It's easy:
Users should be able to construct their own feed and their own trending feed.
It must be like a buffet, with parts and pieces ready to pick and put together.
Each user selects a set of features that shapes a customized algorithm.
The platform can them suggest detect and suggest popular algos based on the frequency of appearance in users choise.
Interesting - but what about non signed in people? Is there a "default" algorithm? If there was an objective way to identify the most popular algo that could work, but in some sense that's just recapitulating the question of how to objectively identify popular content in n+1 dimension
It's not the popular algorithm that we look for, it's the ingredients and styles, which later form a menu, then the platform offers those menus.
Example of the ingredients:
most reaction, most divers reaction, most total sats zapped, most number of zaps, any number of tags, any number of mentions, specific subjects, most quoted, most boosted, and the list goes on...
(Could be the least in every case)
I as L put some of those together with a personal priority and preferance.
Then the client can offer a random or a popular algo based on people's choice.
It can also offer the menu right away and say help yourself, pick what you like.
