@hipsterelectron @aud @ireneista @impactology for example, i still haven't had time to get a MWE together even if all the pieces are in place, but it's simply not true that activitypub can't do custom feeds. the protocol explicitly gives a key streams as "a list of supplementary Collections which may be of interest," and it is trivially implemented in mastodon by modifying the list system to allow lists to contain a list of posts, rather than a list of accounts. one could go further by allowing people to make those feeds public and followable, again both implicitly via mastohacking or explicitly with the protocol since they are Collections like any other.

That would mean that you could have algorithmic feeds computed by a single skiddie with a python script on their laptop that gets run when they feel like it all the way through the same kind of big-world feeds bluesky imagines. your instance is the thing that shoulders all the expensive parts of serving the feed and federation as fedi works now, rather than the atproto model where to have a feed means you need to run a pretty beefy, always-on server. I was working on a json-able DAG system of features and transformations so that if someone had a feed design that they liked, you could just copy it, tweak it, make it your own, and so on - and eventually exposing that via mastodon itself so that you don't even need the additional other computer, your instance handles the whole thing for you.

So algorithmic feeds are things made by people that respect current norms about post access by going through instances with your account as the agent and hosted by the instances we (hopefully) help sustain either with donations or time, rather than products that we are lucky enough to rent.

#ATProto #DIYAlgorithms

@roaldarboel
@DrYohanJohn @neuralreckoning @WorldImagining @elduvelle @NicoleCRust @jonny @PessoaBrain @brembs

beep boop #DIYAlgorithms
https://wiki.jon-e.net/Masto_Forks/DIY_Algorithms

@mala @algorithms @davidslifka

@mala Hi! Thanks for doing this. Here are a hashtag and some links that may be relevant, in case you haven't seen them:

#DIYalgorithms

https://github.com/mitotic/pinafore-mahoot/blob/master/README.md

https://github.com/mauforonda/mastodon_digest

https://octodon.social/@ethanz/109383923163675878

@lili
I agree, but I also think there's a balance to be struck with opinionated design, with masto itself being a decent case study. like inevitably someone will want to make a feed that sorts people they interact with more frequently higher on their feed, which seems very normal to me.

so what might other ways of making that healthier look like? one way might be to add some explicit noise, weighting, or damping parameter that prevents that from being the most dominant feature. if that parameter is under their control, they can decrease it until it isn't a harmful feedback loop. to help with that, another idea might be to offer a toggle switch to generate an "inverse feed" across the whole algorithm or a single parameter on it - what is getting filtered out of my feed as a result of how it's designed? that would make it easier to identify things that might be driving them into some unhelpful bubble or etc.

just thoughts, I think the whole question is v far from simple so thinking about what a healthy social internet looks like is never final or complete for me

(forgive me for tagging and making public, this is largely a way of taking notes to self through another imperfect UI lol #DIYAlgorithms )