Pretty cool! How do you do to build these "stories" based on news?
replies(1):
A lightweight a page design as I've been able to keep; simple, clean, fast. No commercial features or aspirations - this is a passion project, something I've been fooling around with on and off for decades.
There's a "Top" view too with an LLM edited front page & summary, and categorized views for a large number of topics - see the Directory. A few more buried features to explore, but the fundamental use case is pop in, scan, exit - fast and concise.
Your feedback would be appreciated!
Thanks so much for the kind words - its 100% o3-mini for clustering. I have zero editorial input as to what constitutes a cluster, what's "top" news, etc.
The one subtlety is setting up the LLM to understand whether a new story belongs in an existing cluster, or with > 1 neighbors, constitutes a new cluster. The challenge there is scoping the clustering window (hours of stories for consideration) and topic breadth to avoid creating Katamari-super-clusters that just end up with every story associated to them.
At this point I seem to have found a sweet spot re: the hours window, the frequency of processing, and the design of the prompt such that its working consistently.
Very few false positives in terms of spurious clusters being created, or potential clusters being missed.
In my case we went with embeddings and clustering to find close papers to each other because llm were allucinating.