It’s great. Having an algo store anyone can contribute to and consume from is the best solution to my mind. Maximize choice and market dynamic.
Discussion
Solid idea.
Pimp My Ride. Customize your algo experience catered to your demand.
Bring-your-own-algorithm might appeal to normal people and big tech will not be able to offer it substantively.
#[2]
I love the chronological timeline. But one thing I think can be improved upon is boosts.
Boosts tend to happen right after someone posts an interesting note. This limits the reach to those who are on Nostr at that moment. If I boost a note by Jack right after he posts, it has no benefit because everyone is following Jack anyway.
But what if we could schedule a delayed boost? Similar to how I can adjust a ⚡️ amount or anonymity. Hold the 🔄 button and boost in an hour, or at a certain time. This would help filter good content to reappear in people’s feed no matter when they log on.
I had a drama-free TL on Twitter. I've never blocked anyone for their post/ muted keywords (except one).
To me, it had (has?)
1. The best discovery mechanism for events, people, art & literature, random but cool stuff like pudding.cool , Longform's Best of lists
2. Voice of people from different countries. I loved people posting about their daily meals/commute/weather. I could appreciate the everywhere sameness of the banality of life
3. Multiplicity of views, with nuanced discussions if one has the patience
I liked it as it felt like a treasure hunt😅. The algo still works but having lost a bit of its quirkiness 🤷♀️ (plus ads on reply threads 🙈).
---End of missing Twitter rand---
#[0]
Oh oh I know!!! We should be able to set a time limit like on zaps of WHEN we want to boost a post!!!
Boost post in 5min, 10min, 30 min, 1 hr, 1 day.
We can do better: There is an ML framework called federated inference and learning that allows traning algorithms on client side, without data sharing. One then exchanges learned model parameters with other clients using the same algo, so that everyone collectively is improving their local models. One can set up a market for parameter exchange between clients. This can work for any ML/AI application.
Exactly, more of a Vim/Neovim/Emacs model where the client can be the base layer that does all the low level stuff and provides powerful APIs for scripting so devs could create plugins for sorting posts without having to implement a new client or convince the core devs that their idea would be a useful feature for a majority of users.