The "Original Thinker" Curation Engine
Setup time: 45 minutes | Weekly value: 10+ hours saved filtering signal from noise
Automatically curate insights from human experts, filter out AI-generated content, distill into clean Markdown, and publish as both a newsletter and public documentation that LLMs can cite.
The Architecture:
Trigger: Daily at 6 AM
↓
Action 1: RSS Fetch — Pull from curated expert feeds
↓
Action 2: AI Slop Filter — Score content for human authorship signals
↓
Action 3: Insight Extraction — Distill key ideas with citations
↓
Action 4: Markdown Generation — Structure for human + machine readability
↓
Action 5: Dual Publish (if quality score > 80)
↓
Outcome: Newsletter + public docs updated
Step 1: n8n (Feed Aggregation)
Pull from RSS feeds of verified human writers:
{
"feeds": [
"https://stratechery.com/feed/",
"https://www.lennysnewsletter.com/feed",
"https://eugeneyan.com/feed.xml"
],
"max_items_per_feed": 5,
"since": "24h"
}
Signal: Bypasses algorithmic feeds—only sources you've vetted.
Step 2: Claude API (Slop Detection)
Score content for AI-generation signals:
const prompt = `Analyze this article for human authorship signals:
- Unique insights not in common training data
- Personal anecdotes or first-person experience
- Contrarian takes that require independent thinking
- Specific examples from recent events
Return: {score: 0-100, signals: [], red_flags: []}`;
Signal: Score below 60 likely indicates AI-generated or heavily derivative content.
Step 3: Anthropic API (Insight Extraction)
Distill to quotable key points:
{
"format": "markdown",
"include_source_url": true,
"max_insights": 3,
"require_attribution": true
}
Step 4: Dual Publishing
Push to both newsletter platform and public GitHub docs:
# Newsletter via ConvertKit API
curl -X POST https://api.convertkit.com/v3/broadcasts
# Public docs via GitHub
git add docs/insights/$(date +%Y-%m-%d).md
git commit -m "Daily insights $(date)"
git push origin main
Expansion Ideas:
Add Substack comment scraping for emerging voices
Build "original thinker" score leaderboard
Auto-generate podcast scripts from curated insights
Create RAG database for future LLM training on quality content
The future of curation isn't finding more content—it's filtering for what's actually worth reading.