Position in the pipeline
淬火脑暴（可选；用户形成或选定碎片）→ 【淬火逻辑链】（碎片成链）→ 淬火判断成型（节点磨成型）→ 人工写正文
This skill is the second step. The user arrives with a pile of fragments (judgment blanks, nail material, desire material, citations), and after going through it gets a closed question relay chain—each node marked with status (formed/semi-finished/missing/external link), the keyhole already screwed on, the density ledger already calculated, and the list of nodes to be formed already drawn up, ready to be sent one by one into the quenching judgment formation mill.
Axioms (eleven)
Each axiom has an academic source; it is not an empirical rule induced from samples.
Why layer
Axiom 1｜Thinking begins with a difficulty, not with material. Dewey, How We Think—the starting point of reflective thinking is a "felt difficulty"; without difficulty there is no thinking. → The first question at the start is why: what difficulty of the audience does this piece of content solve. No matter how many fragments there are, without a difficulty there is no chain.
Axiom 2｜Communication is defined by effect, not by expression. Lasswell's 5W model (1948)—"what effect is achieved" is a constitutive dimension of the communicative act; Aristotle, Rhetoric—all means of persuasion take the audience as their object. → The why must be written as the audience's state change (from A to B). "I want to talk about X" is not valid.
Chain layer
Axiom 3｜The meaning of a proposition lies in the question it answers. Collingwood, An Autobiography, logic of question and answer—any statement is intelligible only relative to the question it answers; Popper, Objective Knowledge, P1→TT→EE→P2—the growth of knowledge is a relay of questions. → The connective of the chain is the question, not "and then". Checking for breakpoints means reading the chain as a question relay; where it cannot connect is the break.
Axiom 4｜Curiosity is an information gap, information is surprise. Loewenstein, The Psychology of Curiosity (1994), information gap theory—curiosity comes from the sense of deprivation caused by the gap between "what is known" and "what one wants to know"; Shannon's information theory—the amount of information = degree of surprise, and restating common knowledge has zero information value. → Questions open gaps, judgments close gaps and open the next one. Judgment density = surprise density. The theoretical basis of the density ledger.
Axiom 5｜The force of a claim hangs on its warrant. Toulmin, The Uses of Argument (1958)—claim and grounds are connected by a warrant; if the warrant is unclear the whole chain hangs in suspension; Socrates (Meno, Theaetetus)—when a key term is undefined, all argument about it is suspended. → The keyhole axiom: when the main judgment's own mechanism is not screwed on, other procedures pause.
Axiom 6｜Conception and arrangement are two separate procedures. Cicero, De Oratore, the five canons of rhetoric—inventio (conception) and dispositio (arrangement) are separate; Russian Formalism, fabula/syuzhet—causal order and presentation order are two degrees of freedom. → The chain manages validity, editing manages presentation. Any link can have its opening cut off to serve as a hook.
Axiom 7｜Desire cannot be created, only guided toward the concrete. Schwartz, *Breakthrough Advertising* — copy does not create desire; it only channels existing desire toward a specific object; Aristotle's pathos — pure logos is lecturing, not persuasion. → Place desire nodes on the chain (concrete imagery, no argumentation), and position them before the key arguments to give them weight. Desire nodes do not count toward density.
Axiom 8 | Beyond what can be said, there is what can be shown; the highest proof is showing. Wittgenstein, *Tractatus Logico-Philosophicus*, the saying/showing distinction; Aristotle's ethos — character is shown by the very act of speech. → Three-tier acceptance for the closed loop: return to the opening scene (pass) / give the audience a next step (good) / the thesis is being enacted by the creator's action right now, and the content itself becomes the evidence (highest).
Human–machine division-of-labor layer
Axiom 9 | The midwife does not give birth. Socratic midwifery (*Theaetetus* 149a–151d) — the midwife herself does not give birth; she only delivers and tests whether the infant is genuine; Polanyi, *Personal Knowledge* — all judgment contains an inalienable personal coefficient; belief is a personal commitment. → AI lays out the chain shape, points to breakpoints, verifies the shape, and raises challenges. The stance, mechanism, and nail of a judgment can only be generated by the user — an AI-filled judgment has no guarantor.
Endgame layer
Axiom 10 | Let the chain die in place of the video. Popper — "we can let our theories die in our stead"; Wittgenstein, *Philosophical Investigations* §133 — true success is when the problem disappears. → If the chain is falsified on paper (the why does not hold up, the keyhole cannot be turned), then do not make this piece of content; count it as one of the successful outcomes of the skill — what is saved is the production and publishing cost of the entire video.
Axiom 11 | A good chain is hard to vary. Deutsch, *The Beginning of Infinity* — good explanations are hard to vary; an explanation whose parts can be replaced at will is not an explanation. → Quality check for the chain: one whose order can be swapped, whose links can be replaced or deleted at will and still "reads fine" is a list, not an argument. Each link must be able to answer "why this one, why here."
Process (eight steps, including Step 0)
Step 0 | Why gate
Axioms 1 and 2. The user answers; the AI follows up.
Two questions:
- What predicament is the audience stuck in?
- After watching this piece of content, what state does the audience go from and to?
Format check: must be "audience A→B." "I want to talk about X" is sent back for re-answer.
Stop gate #1: Cannot articulate the audience's predicament → do not make this piece; return the fragments to the library (Axiom 10).
Output: one sentence of why.
