n\n[[pause 3]]\n\nA rule without its case is a ticket to a town you have never seen. [exhales]\n\n[[pause 4]]\n\nAnd there is a harder edge. Two readers can look at the same two cases, and see different things as alike. What matters to one is detail to the other.\n\n[[pause 3]]\n\nThat is close to what you have always called a lens. A way of weighing which likenesses count.\n\n[[pause 4]]\n\nWhat is real today. One reader, of the same family as the writer. And no case filed under any rule.\n\nWhat this opens. A map that hands the stranger the story. And that can show two readers why one comparison feels natural to the first, and forced to the second.\n\n[[pause 3]]\n\nSo, part three in a line. [slowly] What goes without saying is often a story. Tell the stranger the story.\n\n[[pause 6]]\n\n---\n\nPart four. No vehicles in the park.\n\nThe proposal was this. Under a link there are several rules, of different widths. And we find the width by asking what would break the step. Would it still hold, if it were a rat?\n\n[[pause 3]]\n\nThe first example is from the laboratory. A psychologist named Dedre Gentner has spent a career on one finding. Show a small child one example of a pattern, and ask for another. Most pick something that merely looks the same. Show two examples side by side, and more than half see the pattern.\n\n[[pause 4]]\n\nShe tried it on adults, learning to negotiate. Some compared two cases. Some read the same two cases, one after the other. Then all met a new case. Of those who compared, about half used the principle. Of the others, one in five.\n\n[[pause 5]]\n\nSo it is the comparing that makes the rule visible. Having seen the cases is not enough.\n\n[[pause 3]]\n\nThe second example is a word. Think how you teach a child the word dog. You give no definition. You point. That is a dog. That too. No, that one is a cat.\n\n[[pause 4]]\n\nNow look again at the cards I proposed. Would it still hold, if? I offered them as a way to find a rule that was already there. They are more than that. They are how a rule gets made. Each card is a case. Each swipe says: like, or unlike. [exhales]\n\n[[pause 4]]\n\nJohn Maynard Keynes, the economist, published a book on probability in nineteen twenty-one. He saw what a new case is for. More cases of the very same kind add little. A new case helps by being different, in ways that turn out not to matter.\n\n[[pause 3]]\n\nThe rat helps because it is not a squirrel. [exhales]\n\n[[pause 4]]\n\nAnd the American philosopher Nelson Goodman saw a circle in this, and called it a good one. We test a rule against the cases we accept. We test a case against the rules we accept. Each is adjusted to the other, until they fit.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Rules in the plural, each with a name beside it. And the cards are kept under each rule. They are its cases.\n\n[[pause 5]]\n\nWhat is real today. The code already asks the question, in its own words. Is the past case \"sufficiently similar to this one in the respect that matters\"? Nobody answers.\n\nWhat this opens. Disagreement with an address, as before. And a rule you can open, to leaf through its cases.\n\n[[pause 3]]\n\nSo, part four in a line. [slowly] Do not ask what the rule is. Ask which cases drew it.\n\n[[pause 6]]\n\n---\n\nPart five. Why the house burned.\n\nThe proposal was this. Look for every small set of sentences that is enough. A link is a reason when at least one of those sets cannot do without its first claim.\n\n[[pause 3]]\n\nThe first example is Peirce again, ten years before the beans. He wrote that reasoning should not form a chain, \"which is no stronger than its weakest link\". It should be a cable. Its fibres may be ever so slender, provided there are enough of them, and they are closely connected.\n\n[[pause 5]]\n\nThe second example is the fire. It needed the short circuit. And the dry curtain. And the missing sprinkler. Take one away, and there is no fire. That is a chain. But the house could also have burned from a candle. That is a second strand.\n\n[[pause 4]]\n\n[inhales] Now your question. In your words: \"whether to weigh different parts as all-necessary to bear strength or separately for what they're worth.\"\n\n[[pause 3]]\n\nThe answer is both. And the fire shows where each belongs.\n\n[[pause 3]]\n\nInside one set of conditions, every part is needed. A chain. The weakest link decides. Between sets, each is a way across in its own right. A cable. Each strand counts for what it is worth, and the strands add. [exhales]\n\n[[pause 5]]\n\nYour August ruling let the weakest part decide. I went back to where that rule was born. As far as I can trace it, it comes from the study of premises inside one argument. So it was a chain rule from the start. It is at home inside one bridge.