Kasra Rahjerdi
@jc4p
I got nerd sniped by the new Trending Topics and wanted to see if I could do it in a better way. Checkout the approaches and results and let me know your fav: https://kasra.blog/2025/04/16/farcaster-trending-topics-analysis/?1
8 replies
15 recasts
62 reactions
Kasra Rahjerdi
@jc4p
Of course, all code open source + the data is from the public HF drops: https://github.com/jc4p/fc-trending-topics
0 reply
0 recast
7 reactions
horsefacts
@horsefacts.eth
big mood
1 reply
0 recast
4 reactions
shoni.eth
@alexpaden
i think the only “serious” improvement is efficiency around personalization nonetheless nice job / cool share
1 reply
0 recast
2 reactions
Jordan
@ruminations
I liked the first and last sets of topics the most. What I'd personally like is the ability to navigate topics through hiearchy - i.e. start at "Miniapps" and then get granular into "Farville..." I think the girvan-newman network algo can take initial topics and then generate those sub-topics without having to run each cast through an LLM multiple times. https://en.wikipedia.org/wiki/Girvan%E2%80%93Newman_algorithm
1 reply
0 recast
2 reactions
Steve
@stevedylandev.eth
Lowkey think your approach finds what’s actually trending on the network. Have a hunch that merkle’s approach is more targeted towards raising DAU vs accuracy
1 reply
0 recast
1 reaction
Koolkheart
@koolkheart.eth
This is awesome. Love seeing folks iterate on discovery UX… what were your biggest surprises when tweaking the logic?
1 reply
0 recast
1 reaction
ted (not lasso)
@ted
cc @ruminations you'll find this interesting
1 reply
1 recast
4 reactions
Zinger ↑
@zinger
None of them really resonated with me but I did find it funny that the embeddings one was a proxy for the overall vibe of Farcaster and thought that generator-critic was an interesting approach Also love the unique share CTA you added to the bottom of your blog!
0 reply
0 recast
1 reaction
