It's not so much an algo as a curator that has a set of parameters and can judge an author/content on multitudes of parameters we've set out for it. Which of course can be changed and adjusted and we'd keep an eye on it and help it improve or downvote if it does bad judgements, it could even learn from our downvotes if the hidden from public but clear parameters we've given it aren't clear enough.

Like, and this doesn't have to be part of them, but imagine it can look into so many things to go into how much it'd vote the content for that manual curators couldn't. Here's some examples off the top of my head if say we had no data limitations or api restrictions on the agent.

How often does the author post
How good are the posts in terms of readability/seo/grammar/uniqueness/originality
How long are the posts
How many users replied to this author, how many of them are unique
How much engagement occurred, how much back and forth, did engagement bring in new commenters to participate
How long has the author been on hive
How many posts and comments has the author generated and are they mostly unique or classified as a bot/service/spammer
How much stake follows the author, how much stake follows the people engaging with the author thus consuming the post
How much stake has the author kept and does he use it well - could even look at the authors curation
How much stake follows the people who reposted this
How many new users are replying to this author, how many of them are new to reply to this author after having followed the author recently
is it simple content or an announcement
is it repetitive content

etc, etc.

a lot of these factors as you can imagine are hard to fake and again, depending on limitations both AI usage, API requests and how much voting power it ends up receiving, it'd have to stabilise and find a middle ground.

Will it be perfect? No, but it'll definitely even as MVP do way better than all autovoters/blind voters/vote sellers and compete with manual curators who hopefully try make this place better and if anything reward more deserving users.

It could do that at scale if there aren't enough manual curators and it could do it at cheap vs rewarding manual curators. Not saying we should punish and remove workload from manual curators, but instead grow the stake that's voting as good as them.

If people then want to use the bot as an algo it could also start following authors it prefers and curates so its following feed could be browsed by people.