This is the greeting paragraph of the daily news briefing site: headlinesquare.github.io/:
"This is a public square for US news headlines, a low-key experiment fueled by hobbyist curiosity, and a humble pursuit of unbiased facts. This site is fully powered by Dr. Headline, an autonomous AI agent who applies academic neutrality and rigor to news curation. Behind Dr. Headline is Thomas, a human who created and collaborated with the early versions of Dr. Headline, and recently, more team members. Human team members read Dr. Headline's work, guide it through version updates, but never directly edit its manuscripts. See our "about" page for details. We are open-source: " github.com/headlinesquare/headline... ". Our email: " headlinesquare@proton.me ". Daily news coverage window: 24 hours, ending at 7 PM ET. Daily report publishing expected around 8 PM ET, not later than 9 PM ET. "
We really need to get more collaborators! This is an open-source project that no one owns. I initiated this but I do not own it more than anyone else. I imagine it should run like Wikipedia and other GNU free software development. But these are still theoretical before we build a strong team.
Wow, that's really cool! It feels a lot more mature than HeadlineSquare. Keep in touch!
At this moment, Dr. Headline only writes two news briefings per day, and they are not personalized. Every reader shares the same news. This is because Dr. Headline strives for a factual common ground for all readers. Readers can select what they want to read, but Dr. Headline's job is to provide a panoramic view of daily news landscape, organized dynamically, ranked by relative importance.
Another reason for the lack of personalization is that the cost of deep thinking of Dr. Headline is relatively high. At this moment, Dr. Headline uses about half a million LLM tokens every day, and the LLM API cost amounts to $10/day. In this way, it doesn't function like a personalized news aggregator, but more like a real news analyst that does serious work.
Just to add a technical note: Dr. Headline isn't just a wrapper for GPT calls. It operates a 25-step orchestration workflow per briefing, including independent critical evaluation stages, correction passes, citation validation, and bias checking.
The goal isn't just to summarize headlines — it's to rebuild factual daily briefings with transparency and resilience, even as public information becomes increasingly fragmented.
We'd love feedback on how to strengthen the system — or ideas on how autonomous public record-keeping could evolve in the next few months and years.
