Why Nostr? What is Njump?
2024-08-04 10:36:18
in reply to

liminal on Nostr: Maybe a stretch, but here's a wall of an idea: (1) Let's move from relay pools as the ...

Maybe a stretch, but here's a wall of an idea:
(1) Let's move from relay pools as the town square, where everyone sharing an intersection of relays shares the same town square and move to relays as islands, closed off rooms etc. Similar to Reddit, but a theme is assigned to the relay, curated by the moderator(s) and community. More closed than what Satelite.earth has.
(2) find better ways to aggregate notes and surface older content - diminishing the recency bias.

Why (1)? With the wide perspectives we see from everyone, groups are not able to maintain a stable set of values. A permissionless wiki or community is not going to bring in specialized groups for fear of vandalism or being lost in the sea of other unrelated notes. Of course you have relay selection, but this is an afterthought in the town square model.

How is a segregated nostr going to help with ingroup behavior? It doesn't explicitly, but it creates an environment at the base level that disincentivizes negative & reactionary behaviors between tribes because now you can choose to not have in your face if you don't care to see it. Set it in your community rules.

Also, the real interesting stuff happens at the boundaries between communities - where relays don't need to be so strict with a specific theme. Where you can share relevant content across communities.

Curation at the level of the user, dictated by what room they choose to be in at any given moment.

(2) Now nostr is more segregated, what about visibility? How can we surface old content?

Searching through tags and labels work, but in a very limited way because you want tags to be specific. If you don't have a defined term ontology for whatever topic it is care about, you're stuck with a resolution problem. You need to balance between trying to find the most relevent tags for a topic and throwing tangentially related tags to increase reach.

Current functionality to share content is boosting, forking articles/notes and quote posting. Quoting, while used to change the context of a conversation or to aggregate multiple posts comes with its own set of problems. Similar to citations in papers, you're stuck temporally. Seeing the top level post means you see the new context, but you're going to have a problem finding where others have taken the conversation if you only see the bottom level. I've also seen quoting as a method for grouping notes together, but argue that its pretty ineffective for the previous reason.

What makes this worse is that many clients are __feed based__. Meaning unless you see it in your feed, you're not going to find it without some text search, which is very limited if you're trying to explore a general topic.

Two problems now, which many if not all clients face: Search resolution in text, tags/labels, and a temporal bias toward recency - incentivizing you to join in on the current conversation.

Here are two specs.
a) Modular articles, aka note collections, note collages, aka renderable lists with context:
Originally developed to help search and segregate focused context within a larger article, modular articles group events together as its core functionality.

Take a set of related events, group them together, and now they're all searchable under a single context.

"Opinions on integrating alt coins for tips"
"Vegetarian recipes"
"Cool notes and events that happened this week on nostr"

https://wikifreedia.xyz/nkbip-01/npub1m4ny6hjqzepn4rxknuq94c2gpqzr29ufkkw7ttcxyak7v43n6vvsajc2jl

https://highlighter.com/laeserin/1719204947236

b) Embedded content:
Without going into detail about the how, embedding models are a computational approximation of __meaning__ where you can compute distances between concepts. You want to find content that wasn't labeled or grouped and you don't know the exact words being said? Your best chance is through embeddings. Expensive, but effective and ideally used for content to be "caught" later like knowledge content or relevant notes. It makes little sense to embed everything someone shouts out their window.

https://wikifreedia.xyz/nkbip-02/npub1m3xdppkd0njmrqe2ma8a6ys39zvgp5k8u22mev8xsnqp4nh80srqhqa5sf


Relays-as-islands helps users stay in their echo chamber if they'd like: Good for community security, but individuals pick the communities they want to participate in. Broadcasting, boosting and quoting now functions as a cross pollination mechanism, rather than a "hey followers check this out"

Note collections help aggregate and surface old content as prescribed by the user given some context. Embedding models help for search resolution, with the optional capacity to throw in for a recommendation algorithm.
Author Public Key
npub1m3xdppkd0njmrqe2ma8a6ys39zvgp5k8u22mev8xsnqp4nh80srqhqa5sf