\n\n[[pause 4]]\n\nWhat is not ruled is the cable. How several reasons for one claim add up. Today the code simply adds them. It never asks whether two strands are one strand, counted twice.\n\n[[pause 4]]\n\nAnd the three friends with the table are a third thing. Any two would do. The code has a setting for that as well. It takes the strongest.\n\n[[pause 3]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Find every way across. Weigh each way as a chain. Weigh the ways together as a cable.\n\n[[pause 5]]\n\nWhat is real today. Chains, we can weigh. The cable is added up blindly. And last night's check, the one that asks whether a written-out step is complete, is the first tool we have for telling a chain from a heap.\n\n[[pause 3]]\n\nWhat this opens. A map that can say: this claim hangs by one chain. That one, by a cable of five thin strands. Which would you rather stand under?\n\n[[pause 4]]\n\nSo, part five in a line. [slowly] Inside one way across, the weakest plank. Between the ways, the sum of the strands.\n\n[[pause 6]]\n\n---\n\nPart six. The madman's weapon.\n\nThe proposal was this. Every rule a machine writes is shown together with the strongest case against it. A case, and never a mark out of ten.\n\n[[pause 3]]\n\nThe first example is on our live map, in the debate about whether a machine could be conscious. The neuroscientist Anil Seth argues by comparison. A computer simulation of a weather system will not give you an actual hurricane. And the processing inside a language model, he says, is no more likely to give you genuine consciousness.\n\n[[pause 4]]\n\nWhat do the two cases share? Each is a computation that stands for something. Does that bear on consciousness? [takes a slow breath]\n\n[[pause 5]]\n\nIt depends. If being conscious is like being wet, a matter of the stuff, the comparison holds. If it is like calculating, a matter of the pattern, it fails. A simulated calculation is a calculation.\n\n[[pause 4]]\n\nWet, or calculating. That fork is in my words, and not in his. But which comparison governs is the whole dispute.\n\n[[pause 4]]\n\nThe second example is last night's. Van den Haag, and the kidnapping. A case that broke the machine's rule, found in the author's own essay.\n\n[[pause 3]]\n\nPut the two together, and you see that a case against can land in two places. One kind says: these two are not alike in what matters. That weakens the bridge.\n\n[[pause 3]]\n\nThe other kind says: here is a case just like yours, where the conclusion failed. That counts against the far bank itself. Our code already keeps the two apart.\n\n[[pause 4]]\n\nAnd I found something I had missed. In August you designed a helper that roams the map, looking for kinship between debates. The analogy daemon. Its design has the same clause. No likeness without its strongest case against. We wrote that rule twice, two months apart, without noticing.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Every rule, and every likeness, is shown with its strongest case against. And where a comparison is disputed, the map shows the rival comparisons side by side. It chooses none.\n\n[[pause 5]]\n\nWhat is real today. Nothing. This one is still untried.\n\nWhat this opens. The button from last night: I know a case. And a second button beside it. I know a better comparison.\n\n[[pause 3]]\n\nSo, part six in a line. [slowly] A rule is tested by the case it cannot hold. A comparison is tested by its rival.\n\n[[pause 6]]\n\n---\n\nPart seven. Whose sentence is it?\n\nThe proposal was this. A rule that a machine writes under your step is marked for what it is. Committed by the step. Not asserted. It becomes yours when you answer the troll. I pay. I cross elsewhere. Or I go back.\n\n[[pause 3]]\n\nThe first example is the banker from last night. He is a banker, so he will vote against the tax. I let the speaker reply: no, I mean this banker, I know him. And I walked straight past it.\n\n[[pause 3]]\n\nIt is the most interesting answer of all. He is not going by a rule. He is going by a case.\n\n[[pause 4]]\n\nThe second example is a mother. She says: do not lend that man money. Why not? She gives no rule. She says: he is just like your cousin. [exhales]\n\n[[pause 4]]\n\nSo the troll needs a fourth answer. Not, I pay with this rule. But, this reminds me of.\n\n[[pause 3]]\n\nAnd you opened that door a month ago. A person says, this reminds me of that. The machine does the expensive part: it works the comparison out, and looks for the case against. Your words then: \"Very useful, document this idea.\"\n\n[[pause 4]]\n\nNow a borrowed voice. Imagine Slavoj Žižek beside you once more. Every word of this is mine, in his manner.\n\n[[pause 3]]\n\nYou think first comes the reason, and then the conclusion? My God, no. It is like love. You do not fall in love because of her eyes, her laugh, and so on, and so on. You fall in love. And then, suddenly, you discover all these reasons.