Nice work. I wonder if there may be a better application for the Vectara capabilities than search?
Algolia has already done the search thing, can the Vectara search be 10x better?
What I do find missing from HN is the ability for me to see things that may be of interest to me, but that I may have missed. I like how I get everything in the main feed which is pure popularity, but I don't have the time to go through all posts, and definitely likely miss things I would probably have been interested in.
Though this can be done with collaborative filtering, or other non-AI methods, might this be a decent use case for your AI?
This may be a good use case for RSS. Feed readers can filter posts by keyword so you can take the unofficial RSS feed [1] and filter it down by your interests.
I posted an RSS reader that can do this recently [2] and I'm actively hacking on another [3]. But there's many RSS tools that can do this.
Unclear that the HN feed is pure popularity. There's got to be something else to it [1], otherwise it would look like all the medium-like crap you'll find elsewhere.
[1] my hunch is that some human expert curation is involved.
It is not pure popularity, I suspect it is a combination of clicks/upvotes/comments over time. I know time is a key component in the algorithm. If you don't rise fast, you don't rise. Who is commenting/upvoting is also likely an easy metric to add in. If I have more points, my votes are probably worth more than someone without.
Human curation also exists, but I think that is aimed at removing spam and uplifting YC company posts.
I've been wondering if segmentation might be the way to go. Have the chatbot look at all the items, cluster them into a few buckets of its choosing, then throw each new item into the most appropriate bucket.
(I've been thinking about this not just in terms of HN, but treating all my RSS feeds as one undifferentiated stream and just having a chatbot sort incoming items into whatever bucket it deems most appropriate).
What's stopping me is that it might work, and I doubt making the internet even stickier is good for me long term.
I've been thinking about that at length recently. RSS feeds often don't have category elements specified and there isn't a widely used taxonomy of category names. I'd prefer not to use AI to solve the problem, although encouraging the use of RSS categories will be slow work.
Thanks for the link. It's interesting, and I hope you find a way forward with that, it would undoubtedly be a useful addition to the ecosystem.
But my gut feeling is that there's not enough interest in RSS right now to drive widespread adoption of a new version of the spec. My approach would be to focus on improved UX over existing feeds, rather than speculatively expanding the spec to make feeds richer.
The main advantage of my approach, I think, is that it adapts to the individual end user's needs. If all my subscribed feeds are tech-focused and I use a generic published taxonomy, I'm going to end up with 60% of my items in "Technology" and 30% in "Computing". If I use a chatbot to dynamically bucket stuff, I'll get "Micro PCs", "Graph theory", "Golang", etc etc.
Algolia has already done the search thing, can the Vectara search be 10x better?
What I do find missing from HN is the ability for me to see things that may be of interest to me, but that I may have missed. I like how I get everything in the main feed which is pure popularity, but I don't have the time to go through all posts, and definitely likely miss things I would probably have been interested in.
Though this can be done with collaborative filtering, or other non-AI methods, might this be a decent use case for your AI?