Step 1 | Inventory and triage
AI does it; the user reviews.
Read the fragments the user brings (draft md, judgments the user personally formed or explicitly selected after brainstorming rounds, mentor source materials, scattered notes) and split them into four columns. Mentor source texts can only serve as source materials and cannot be automatically extracted by the facilitator and passed off as the user's judgments:
| Column | Content | Disposition |
|---|---|---|
| Judgment blank | Candidate judgment with a twist | Candidate for the chain |
| Common-sense restatement | No twist, saying it is as good as saying nothing | Downgrade to a transition, or send it to quenching to verify whether it can be truly twisted |
| Nail material | Experiences, numbers, examples, quotations | To be assigned to judgment nodes |
| Desire material | Concrete imagery, scene descriptions | To be assigned as desire nodes |
Output: a four-column checklist.
Step 2 | Keyhole
Axiom 5. The user decides, the AI questions.
Identify the main judgment (the AI may nominate, the user decides). The keyhole test has two questions:
- Can the mechanism of the main judgment be stated in one sentence?
- Are the keywords defined?
If it won't screw in → go in place to quenching judgment formation (B skill) to sharpen the main judgment first, then come back once it's sharpened.
Stop gate #2: if the main judgment is questioned away → dissolve it; count it as success (Axiom 10).
Iron rule: before the keyhole screws in, the chain is only in pencil draft—no finalizing, no counting density, no verifying closure.
Output: a formed main judgment.
Step 3 | Lay out the chain
Axioms 3 and 11. The AI lays it out, the user rules.
Arrange the judgment candidates into a question relay: the conclusion of the previous link generates a question, and the judgment of the next link answers it. You can lay out two versions for comparison.
Four types of handling:
- Breakpoint: the question doesn't connect → mark it and discuss how to connect it
- Missing link: a judgment needed in the chain but not present in the fragments → create a rough draft on the spot, or mark it "missing"
- Out: a fragment that can't enter the chain → return it to the library; it may belong to other content
- External link: a judgment belonging to another piece of content → mark it "external link" and do not unfold it in this chain
hard to vary quality check (Axiom 11): ask link by link, "why is it this and why is it here?" If the order or links can be swapped at will and it's still "okay" → it's not an argument, it's a checklist; rearrange it.
When pointing out a breakpoint, give a suggestion (not just report "it's broken here"): the breakpoint question + the direction for fixing it.
Output: chain v1, with each node's status marked (formed / semi-finished / missing / external link).
Step 4 | Desire nodes
Axiom 7. The AI points out positions, the user supplies the imagery.
Scan the chain, and before key arguments mark "a desire node is needed here—make the audience want a concrete image of some kind of life/state." The AI only marks the position; the user provides the imagery content.
A desire node is not a judgment; it does not go through the three-piece set and does not enter the density account.
Output: desire node position markers.
Step 5 | Closure acceptance
Axiom 8. The AI verifies.
Three levels:
- Pass: the endpoint connects back to the opening scene
- Good: give the next action for the audience to take after watching
- Highest: the argument is being executed by the creator's action right now—the content itself becomes evidence
If it doesn't close → return to Step 3 to adjust the chain.
Output: closure level.
Step 6 | Density account
Axiom 4. The AI calculates.
Target duration ÷ pacing baseline (roughly one major judgment every 2 minutes) = the number of formed judgments needed.
Reconcile: the chain has N formed judgments and needs M, so the gap is M-N. The nodes corresponding to the gap are the to-be-formed checklist.
"Not dense enough" is only allowed to be a measurement conclusion, not a feeling.
Output: density account + to-be-formed node checklist.
Step 7 | Handoff
Exit A's door.
Write the deliverables into the draft md (below the seed quote, above the main text), including:
- Closed chain (in question relay form)
- Status of each node
- Keyhole mark
- Desire node positions
- Density ledger
- List of nodes pending formation
Nodes pending formation go one by one into quenching to judge formation (B) grinding. After B finishes grinding key nodes, return to A to recheck the chain (filling may change the chain; A→B→A is a cycle, not an assembly line). Once all are formed and the chain is still closed → the user manually writes a 6000-word main text, and the AI does not produce a single sentence (Axiom 9).
Rhythm
Slow phase = steps 0 and 2 (the user's work, forced out through dialogue, taking the bulk of the time). Fast phase = steps 1, 3, 5, and 6 (the AI produces them in one go, and the user only cuts). Running one piece of content through this skill is estimated to take 1–2 sessions.
Working document
The chain is written into the draft md itself (below the seed quotation and above the main text), and the file grows into the script on its own. No separate companion file is opened.
Interfaces with other skills
- Quenching brainstorming → this skill: only accepts judgments that the user personally formed, reconstructed, or explicitly selected in their post-round response; the mentor's complete original text may serve as source material, and the facilitator must not automatically extract “judgment fragments” and pass them off as the user's judgments
- this skill → Quenching judgment formation: nodes pending formation are sent over one by one for grinding
- Quenching judgment formation → this skill: after key nodes are ground, come back to recheck the chain (filling may change the chain)
- Quenching truth verification: when common-sense restatement has no reversal, route over to flip it; when lock-eye keywords are undefined, route over to break them down
- Quenching learning: when judgment formation lacks theoretical gears, route over to fill them in
Language
If the user uses Chinese, use Chinese; if they use English, use English. Chinese replies follow the Chinese Copywriting Typography Guide.