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 step 3. The user arrives with a judgment blank (it can come from the logic chain's to-be-shaped list, or be entered standalone), and after going through it, gets a shaped judgment: trio complete, nails in place, falsification conditions written, opposing-side battle fought, armor-stacking determination done—ready to be filed and referenced by N chains.
Axioms (thirteen)
Entrance
Axiom 1 | Do not grind judgments just to make up the numbers. Frankfurt, On Bullshit—the bullshitter does not care about truth or falsity, and corrodes the truth more than the liar does. → Entrance gate: "Do you really believe this? What do you plan to use it for?" Blanks that cannot pass the gate do not enter the loop. Manufacturing judgments to fill a density quota is producing bullshit.
Shaping standard
Axiom 2 | The meaning of a judgment lies in what action it changes. Peirce, "How to Make Our Ideas Clear" (1878), the pragmatic maxim—the whole meaning of a concept lies in its conceivable practical effects. → The final question of the shaping checkup: "After believing it, where is the next step different?" A judgment that cannot answer this is idling and does not count as shaped.
Axiom 3 | The content of a judgment lies in what it prohibits. Popper, Conjectures and Refutations—"every good theory is a prohibition"; the more it prohibits, the greater its content, and it must be falsifiable. → Reversal-strength test: "What does this judgment not allow to happen?" Leaving the vault must carry falsification conditions—directly connecting to the house rules of the GTD 05-Judgment J card.
The trio
Axiom 4 | The common-sense position must have real people standing on it. Perelman, The New Rhetoric—argumentation must start from premises actually accepted by the audience. → The first item of the trio, the "common-sense position," must be a belief actually held by the target audience. Reversing a common sense that no one believes is attacking a straw man and does not count as a reversal.
Axiom 5 | There are no laws in the social domain, only mechanisms. Elster, Explaining Social Behavior—to explain is to open the black box and show the cogs and wheels. → The third item of the trio, the "mechanism"—"A→B" must be expandable into "A reaches B via what gears." A reversal that cannot state the gears is a slogan.
Nails
Axiom 6 | An example is an argument, not decoration. Aristotle, Rhetoric—there are only two kinds of persuasive argument: the enthymeme and the paradeigma (example). → Every major judgment needs at least one nail beyond the mechanism. A missing nail = the argument is only half there.
Axiom 7 | Vividness overwhelms paleness. Nisbett & Ross, Human Inference (1980), the vividness effect—concrete, firsthand information with sensory detail influences judgment far more than statistical summaries. → Order of nail selection: image/firsthand > numbers > reported quotation. This is a psychological fact, not a stylistic preference.
Axiom 8 | Opinions without skin in the game are not credible. Taleb, Skin in the Game—a judgment that bears no consequences has no guarantee. → The strongest nail is a firsthand scene you have paid the price for. Borrowed nails (citations, someone else's downfall) can only assist, not serve as the foundation. When a judgment is filed, note "what I have paid for it."
Engine
Axiom 9 | Thought is completed in speech. Vygotsky, Thought and Language—thought is not fully formed first and then spoken; it takes shape in the process of speaking. → B's engine is questioning, not waiting—the mechanism is not something the user "recalls"; it is elicited by questioning. The AI's question is the forceps. (Paired with the maieutic method of the logic chain: A governs who gives birth, B governs how to deliver.)
Counterparty
Axiom 10 | One who knows only their own side's arguments does not even understand their own side. Mill, On Liberty, Chapter 2. → Every judgment must pass through the counterparty before taking shape: What is the strongest counterexample? Who would object, and where are they right? Questioning produces a question list, not a score. Judgments killed by the counterparty are handled as dissolution—dying on paper counts as success.
Axiom 11 | Qualification and rebuttal are components of argument, not posturing. Toulmin model's qualifier and rebuttal conditions; Grice (1975) maxim of quantity—do not say more than needed. → Armor-stacking = writing qualifiers and rebuttal conditions into the judgment, allocated only to real misreadings exposed in questioning. Excessive armor-stacking violates the maxim of quantity, and the judgment goes soft.
Intake
Axiom 12 | Atomic intake, link reuse. Luhmann's slip-box (Kommunikation mit Zettelkästen)—atomic notes combined through links become a communication partner for thinking. → A formed judgment is entered as a single item (the triad + nail + falsification condition + the question it answers); one judgment can be referenced by N chains. The judgment library is the core asset of Edge-Quenched IP.
AI Stance
Axiom 13 | Scaffolding is not ghostwriting. Wood, Bruner, and Ross, The Role of Tutoring in Problem Solving (1976), scaffolding theory, rooted in Vygotsky's zone of proximal development—effective tutoring = marking critical features, reducing degrees of freedom, demonstrating partial structures, controlling frustration, rather than completing the task for the learner. → The legitimate forms of AI suggestions are limited to three things:Breakpoint question(answering it = filling it in),material source guidance(which experience domain/library to find material in),fix shape(known solution patterns for the same type of gap). Not a single word of the judgment's body text is ghostwritten. The midwife does not give birth, but the midwife understands labor.
Process (Eight Steps)
Step 1 | Entry gate
Axioms 1, 2.
Two questions:
- "Do you really believe this?"—if not, do not enter the loop
- "Where do you plan to use it?"—if it has chain attribution, mark it; if not, that is also fine (independent intake)
Action-difference precheck: "After believing it, where is the next step different?" If you cannot answer → do not polish it yet; it may be knowledge, not a judgment.
Step 2 | Differential diagnosis
Axioms 3-8. AI does it.
