TL;DR: I created a public Telegram channel that shares news and posts about Data Science, Machine Learning, and AI that I find interesting. You can preview content here and join the Telegram channel here.
Problem
Keeping up with the tech news, whether it’s a new development framework, new ML model, or just a hands-on tutorial, has been bothering me for quite a long time. There are so many information sources: blogs, Twitter/X, social networks, arXiv papers, podcasts, and many others, all contributing to information overload powering FOMO. Even newsletters, which are meant to aggregate information, are just another source of information overload and clutter in your inbox. Search for “python newsletter”, and you’ll find a dozen of them.
Tracking all these sources takes a lot of brainpower and time, with even more time wasted due to ads and duplicate content circulating across different channels.
Solution
To manage my information intake better, I started a small pet project almost a decade ago. I enjoy building pet projects to experiment with new technologies — quickly hack a prototype in a few days, learn something new, and move on. I rarely make those projects public.
But this project survived many iterations. It started as a bunch of scripts to grab RSS feeds and scrape web pages. And over the years, this project has become a playground for different technologies: web-scraping with headless browsers; experiments with Elasticsearch as a main database for link collection, aggregation and search; classic NLP/ML to build taggers and classification models to predict what I’d be interested in (the idea was that I’d only get news and articles from my areas of interest); integrations with various APIs (some don’t even exist anymore, like OpenCalais, if anyone remembers it). And it still is a playground, for example using LLMs to extract data from HTML, or one of the recent experiments was playing around with ModernBert for classification (maybe that deserves a dedicated post too).
LLMs, AI Agents and llama.cpp have changed the project a lot, making everything way simpler, no need for custom parsing rules anymore, even small LLMs can extract data like a charm.
Channel
Like many pet projects, it wasn’t supposed to be published. But a few people knew about that project, and asked me for access to the aggregated news stream — one of the byproducts of the project. So I created a Telegram channel to publish that stream.
Why Telegram? I wanted a solution that I’d actually use myself, that’s why I didn’t want to create “another” newsletter that clutters an inbox. Posting to Twitter/X was not an option either, I stopped using twitter actively when they blocked alternative clients. But people use messaging apps pretty often, and between WhatsApp and Telegram, I personally prefer Telegram, I do not quite like channels in WhatsApp.
After testing this approach with my friends for almost a year, I settled on publishing 5-10 short news posts (title, short description, link, and tags) as separate messages once a day. And after some encouragement from my friends, I finally decided to make this channel public. So, please, check out The Boring AI News.
Why Boring? Keep in mind that the content reflects my personal interests, I built this project to filter posts relevant to me. So the content is mostly AI/ML/DS oriented with the focus on practical applications, hands-on experience and fundamental research, and less about business news, so you won’t find news about startups acquisitions, money raising, etc.
Personally I use this channel as a news feed, I keep notifications muted and scroll through it whenever I have time. If you don’t use Telegram or just want a sneak peek, I republish the last three updates here in my blog.
If you have any feedback, positive or negative, you liked it or have a suggestion for improvement, feel free to get in touch with me.