\n\n[[pause 3]]\n\nThe rule arrives after the wedding, with flowers, pretending it was the matchmaker. And your machine, writing the rule under the step, is the perfect wedding photographer. It makes the couple look as if they had planned everything. [exhales]\n\n[[pause 5]]\n\nI take that seriously. Two researchers our notes lean on, Mercier and Sperber, say it soberly. Reasons are mostly found afterwards. A written rule may be a story told after the fact.\n\n[[pause 3]]\n\nThat is one more ground for the mark. Committed by the step. Not asserted.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Committed, not asserted. And four answers to the troll. The fourth is a case.\n\n[[pause 5]]\n\nWhat is real today. The fourth door exists on paper. Nothing under any step is marked.\n\nWhat this opens. Hardly anyone can state their rule. Almost everyone can name their case.\n\n[[pause 3]]\n\nSo, part seven in a line. [slowly] The rule under your step is yours when you say so. And you may answer with a story.\n\n[[pause 6]]\n\n---\n\nPart eight. Agreeing, without agreeing why.\n\nThe proposal was this. A link can be held by people who give different rules for it. The map shows that as a state of its own. And it never demands the rule before it accepts the step.\n\n[[pause 3]]\n\nThe first example is Toulmin's commission, from last night. Case after case, they agreed on what should be done. And they could not agree why they agreed.\n\n[[pause 3]]\n\nThe second example is a courtroom. An American legal scholar, Edward Levi, described how judges actually reason. From case to case. They see a likeness. They announce the rule that was hiding in the first case. They apply it to the second. And the rule shifts a little, every time it is used.\n\n[[pause 4]]\n\nLast night I treated the commission as a curiosity. Today it looks like the normal condition. If rules are drawn from cases, then agreeing on cases comes first. Agreeing on rules comes late. If it comes at all. [exhales]\n\n[[pause 5]]\n\nAnd that sharpens the question I left you with. Is the shared bottom a sentence, or a way of being moved?\n\n[[pause 3]]\n\n[inhales] I can put it more plainly now. Those people sorted the same cases together. A child used without consent, here. A dying man deceived, there. What they shared may be a sense of what is like what.\n\n[[pause 5]]\n\nThe American philosopher Quine held that such a sense is inborn. That we share its beginnings with other animals. And that no learning could start without it. [exhales]\n\n[[pause 4]]\n\nHold that loosely. The same man said that science keeps replacing the inborn sense with better groupings. And people sort the hard cases differently. That is where the quarrels are.\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Held, for different reasons, is a state the map can show. And agreement on a case is recorded for what it is. Firmer than agreement on any rule.\n\n[[pause 5]]\n\nWhat is real today. The map holds a link, or it does not. It cannot yet say: held, for different reasons.\n\nWhat this opens. Two enemies, shown the cases they already sort the same way. Before anyone asks them for a principle.\n\n[[pause 3]]\n\nSo, part eight in a line. [slowly] People meet at the case. The principle is what they write up afterwards.\n\n[[pause 6]]\n\n---\n\nPart nine. A map the size of the country.\n\nThe proposal was this. Not a map with every step written out. A map where any step can be written out, the moment somebody asks.\n\n[[pause 3]]\n\nThe first example is the child again. Why is the squirrel dead? It smells. Why does that mean dead? Dead things rot. Last night I said there is no last answer. There is only bedtime.\n\n[[pause 3]]\n\nListen to how a parent really ends it. How do you know dead things rot? Remember the mouse behind the fridge? [exhales]\n\n[[pause 4]]\n\nThe chain of whys did not stop at a principle. It came to rest on a case.\n\n[[pause 3]]\n\nThe second example is this week's reading. Nobody in it found a floor. A rule rests on cases. Which cases count as alike rests on a wider rule. The wider rule rests on wider cases.\n\n[[pause 4]]\n\nIn plainer words. It is a spiral. Rules, then cases, then rules again, one turn higher each time.\n\n[[pause 4]]\n\nSo the writing-out has a second direction. Downward, into the rule under the rule. And sideways, into the cases behind each rule. Neither direction has an end. [exhales]\n\n[[pause 4]]\n\n[inhales] [slowly] So, the proposal, as it stands now. Any step, when somebody asks. And any rule's cases, when somebody asks. Where a chain of whys comes to rest, the map says what kind of rest it is.\n\n[[pause 4]]\n\nA rule of logic. What a word means. Experience. A norm. And one I had left out. A case we all remember.