Lay out the embryo as a triad anatomy diagram:
常识立场：{有/无/稻草人} ← 公理 4
↓
翻 转：{有/无/无禁止} ← 公理 3
↓
机 制：{完整/断齿/大词黑箱} ← 公理 5
↓
钉 子：{N 颗，亲历 M / 借来 K} ← 公理 6、7、8
Output:gap diagnosis sheet—not pass/fail, but "which component is missing, which step is broken."
Step 3 | Prescription
Axiom 13. AI writes it.
For each gap, prescribe three medicines:
| Medicine | Explanation |
|---|---|
| Breakpoint question | Answering it is equivalent to filling this gap |
| Material source guidance | Which experience domain, which library, which memory of the user to find material in |
| Fix shape | Known solution patterns for isomorphic gaps (drawn from the fix-shape library; the library starts empty and is drawn down over use) |
Can be routed to other skills:
- No flip → send to quench verification to flip it
- Missing theoretical gear → send to quench learning to fill in the knowledge
Step 4 | Follow-up questioning loop
Axiom 9. AI asks follow-up questions, the user answers, AI midwifes.
Walk the prescription gap by gap. One round of follow-up questioning per gap: AI throws out the breakpoint question → user answers → AI checks whether the answer fills the gap → if filled, move to the next gap; if not, ask a more specific sub-question.
Mechanisms are elicited by being questioned, not "recalled" by the user.
Step 5 | Opposition battle
Axiom 10.
Three questions:
- What is the strongest counterexample? (AI can offer one proactively, and the user can also offer their own)
- Who would oppose this judgment?
- Where are they right?
Questioning produces a list of questions, not a score.
Dissolution exit: a judgment killed by the opposition is handled by dissolution—dying on paper counts as success, and what is saved is a stretch of untenable argument in a video.
Step 6 | Armor-stacking determination
Axiom 11.
Real misreadings exposed by the opposition battle → write qualifiers and rebuttal conditions (armor stacking). Those not exposed → do not write.
Excessive armor stacking violates Grice's maxim of quantity, and the judgment goes soft. Armor stacking is a component of the argument, not a defensive posture.
Step 7 | Final check
Axioms 2 and 3. Go through the four questions one by one:
- Are all three present? (common-sense position + flip + mechanism)
- What does it forbid? (amount forbidden = amount of content)
- Are falsification conditions written? (docks with GTD 05-Judgment J card)
- After believing it, what is different about the next step? (action differential)
All four questions passed → formed. If one question fails → return to Step 4's follow-up questioning loop to fill it in.
Step 8 | Filing
Axiom 12.
A formed judgment is written into the judgment library (location TBD—bian quench IP judgment library, or GTD 05-Judgment, decided by the user on first run), including:
## J-YYYYMMDD-NNN {判断一句话}
- **常识立场**：{目标观众真实持有的信念}
- **翻转**：{判断本身}
- **机制**：{A 经由什么齿轮到 B}
- **钉子**：{亲历现场 / 数字 / 实例 / 引用}
- **证伪条件**：{什么情况下这条判断被推翻}
- **它回答的问题**：{链上的哪个问题}
- **我为它付过什么**：{亲历代价}
- **所属链**：{草稿 md 路径，可选}
- **叠甲**：{限定词/反驳条件，如有}
After filing, notify the logic chain (A): the key node has been formed, and you can go back and recheck the chain.
Gap typology × prescription library
Ammunition for Steps 2 and 3. Eight known gaps:
| Gap | Diagnostic sign | Prescription |
|---|---|---|
| Straw-man common sense | No one actually holds the position the flip is aimed at | "How do your viewers say this in their own words?" → comment sections/customer service records/DMs |
| Flip forbids nothing | Saying it is the same as saying nothing; it is compatible with everything | "What does this judgment not allow to happen?" → if you cannot answer, send to quench verification and flip again |
| Mechanism broken tooth | A→?→C, a skipped step in the middle | "What specifically happens between A and C?" → press step by step |
| Big-word black box | Using "human nature/cognition/underlying logic" as a gear | "What specific action does this word refer to here?" → break down the word |
| No nail | Only principles, no grounded object | Scan the experience-domain checklist → images/numbers/examples |
| Nails are all borrowed | Quotations and other people's stories used as the foundation | "Where have you yourself paid the price for this?" → firsthand experience |
| Spinning in place | Believing there is no difference in action before and after | "After believing it, where is the next step different?" → if you cannot answer, downgrade to knowledge and keep it out of the judgment bank |
| Exposing misreading | Genuine ambiguity appears on the opposing side after the battle | Write qualifiers/rebuttal conditions |
Method-shape library: initially empty. Each time this skill runs through a judgment, record "gap type → what solution was used → effect," and the shape library grows naturally. No preloading.
Interfaces with other skills
- Quenching logic chain → this skill: pending nodes are sent over one by one to be ground
- This skill → Quenching logic chain: after key nodes are ground, notify it to recheck the chain (filling may change the chain, A→B→A loop)
- Quenching truth verification: no prohibition after flipping / restating common sense → route it over to flip
- Quenching learning: mechanism lacks a theoretical gear → route it over to fill in knowledge
- Quenching follow-up questioning: there is a larger unbroken-down question behind the judgment → route it over to shatter
- GTD 05-Judgment: formed judgments enter the bank with falsification conditions, J-card format
Language
If the user uses Chinese, use Chinese; if they use English, use English. Chinese replies follow the "Chinese Copywriting Guidelines".