AI news moves quickly. A model release, a research paper, a useful open-source tool, and a thoughtful critique can all appear in the same day. I want to notice the important changes without turning every spare minute into a search for updates. RSS gives me a way to choose my sources and read them in one place.

Why RSS works for me

RSS is a simple subscription system: when a site publishes something new, it appears in my reader. I choose the publishers and writers; the reader collects their updates. That makes it easier to follow original work instead of relying on whatever a social feed happens to show me.

The hard part is finding good feeds and keeping the list manageable. That is where Tidings RSS helps. Its public directory groups feeds into downloadable OPML files, a format that many RSS readers can import. The collection covers AI, research, engineering, news, blogs, podcasts, and more. I use it as a starting map, then keep only the sources I actually want to read.

Build a small, varied reading shelf

The repository's AI bundle is broad, so I treat it as a menu rather than a checklist. I want a mix of sources that answer different questions:

  • First-party updates: OpenAI News, Google DeepMind News, Anthropic News, and the Hugging Face Blog help me see what teams actually released.
  • Research: arXiv feeds and lab publications point me toward papers and technical details behind the headlines.
  • Independent interpretation: writers such as Simon Willison help connect releases to what developers can build and test.
  • Another language and viewpoint: the catalog includes Chinese-language AI sources, which broadens the conversation beyond English-language coverage.

That mix matters more to me than the number of subscriptions. A press release tells me what a company claims. A paper shows the method. An independent writer may reveal a practical limitation or an unexpected use. Reading across those layers gives me a more useful picture.

How I keep the stream useful

  1. Import, then prune. I start with the AI bundle rather than the complete directory. If I want a wider view of technology and culture, the Top 200 bundle is another starting point. The bundles overlap, so I avoid importing everything at once.
  2. Scan headlines, save selectively. A headline is a signal, not an assignment. I open the original article when the topic connects to my work or challenges something I thought I knew.
  3. Go deeper at the source. For a major announcement, I look for the paper, documentation, model card, or code before forming an opinion. Summaries are useful for triage; they are not a substitute for the underlying work.
  4. Review the feed list. If a source is repetitive, noisy, or no longer relevant, I remove it. A small reading shelf I trust is more valuable than an unread count I cannot finish.

Try it yourself

Choose an RSS reader that can import OPML. Download the AI collection, use the reader's “Import OPML” option, and let the first refresh finish. You can also browse the source directory and subscribe to individual feeds. The project's RSS guide has reader links and troubleshooting steps.

AI will keep moving. My goal is not to read every update. It is to build a reliable path from an interesting signal to the original work—and leave enough quiet space to think about what it means.