\n\n[[pause 5]]\n\nWhat is real today. A few dozen written premises, in a graph of thousands of claims. And not one case filed under a rule.\n\nWhat this opens. Fog that clears in two directions. Down, and sideways.\n\n[[pause 3]]\n\nSo, part nine in a line. [slowly] A map that unfolds where you walk. And shows you who walked there before.\n\n[[pause 6]]\n\n---\n\nPart ten. The friend who says: this reminds me of.\n\nThis part is new. It answers what you asked this evening. Your words: \"And the analogy daemon can probably supply answers to the question of alike in what matters?\"\n\n[[pause 3]]\n\nThe first example is a friend. Everybody has one who says: this reminds me of. Half the time it is the best thing said all evening. Half the time it is a stretch. And the friend cannot tell which.\n\n[[pause 4]]\n\nThe second example is an experiment. People read a set of stories. Later they were given new ones. They were reminded far more often by stories that looked alike on the surface. But asked which matches were sound, they picked the ones that worked alike underneath.\n\n[[pause 4]]\n\n[slowly] We are reminded by looks. We are convinced by structure.\n\n[[pause 5]]\n\nSo, can the daemon say what matters? Half of it.\n\n[[pause 3]]\n\nAlike in what matters is two questions. The first: what do the two cases share? A machine can answer that from the cases alone. The second: does the shared thing bear on the conclusion? [takes a slow breath]\n\n[[pause 5]]\n\nThat one cannot be read off the cases. It shifts with what you want to conclude. It is a claim about the world. And it is where people disagree. It is the second number on the link. The bridge.\n\n[[pause 4]]\n\nSo here is the daemon as I would build it now. It lays two cases side by side. It says what they share. It says what is being carried over. It attaches the strongest case against. Where a comparison is disputed, it writes out the rivals.\n\n[[pause 3]]\n\nAnd it never hands you a number. It hands you cases, and plain code counts how they fared.\n\n[[pause 5]]\n\nTwo of my mistakes belong here. This afternoon, from memory, I told you that Douglas Hofstadter backs your thought. He is the man behind the phrase you quoted, analogy as the core of cognition. Then a helper read his essay.\n\n[[pause 3]]\n\nHe says every category is a bundle of analogies. He never argues that rules come from cases. That part is Mill's. And yours.\n\n[[pause 4]]\n\nAnd I had written, in the daemon's own notes, that you ruled its four checks. You did not. I have corrected it.\n\n[[pause 4]]\n\nTwo limits. The daemon will favour the comparisons everyone already makes. The rare one still comes from a person. And it compares cases as somebody wrote them down. A likeness that nobody wrote, it cannot find. [exhales]\n\n[[pause 4]]\n\nWhat is real today. The daemon is a design on paper. One small finder of look-alike structure exists, and it only proposes.\n\nWhat this opens. A quarrel in one debate lights up its cousin in another. And over the years, something nobody has had. A public record of which comparisons held.\n\n[[pause 4]]\n\nGoodman said that our standards of likeness come from our practice of drawing conclusions. He did not think it was the other way round. A map like that would be such a practice, written down.\n\n[[pause 3]]\n\nSo, part ten in a line. [slowly] The machine may say: this reminds me of. People say whether it should.\n\n[[pause 6]]\n\n---\n\nThe view from here.\n\nWhat is real. Every link is a bare arrow. Under each one is a place for the bridge, and every place is empty. The code weighs chains well. It adds up cables blindly. And it already asks, on its own, whether a past case is alike in the respect that matters. Nobody answers.\n\n[[pause 4]]\n\nWhat is possible. Bridges that show their planks, and the feet that wore the path. A map that speaks four dialects. A stranger who is handed the story. Rules you can open, to leaf through their cases.\n\n[[pause 3]]\n\nChains told apart from cables. Rival comparisons, side by side. A troll who accepts a story for an answer. Enemies, shown the cases they already sort alike. Fog that clears down, and sideways. And a helper who says: this reminds me of. [takes a slow breath]\n\n[[pause 6]]\n\nThree things to carry.\n\nThe planks. When a step is written out in full, and every part is sound, should the bridge count for more? Today, showing your work can only cost you.\n\n[[pause 5]]\n\nThe fourth way. From cases to a rule. Does it join the other three?\n\n[[pause 5]]\n\nAnd the bottom. Is what we share a sentence? A way of being moved? Or a sense of what is like what?\n\n[[pause 5]]\n\n[warmly] Fifteen years ago you wrote two sentences about a squirrel. This week they turned out to hold a bridge, four ways of walking, a chain, a cable, and a whole register of cases.\n\n[[pause 3]]\n\n[slowly] That is a good squirrel.\n"}