Stop just copying "benchmark accounts" anymore! Teach you how to distill your favorite bloggers into AI skills in 10 minutes, for unlimited reuse.

Many people do benchmarking, just copying their sentences (What), without learning their thinking (How). The result is: every time you write, you have to start from scratch, exhausting.
The real approach is: extract the blogger's "operating system" and turn it into your AI skill package.
👇 Sharing a set of the RIA-TV++ five-step extraction method I summarized:
1️⃣ R - Read (读取) Don't just look at a few articles, use tools (web crawlers/MCP) to batch capture the 50-60 most popular notes from the blogger.
Key fields: title, content, likes, favorites, comments. Data is insufficient, AI can't see patterns.
2️⃣ I - Identify (识别三层系统) This is the core! Let AI analyze these 60 pieces of content and extract three levels of logic:
Cognitive layer (Core Belief): What does he believe? (Example: Finance is not analysis, but translation. Translate major events into their impact on the user's wallet.)
Strategy layer (Strategy): How does he approach? (Example: 97% pure video, never do serials, each episode independent.)
Content layer (Formulas): What are the viral formulas? (Example: Macro events + specific impact on your life.)
3️⃣ A - Archive (归档) Organize the identified three layers into a structured Markdown file. ⚠️ Key point: this is not a note for people to read, but a "static asset" for AI.
Turn flowing notes into static assets—that's the core of Skill.
4️⃣ T - Template (模板化/翻译规则) Convert specific examples into general rules.
Original sentence: "Trump tariffs, how do they affect your mortgage?"
Rule: [Grand narrative/Hot topic] + [Impact on user’s personal interests]
Reuse: AI boom, how does it affect your salary and career longevity? That’s modeling, not copying.
5️⃣ V - Validate (验证) Take a topic the blogger has never written about, use this Skill to have AI write an article.
If it sounds like him, congratulations, the Skill is successful! If not, go back and redo step 2.
💡 Core principles:
✅ Benchmarking ≠ Modeling. Benchmarking is pointwise learning; modeling is compound system.
✅ Produce two outputs: one is a "report" for people (a副产品), the other is an "AI Skill" (true asset) for AI.
✅ 30/70 division of labor: you handle 30% (select bloggers, set directions, make judgments), AI handles 70% (read, summarize, write).
After three months, while others are still struggling with benchmarking, you will already be calling 100 blogger models.