This skill is used to write Edge Quenching 2.0 long-form articles. The core of 2.0 is not explaining knowledge, but pushing readers from "I am preparing" to "I must enter market feedback."
2.0 themes include AI, business, making money, entrepreneurship, content monetization, one-person business, product hypotheses, market feedback, traffic, communities, consulting, tools, and value creation. 1.0's inner game material can enter the main text as evidence or a mirror, but it cannot steal the commercial main axis.
Hard Rules
- Do not write from memory. When benchmarks, topic libraries, original texts, or nodes are involved, local files must be read first.
- Do not write content as "explaining a problem." The content should make readers feel that the old problem is beginning to loosen, and be willing to hand over the real problem.
- Do not treat the product as a preemptive answer. Before the product is finalized, the content serves a paid problem hypothesis.
- Do not write AI as an independent trend. AI can only be placed in the relationship of market feedback, value creation, user service, or amplification of business capability.
- Do not use parallelism and juxtaposition of synonymous sentence patterns to create momentum.
- Do not write standalone golden lines. Every sentence must serve a behavior, mechanism, case, or feedback chain.
- It must pass through user confirmation step by step. Reading the library, expanding nodes, setting the nail, assembling the logic chain, writing the first draft, and reviewing and revising are a continuous process, but key judgments cannot automatically skip the user. At each step, first give a high-density intermediate result, and wait for the user to confirm, revise, or reject before entering the next step.
Anti-Parallelism Prohibition
When writing and revising, actively scan for these bad smells:
- Starting consecutive sentences with the same sentence pattern.
- Using "not X, but Y" consecutively.
- Listing three synonymous judgments in a row.
- Ending with "let A, let B, let C."
- Using "no A, no B, no C" to create pressure.
- Using stacked abstract words instead of concrete behavior, for example only saying "cognition, energy, system, inner core, underlying logic."
- Using pretty short sentences to skip causality, for example "The market will educate you. Feedback will hurt you. Fate will screen you."
Lists are allowed, but lists must carry a structural function: node summaries, operational steps, review items, evidence tables. In the main text, try to advance with causal sentences, scene sentences, and action sentences.
Workflow
1. Read Entry Materials
Prioritize reading the topic file given by the user. If the user only says "the next 2.0," first look in the current workspace or the content library specified by the user for:
10-选题库/01-待发布10-选题库/02-已发布
If the content library cannot be discovered, only ask the user for a content library or topic file path; do not guess a fixed directory on some computer.
If the topic file cites external benchmark nodes, continue reading the corresponding files:
内容对标库/01-原始素材区内容对标库/02-内容单元区
Do not read only the topic unit. Trace at least to the original material, viewpoint unit, problem unit, desire unit, and method unit. Opening units and case units should be read as needed.
2. Expand Surrounding Nodes
Use rg to search source_id, title keywords, core concepts, and author names, and pull out three layers of material:
- Main node: the original text, topic, viewpoint, or method directly cited by the current topic selection.
- Same-source node: the QST, OPI, DES, MTD, HOK, and CAS split out from the same source_id.
- Adjacent node: nodes with similar themes but different sources, used to supplement counter-evidence, operational steps, and personal material.
Nodes are not about quantity; they must be able to enter the logic chain. Each node must be labeled with its function in the article: entry, counter-evidence, mechanism, case, operation, closing.
3. Nail review
Before writing, nail down these four judgments:
- Nail: the single judgment this article must drive through.
- Old belief: what the reader currently believes, and therefore where they get stuck.
- Behavioral evidence: what action this old belief manifests as in reality.
- Paid-question hypothesis: which type of person this piece of content wants to verify is willing to pay for which type of question.
Without a paid-question hypothesis, do not directly write a 2.0 long-form article. First downgrade the topic selection into research or conversation.
4. Assemble the logic chain
By default, use this staircase, adjust according to the material, but it cannot become parallel viewpoints:
- The reader thinks they are starting a business, learning AI, building a personal IP, or preparing a product.
- Behavior exposes the real goal: they would rather preserve the identity of “not having failed yet.”
- This identity protection traps people in the preparation zone: building systems, asking for advice, looking at benchmarks, changing tools, thinking about products.
- Business starts from feedback: content, consulting, communities, and light products are all just ways to bring value hypotheses into contact with reality.
- The market will correct the product: who approaches, who asks questions, who pays, who stays silent—these are more reliable than analysis in the head.
- AI can only amplify this feedback system; if there is no contact with the market, AI will amplify the preparation zone.
- Land on one action executable that same day: publish a piece of content that verifies the paid-question hypothesis, or have a conversation with a real user.
Each layer must be more specific or deeper than the previous one. If it still holds after the order is swapped, the logic chain has failed.
5. Main body writing
By default, write 2000-4000 Chinese characters, unless the user specifies a length. The main body may have no subheadings; if subheadings are needed, use them only to change the reading rhythm, not to explain reasoning.
The first 100 characters must complete one belief disruption. Do not set up the background of the era, and do not explain concepts first.
Personal experience accounts for 15-25%. Experience does only three things: prove that you have also fallen into it, show the cost, and turn external viewpoints into something quenched by your own edges.
Take only one cut from external theory. Do not explain the complete theoretical system.
The action section should be short and not a universal tutorial. The purpose of a 2.0 long-form article is not to solve the problem all at once, but to push the reader forward to a next step that can generate feedback.
6. Final draft review
Do one round of internal review before delivery:
- Does the first 100 characters kill the old belief?
- Is there only one nail?
- Do three or more synonymous sentence patterns appear in parallel?
- Does continuous parallelism appear?
- Is there a sentence that only sounds good but does not carry a function of advancement?
- Is it clearly written how behavior exposes the real goal?
- Is it clearly written how market feedback enters?
- Is it clearly written what the relationship is between content and product?
- Is it clearly written where AI is positioned in the feedback system?
- Can the reader finish and want to ask themselves a question, rather than only thinking the author is smart?
When you find parallelism or vague inspirational lines, prioritize changing them into specific actions, real-world costs, user feedback, or personal cases.
Common local nodes
When needed, prioritize searching these relative directories in the user-specified content library or the current workspace; do not assume they definitely exist:
内容管理库内容对标库内容管理库/02-内容单元库/案例单元内容管理库/03-欲望库/欲望单元
2.0 Common keywords:AI、商业、赚钱、创业、产品假设、市场反馈、一人事业、内容变现、咨询、社群、客户、价值创造、工具。
Output method
When the user asks to "write it out," directly provide the complete first draft. When the user asks to "see the logic chain," provide only the nodes, nails, and stairs. When the user asks to "put it into the library" or "save it," then write it to the specified location; do not force the production of files just for a sense of process.
Even if the user asks to "write it out," you must not skip confirmation. First complete the node expansion and the nail/stair design, and wait for the user to confirm; after confirmation, then write the complete first draft.
Step-by-step confirmation points:
- Node expansion confirmation: show the main node, surrounding nodes, and their respective functions, and wait for the user to confirm whether they are sufficient.
- Nail confirmation: show the nails, old beliefs, behavioral evidence, and paid question hypotheses, and wait for the user to confirm.
- Stair confirmation: show the progressive logic chain, and wait for the user to confirm whether it holds.
- First draft confirmation: after writing the complete first draft, let the user revise the wording, materials, and tone.
- Review confirmation: execute the anti-parallelism and 2.0 quality review, provide revision suggestions, and wait for the user to decide whether to revise v2.