Feeds Hub: Turning Information Streams Into Executable AI Automation

I built Feeds Hub because the information I track every day is scattered across too many places: AI, financial markets, open source, Rust, startups, product design, and global news.

Reading news is not the hard part. Turning scattered updates into personal judgment is the hard part.

Feeds Hub is not meant to be a fixed content site. It is a repeatable information-processing workflow: search, filtering, source judgment, summaries, Markdown writing, cover generation, and static publishing.

Project repo: idaibin/feeds-hub

Thinking Process

At first, I thought the main problem was daily content updates. Later I realized the real problem was editorial judgment.

A useful feed automation run has to answer several questions:

• Is there anything worth writing about today?
• Is the source reliable enough?
• Is the fact fresh enough?
• Is this a real update or repeated noise?
• Will the entry still be readable later?
• Is the cover image clear enough for card-based display?

That is why Feeds Hub does not focus on RSS ingestion alone. It focuses on rules an AI agent must read before execution: topic rules, source rules, content format, poster prompts, quality checks, and repository boundaries.

Execution Model

Feeds Hub separates repository ownership clearly:

• idaibin/skills stores independently installable specialist capabilities
• feeds-hub stores the feed site, topic rules, source rules, and generated entries
• blog stores public Prompts, long-form reflections, and project memory

This keeps short-cycle feed entries away from reusable AI assets, and keeps long-form reflection out of the feed repository.

A normal update behaves like a small editorial pipeline:

1. Read automation rules
2. Search recent information by topic
3. Judge whether each item is worth writing based on source level and freshness
4. Skip topics without enough signal
5. Write valuable items as structured Markdown
6. Generate or update the cover image
7. Run content checks and build checks
8. Commit to the selected content or production branch

Result

The most useful result is not the page itself. Feeds Hub proved a practical point:

AI can do more than write articles. It can run part of an editorial workflow.

It turns “reading a lot of updates” into “saving a small number of high-signal entries.”

For me, it produced three concrete outcomes:

• It proved that ChatGPT scheduled tasks can operate a real content repository
• It forced the source, topic, article-format, and poster rules to become explicit documents
• It lets the blog focus on long-term reflection instead of short-cycle feed entries

Feeds Hub is not the end state. It is a reference pattern for any topic that needs repeated tracking.