AI Newsletter Curation

Fetch, deduplicate, and rank AI newsletter content into a daily digest.

RSS Sources

• TLDR AI: https://bullrich.dev/tldr-rss/ai.rss
• The Rundown AI: https://rss.beehiiv.com/feeds/2R3C6Bt5wj.xml

Workflow

1. Check cache: Look for 50_资源/Newsletters/YYYY-MM/YYYY-MM-DD-摘要.md. If exists with today's date, return cached content.

2. Fetch feeds: Use WebFetch on both RSS URLs. Extract title, link, pubDate, description for each item.

3. Deduplicate: Merge items with similar titles (80%+ word overlap). Keep longer description, track both sources.

4. Rank items by:

   • AI relevance (LLM, GPT, Claude, agents, ML keywords)
   • Productivity relevance (workflow, automation, tools, PKM)
   • Recency (newer = higher)
   • Novelty (check recent archives, penalize repeats)

5. Generate digest: See TEMPLATE.md for format. Include:

   • 精选推荐 (3-5 highest scoring) with content creation angles
   • AI动态 section
   • 生产力工具 section
   • Stats footer

6. Save files:

   • 50_资源/Newsletters/YYYY-MM/YYYY-MM-DD-摘要.md (curated)
   • 50_资源/Newsletters/YYYY-MM/原始数据/YYYY-MM-DD_TLDR-AI-Raw.md
   • 50_资源/Newsletters/YYYY-MM/原始数据/YYYY-MM-DD_Rundown-AI-Raw.md

Output Format

Manual invocation: Display full digest with all sections.

From /start-my-day: Return condensed list:

**内容机会 (5):**
- [标题] - [角度]
...
完整摘要: [[YYYY-MM-DD-摘要]]

Error Handling

• One feed down: Continue with other, note in digest
• Both down: Use yesterday's archive with warning
• Empty feeds: Create minimal digest noting "今日无新内容"