đ by @davehappyminion
Curate Reading List
Gather the best articles from across the web, filter out noise, and deliver a focused reading list.
1. Load User Interests
Check the user's config or memory for topic preferences (e.g., "AI/ML", "systems programming", "startup strategy"), blocked domains, and past reading lists. If no interests are set, ask for 3-5 topics before proceeding.
2. Gather Sources
Pull recent content from:
- Hacker News â Top 30 stories and Show HN posts
- Reddit â Top posts from subreddits matching user interests
- RSS feeds â If configured, fetch latest entries
- Newsletters â Check for forwarded newsletter content in inbox or saved directory
Capture: title, URL, source, brief description, and approximate reading time.
3. Filter and Deduplicate
- Remove articles the user has already seen (check previous lists)
- Deduplicate by URL and content similarity (same story from multiple sources â keep the best version)
- Filter out paywalled content unless the user has a subscription
- Remove clickbait and low-substance content
4. Rank and Select
Rank by relevance to stated interests, source quality, and freshness. Select the top 5-10 articles.
5. Present Reading List
For each selected article, provide:
- Title, source, estimated reading time
- 2-3 sentence summary of the key takeaway
- Why it was selected (which interest it matches)
Include a few honorable mentions that did not make the top picks.
6. Save and Archive
- Save the reading list to the user's configured output location
- Update the "already seen" list to avoid future duplicates
- Offer to send articles to a read-later service (Pocket, Instapaper) if configured
Notes
- Runs daily at 8am but can also be triggered on-demand.
- Never auto-subscribe the user to anything. Read-only gathering.
- If a source is unreachable, skip it and note the failure.
- Five great articles beat ten mediocre ones. Quality over quantity.
