From Topic Selection to Publishing: dontbesilent's Claude Code Workflow

💡 This article showcases a mature systematic content production system: from topic recording, material retrieval and reuse, title/cover generation, data-driven methodology refinement, to business operations — forming a complete creative loop. The output of 13,000 pieces of content per year is backed by a system-thinking approach worth learning from.

Due to the complexity of the process, this article was written by Claude Code on my behalf.

Why Read This Article?

Many people see my Claude Code workspace directory screenshot and think: “Does a content creator really need something this complex?”

But when I tell them I use this system to:

• Publish 13,000 pieces of content per year
• Simultaneously run 7 platforms with 10K+ followers each
• Gain 700K new followers annually

Their response becomes: “Can you teach me?”

This article is the answer.

Core Philosophy: From Fragments to Systems

Most people use AI like this:

Have an idea → Ask AI → Get answer → Publish → Forget → Repeat

This is fragmented creation — reinventing the wheel every time.

My approach:

Have an idea → Record in topic library → AI searches material library → Reuse validated frameworks → Publish → Data review → Refine methodology

This is systematic creation — every creation adds to the system.

Complete Workflow: From Idea to Publication

Step 1: Topic Recording (Fragment Management)

When I have an idea, I say: “Record topic”

AI automatically saves it to 01-Content-Production/Topic-Management/00-Topic-Records.md

This file is my “idea inbox” — all fragments go here first.

Step 2: Topic Development (From Idea to Draft)

When I decide to develop a topic, AI will:

1. Search the material library (this is key)

   • Check Core-Concepts/: any relevant theoretical frameworks?
   • Check Quotes-Library/: any quality expressions?
   • Check 03-Published-Topics/: any related drafts?

2. Suggest reuse — if related content exists, AI suggests reusing rather than creating from scratch.

3. Generate draft — if nothing reusable exists, AI generates based on my style.

Step 3: Title & Cover Generation

AI generates based on my methodology doc:

• 3 title options
• Cover text suggestions
• Title logic explanation

Step 4: Short Video Opening Optimization

AI optimizes the first 5 seconds for completion rate, based on historical performance data.

Step 5: Publishing & Data Recording

After publishing, I move the draft to 03-Published-Topics/ and record:

• Publication time
• Performance metrics (views, likes, comments)
• Review thoughts

This data feeds back into Content-Data-Statistics/ and Methodology-Refinement/.

AI Automation Capabilities

Defined capabilities include:

• Record topics: Quick fragment capture
• Generate titles: Based on methodology
• Optimize openings: Improve 5-second completion rate
• Search materials: Find reusable content
• Data review: Record and feed back to methodology

These make AI not just “answer questions” but “execute workflows.”

Material Library: From Luck to System

This is the system’s core.

Traditional approach: Start from zero every time, lose track of past work, good expressions scattered everywhere.

My approach: AI searches material library before writing, suggests reuse, every creation enriches the library.

Content-Material-Library/
├── Core-Concepts/           # Reusable frameworks
├── Quotes-Library/          # Quality expressions
├── Hit-Drafts/              # Validated content structures
├── 100-Thoughts-Series/     # 100 thoughts from 10,894 tweets
└── Tweet-Library/           # 10,894 raw tweets

Data-Driven Methodology

Not “by feeling” but “by data.”

• Content statistics: Performance data for all content
• Methodology docs: Data-derived title patterns, topic-title-opening relationships
• Topic research: Benchmarking against Dan Koe’s popular topics

Every publication feeds data back into methodology. The methodology is “alive” — constantly iterating with data.

Business Operations: Content Is the Means, Business Is the Goal

Many creators focus only on traffic, not monetization.

My 02-Business-Operations/ directory tracks:

• Revenue data and analysis
• Business logic and strategic decisions
• Per-business-line statistics

Content is the traffic entry point; business is the monetization exit.

Core Value: Systematic > Fragmented

The system’s core value isn’t “AI can write drafts” but:

1. Memory system: AI knows what I’ve written before
2. Material reuse: Good frameworks and expressions get repeated use
3. Methodology refinement: Every creation strengthens the system
4. Data-driven: Iteration by data, not feeling

How to Start?

1. Build directory structure: Organize by business process, not file type
2. Define workflows: Teach AI your repetitive processes
3. Build material library: Capture good content, frameworks, expressions
4. Drive with data: Record from day one

Don’t pursue perfection from the start. Begin with one small workflow and iterate.

Let Claude Code Build It For You

Just send this article to Claude Code with your specifics:

I want to build a content production system like dontbesilent's.

My situation:
- Content type: [short video/graphics/audio]
- Platforms: [Xiaohongshu/Douyin/WeChat/...]
- Pain points: [starting from zero each time/can't find past materials/...]

Please help me:
1. Design a directory structure for my needs
2. Build a material library management system
3. Define AI automation capabilities I need
4. Create a CLAUDE.md project guide

Reference this article's approach:
[paste this article's link or content]

Tips:

1. Start small — solve one specific pain point first
2. Iterate as you go — systems are built through use
3. Record data from day one
4. Review weekly or monthly

Summary

This system exists to enable:

• 13,000 pieces of content per year
• 7 platforms simultaneously
• Stable content quality
• Continuous methodology refinement

If you want to upgrade from “fragmented creation” to “systematic creation,” this article is your starting point.

Source: Reposted from @dontbesilent12 (dontbesilent)’s original post, published February 1, 2026.