That's a fun visualization and interesting to see. The bottleneck here is probably that it looks like the channels need to be submitted manually, but optimally that gets resolved over time.
The larger challenge might be that given your vision this should help with choosing what to watch. But seeing this, it might not give me enough to actually make the choice. I do recognize some of the channels there and I like that clicking on the ones I know I see connected ones I don't recognize – so those might be ones to check out. But how and why are they connected? Are the channels covering the same topics, or are they sharing a lot of the same viewers?
I would also like the text to be a bit more readable on when clicking on a channel. The contrast is such that there's a bit of friction to read it, and the fact it appears in the top right corner of my screen makes me move my eyes back and forth constantly. Maybe the information could pop up closer to the bubble (but I understand it might then overlap the links).
Just some thoughts! I did find some channels I might want to check out!
Also, sometimes the connections extend far away from the channel I click on and it's hard to find what is it connected to – but I find that actually really interesting to see what the channel is that's in a different domain. Maybe the side panel on the right could list out the connections so it would be easy to find?
Connections are done this way by purpose, to see, why AI use this subniche for another cluster and is it right or not. This is part of experiment.
Thanks for output! I missed that right side popup issue, because i work mostly on macbook 13', so it was closed for me!
Manual input was by purpose, i want users to feel mlike a part of social concept.
And if that is unclear, those lines connect nearby channels by subniche(you can click each subniche which will show you connections and list of related channel, which share same subniche). Problem is, that i can get limited data from YT on bigger scale, and because im limited to this data, AI got some issues to recognise subniche pretty acurately. Im thinking how to resolve it in clever way then to burn a lot of openAi tokens. Thanks for info, that's very important for me, because its my first standalone project ever! You cannot imagine how it turned my day to bright side!