Content
@
https://warpcast.com/~/channel/fc-updates
0 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
Introducing Trending Topics Algorithmic collections of casts about topics people are talking about. They update a few times per day. Mobile only to start. Update to the latest version of Warpcast from the your app store and restart your app 1-2x. Topics will appear on the search tab and in your feed while you're scrolling. We’re running this as an experiment in Warpcast for the next couple months. --- > Are these channels? Trending topics **are not** channels. They are collections casts about a popular topic identified by an AI-powered algorithm. Channels have no bearing on trending topics. > Are the topics on the protocol? The casts themselves are on the protocol but the categorization and feeds are generated by Warpcast. There are developer tools like @mbd that can provide this for your app.
49 replies
155 recasts
728 reactions
Michael
@michael
Am also curious how you decided to implement this
1 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
how? as in technical implementation?
2 replies
0 recast
1 reaction
shoni.eth
@alexpaden
yeah would be helpful to get a surface level explanation, i meant to ask in the dev call but didn’t. was just wondering if you’re even using embeddings or just putting all casts into a small llm with a prompt to classify them in a couple words then grouping by count or something else outside the box
1 reply
0 recast
1 reaction
Dan Romero
@dwr.eth
> just putting all casts into a small llm with a prompt to classify them in a couple words then grouping by count More or less this. Ask LLM to interpret and come up with best topics for each cast evaluated. Then aggregate.
2 replies
0 recast
1 reaction
Michael
@michael
No vector embeddings or anything, just raw context window? interesting that it works so well, cheers
2 replies
0 recast
0 reaction
shoni.eth
@alexpaden
will you share what param model is used?
0 reply
0 recast
0 reaction
