The flood
Since the launch of GPT-3, the volume & velocity of AI news have gone parabolic while the quality has gotten worse. Even when immersed in the space, trying to stay on top of relevant, useful news can be a Sisyphean act.
Every firm, company, or side project boasts about its new AI feature, which makes actual breakthroughs or useful products feel few & far between. Especially in the news, it’s currently only a handful of companies worth following: OpenAI, Anthropic, DeepSeek, DeepMind, & a few other open-source ones.
Even these firms still publish meaningless fluff for stockholders or attention, but it’s less than the full firehose of AI news from the entire news media. An annoying problem arises when you want to follow the top firms: they don’t all have a standardized way to publish their news.
The problem
I found this when I wanted a simple AI-news RSS feed. I looked for a feed from each top firm. OpenAI, DeepMind, & Hugging Face have one, but Anthropic, Meta AI, & DeepSeek don’t. I don’t know why; it’s easy to make, but these multi-billion-dollar firms can’t manage.
“The AI news landscape moves fast, but reliable updates are scattered. Building a unified feed is the only way I’ve been able to stay current without drowning in noise.”
This led me to make my own feed that has all the top firms’ news in one place. My plan: create a DB that will auto-populate with the latest news and allow me to search for them. First up, OpenAI & Hugging Face. They were easy to work with out of the box. DeepMind took some fighting, but after a while, it worked too.
At this point I have a DB with news from three of the top firms, but that’s not enough. Next up, Anthropic. This one surprised me. A company based around being the Coders AI, yet they don’t have an RSS feed. So, I went about making my own. Luckily, I’m not the first person to want to turn a blog into an RSS feed.
The final stretch
Full credit goes to Daniel Olshansky & his tool, RSS Feeds. This made it possible to take the Anthropic news blog & turn it into an unofficial RSS feed. Now with the feed, it gets plugged into a worker to feed the DB.
Now onto Meta AI. The AI is meh, but a great open-source tool. They didn’t have an RSS feed or a good blog to transform. So I went to the dark arts: web scraping. This was a pain in the ass. Not only are they super finicky, but they tend to trash data a lot of the time. After hours of pain, struggle, & swearing, it worked.
The final boss was DeepSeek. Compared to that, the problems we face now seem mild. DeepSeek doesn’t have an RSS feed, blog, news site, or anything. Nothing to scrape, nothing to collect. My solution, after a long, hard think: fuck it, remove them from the pool of feeds. This pained me; I like their models, their tech, and their open-source nature. Yet, they don’t have a place to get news directly from them.
The learnings
In the end now, what I learned.
- RSS feeds are awesome. I wish more companies would use them for news & info delivery. I have more respect for OpenAI, Hugging Face, & DeepMind for having their feeds.
- Creating your unofficial feed is easier than ever! As well as the tools to store and automate collection. Creating the backend for this project was slow due to my lack of knowledge of the tools. Shoutout to Supabase for being so beginner friendly.
- Web scraping sucks; it is annoying, cumbersome, & inaccurate.
If you want to check out the site to keep up to date on AI news, check out Ai-news **or the GitHub repo: https://github.com/JonathanRReed/ai-news.
2025 Update: Its been updated to be more useful. Now with a search bar, and the ability to filter by company.
Thanks for reading!
Feel free to share this blog or reach out to me on LinkedIn.