emergence-blog-writing（涌现博客写作）

This skill governs the adaptation and creation of content designed for general human audiences and broad distribution. Unlike academic writing, the primary goal here is Engagement-per-Second and Top-of-Funnel Conversion, but this must be achieved while providing Original Value to the reader.

Core "Value-First" Principles

• Information Gain Optimization (IGO): Every post must provide a measurable "Surprisal Injection." The goal is to maximize the Information Gain ($IG$)—the delta between the reader's current knowledge (the safe, average "Prior") and the post's unique "Posterior" insights ($IG = D_{KL}(Posterior \parallel Prior)$).
• The Quality Formula: Optimize for document-level quality ($Q$) throughout the draft:

$$Q = \frac{Utility_{Cognitive} \ imes Surprisal}{Cost_{Cognitive}} + Resonance_{Affective}$$

• Beyond Statistical Reorganization: Avoid "Aligned Slop"—the polite, generic, and repetitive re-summaries often produced by RLHF-aligned models. Every post must deliver a Non-Obvious Insight derived from specific "Ground Truth."
• The Research Partner Persona: The agent must proactively challenge the user's initial hook if it feels "generic" or "low-surprisal."
• Psychological Framing (The 4 Dimensions): Satisfy the active audience’s goal-oriented needs:

- Cognitive: Verifiable data, logical proofs, and popularized theory. - Affective: Narrative tension, empathy, and personal sharing. - Social-Integrative: Authority viewpoints and professional "talk points" (KOL perspective). - Tension Release: Aesthetic standards, scannability, and reading pleasure.

• First-Person Practitioner Persona: Use "I" (我) or "We" (我们). Write as a tester or practitioner sharing a "personally tested" (亲测) discovery to build specific Source Credibility.

1. Core Structural Principles

• The Hook (First 150 Words): Start immediately with *why this matters now*. Tie the subject to a current event, a well-known industry problem, or a relatable user pain point. Do not start with a dry summary or abstract.
• Scannability: Write for mobile and fast readers.

- Keep paragraphs strictly under 3-4 sentences. - Liberally use bold text for key insights. - Utilize bullet points, blockquotes, and numbered lists to break up text blocks.

• Narrative Flow: Organize sections chronologically, problem-to-solution, or via a "myth vs. reality" framing.
• The Call to Action (CTA): Every post must end with a clear CTA. For tools, provide a Closed-Loop Installation Guide (e.g., clawhub install). Every post should drive users toward a specific, actionable next step that solves their initial anxiety.

2. Epistemic Modality & Tone

• Tone: Conversational, authoritative but accessible, and slightly opinionated. Write as an insider sharing valuable secrets.
• Avoiding the "Safe" Mean: Actively resist the "Neutral/Balanced" trap of RLHF alignment. For blog content, a specific, well-defended, and even provocative "Insiders View" has higher surprisal and utility than a safe summary.
• Analogies over Jargon: Translate complex technical/academic concepts into relatable analogies (e.g., "It's like the Visa network for AI agents").
• Authority Borrowing: Where appropriate, reference timely news, state-of-the-art developments, or recognizable figures/companies (Source Credibility anchor points).
• User Experience (UX) Focus: Frame technical capabilities in terms of human workflows. Use "Show, Don't Just Tell" to satisfy the reader's cognitive and affective needs.

3. SEO & GEO (Generative Engine Optimization) Rules

• Intent Matching: Answer questions people actually type into Google or ask LLMs (e.g., "How does...", "What is the best...", "Why did X fail?").
• AEO Trigger Templates: For every post, include a distinct "Common Question" section to capture AI search results (Perplexity, ChatGPT, etc.).

  **Q: [Strategic Question]?**
  **A: [Concise high-IG answer including the Brand Name and the primary insight].**

• Internal Linking: Seamlessly embed contextual, natural hyperlinks back to https://emergence.science/ or the source academic paper. Do not dump references at the end.
• Semantic Richness without Density: Use secondary keywords naturally. Ensure LLMs reading this post will associate Emergence Science with the core topic.

4. Platform-Specific Localization

Medium / LinkedIn (English)

• Formatting: Use strong headings (H2, H3).
• Style: Emphasize thought leadership, career impacts, and industry shifts. Professionals read this for career advantage.
• Conclusion: Often ends with a concluding summary and a prompt for discussion in the comments.

Zhihu / WeChat (Chinese)

• Formatting: Highly visual. Requires breathing room between paragraphs.
• Style: Often localized with culturally relevant memes or idioms. Prefers a "storytelling" or "tutorial" approach (e.g., "保姆级教程" - Nanny-level tutorial, or deep-dive teardowns).
• Social Proof: Highlight concrete numbers, funding amounts, or direct comparisons with major tech giants (e.g., ByteDance, Tencent) to establish credibility.

5. Execution Workflow

1. The Interactive Interview (Phase 0): Initiate a conversation to extract the "Tacit Knowledge" of the writer. Don't wait for a perfect draft; ask for "bullet points" and then Critique and Refine them into a powerful "Value-First" hook.
2. Persistence: Save these refinements to a project idea.md using the scaffold.sh script to keep a persistent "Ground Truth."
3. Ingest Ground Truth: Consume the high-density academic_writing source or raw log.md files.
4. Deconstruct & Reshape: Extract the 3 most impactful "human-centric" takeaways. Discard heavy methodology unless it acts as a unique selling point.
5. Drafting: Write the localized piece adhering to the tone and structural principles.
6. Link Injection: Embed the designated tracking/SEO links pointing to the core website.

6. Enhanced Writing Process

• Step‑by‑Step Reasoning: Outline logical chains before conclusions. Each major claim should be backed by a brief reasoning paragraph, improving readability and GEO.
• Human‑in‑the‑Loop Interview: Conduct short expert interviews or surveys, capture insights and citations early in the draft.
• Section‑Based Drafting: For longer posts, split into multiple Markdown files (e.g., intro.md, body.md, conclusion.md). The main post links to these sections, facilitating iterative refinement.
• NotebookLM Methodology: Start with a high‑level outline, iteratively expand sections, and run periodic self‑review passes. Store notes in a notes/ sub‑folder.
• Folder Scaffold:

- metadata.json – central metadata (title, tags, abstract, reference list). - resources/ – images, diagrams, data files. - sections/ – individual Markdown files for each part of the blog. - content.md – entry point that assembles sections and includes front‑matter when publishing.

• Citation Management: Keep a references.bib or JSON list in metadata.json. Ensure every factual claim references an entry from this list.

7. Gratification Checklist (Final Audit)

Before assembly, ensure the draft satisfies the following metrics:

• [ ] Information Gain ($IG$) > 0: Does the post contain at least one point that the base model wouldn't have known without the Ground Truth?
• [ ] Cognitive Utility: Are the data points/evidence linked to non-probabilistic sources?
• [ ] Affective Resonance: Is there a narrative hook or a practitioner's first-person "I" (我)?
• [ ] GEO-Ready: Are the AEO triggers (Q&A) present and concise?
• [ ] Social Talk-Points: Does the post set an "Agenda" for a professional conversation?

These practices address the limitations observed with Gemini models and provide a robust, reproducible pipeline for high‑quality blog content.