Avatar
hodlbod
97c70a44366a6535c145b333f973ea86dfdc2d7a99da618c40c64705ad98e322
Christian Bitcoiner and developer of coracle.social. Learn more at info.coracle.social. If you can't tell the difference between me and a scammer, use a nostr client with web of trust support.

Yep, click on the sliders icon on a person's feed

Replying to Avatar hodlbod

I think we need different types of "follow". What about something like this:

1. Friend — this person doesn't show up in my feed. But I like them, and want to keep track of them, and maybe be reminded of them on occasion (hey hodlbod, here's what your friend X has been up to this summer). I trust them, and want to factor their opinions in to content recommendations generated for me.

2. Follow — I want to know what this person says if it has engagement, is popular, or matches some other filter, like topics I've expressed interest in. The Hacker News bot might fit in this category.

3. Super Follow — I want to see everything this person says. These are the people on my "pure signal" list currently.

4. Subscribe — these are people you don't care about, and whose opinions you don't care about per se, but are people who get paid (either by you or by advertisers) to recommend content or products. Could be bots or influencers. You'd never see their content, but you would see things recommended by them.

Items #2 and #3 are variants of the same thing, and could be conflated by assigning a decimal value to your follow (suggested by nostr:nprofile1qqsfcts2suzpxaeuhy2mnjwd9cwt69l98t3tp2r2hf09hu8uz0zzp5spzfmhxue69uhhqatjwpkx2urpvuhx2ucpz3mhxue69uhhyetvv9ujuerpd46hxtnfduq3vamnwvaz7tmjv4kxz7fwdehhxarj9e3xzmnyuurtjm earlier, but I used to have something like this in Coracle).

The first category is entirely different though, because the value of the relationship is not based on what they say, but who they are. I honestly don't want to see anything my mom posts to social media unless she tags me in it (in which case she'll email me). But she's one of the most important people in my life.

Likewise the final category. This is an entirely transactional-type relationship, and is exploited to provide additional social signal to otherwise neutral content.

So, any other categories? This is really quite similar to nostr:nprofile1qqsyvrp9u6p0mfur9dfdru3d853tx9mdjuhkphxuxgfwmryja7zsvhqpr4mhxue69uhkummnw3ez6ur4vgh8wetvd3hhyer9wghxuet5qyt8wumn8ghj7anfw3hhytnwdaehgu339e3k7mgpzpmhxue69uhkummnw3ezuamfdejsxp7af9 's "relationship status" nip, but with more ability to quantify what clients can do based on the relationship. It could also be implemented (of course) using NIP 32.

I don't think so. A fourth positional argument to `p` tags seems like a bad idea. I don't think lists are a good data type anyway, prone as they are to race conditions.

No, I unwrap. Unless we're talking about an authenticated account switcher, filtering is just a matter of showing the right information in the UI

I don't think category #1 is exclusive to family, it could also apply to anyone you want to remember but is prone to fall off your contact list, like friends from high school, favorite authors (so you know when they publish a new book), local businesses, etc. Basically stuff where you want several months of activity artificially compressed into a digest that you can spend 5-15 minutes on once per quarter to keep up, rather than 1 minute every day (a total of 90 minutes per quarter).

I think we need different types of "follow". What about something like this:

1. Friend — this person doesn't show up in my feed. But I like them, and want to keep track of them, and maybe be reminded of them on occasion (hey hodlbod, here's what your friend X has been up to this summer). I trust them, and want to factor their opinions in to content recommendations generated for me.

2. Follow — I want to know what this person says if it has engagement, is popular, or matches some other filter, like topics I've expressed interest in. The Hacker News bot might fit in this category.

3. Super Follow — I want to see everything this person says. These are the people on my "pure signal" list currently.

4. Subscribe — these are people you don't care about, and whose opinions you don't care about per se, but are people who get paid (either by you or by advertisers) to recommend content or products. Could be bots or influencers. You'd never see their content, but you would see things recommended by them.

Items #2 and #3 are variants of the same thing, and could be conflated by assigning a decimal value to your follow (suggested by nostr:nprofile1qqsfcts2suzpxaeuhy2mnjwd9cwt69l98t3tp2r2hf09hu8uz0zzp5spzfmhxue69uhhqatjwpkx2urpvuhx2ucpz3mhxue69uhhyetvv9ujuerpd46hxtnfduq3vamnwvaz7tmjv4kxz7fwdehhxarj9e3xzmnyuurtjm earlier, but I used to have something like this in Coracle).

The first category is entirely different though, because the value of the relationship is not based on what they say, but who they are. I honestly don't want to see anything my mom posts to social media unless she tags me in it (in which case she'll email me). But she's one of the most important people in my life.

Likewise the final category. This is an entirely transactional-type relationship, and is exploited to provide additional social signal to otherwise neutral content.

So, any other categories? This is really quite similar to nostr:nprofile1qqsyvrp9u6p0mfur9dfdru3d853tx9mdjuhkphxuxgfwmryja7zsvhqpr4mhxue69uhkummnw3ez6ur4vgh8wetvd3hhyer9wghxuet5qyt8wumn8ghj7anfw3hhytnwdaehgu339e3k7mgpzpmhxue69uhkummnw3ezuamfdejsxp7af9 's "relationship status" nip, but with more ability to quantify what clients can do based on the relationship. It could also be implemented (of course) using NIP 32.

Or suffix? Seems like a 10 byte prefix would have more collisions because of PoW

Here's a quick rundown of the new content filtering functionality in Coracle, including content warnings and word muting.

https://media.nostr.build/av/d43449f0587e2c697fe8f13633f1d1339a57b8154d7ac573841098ec873feed2.webm

I tried adjusting the error rate to 1/10k, that should help, but I think nostr:nprofile1qqs99d9qw67th0wr5xh05de4s9k0wjvnkxudkgptq8yg83vtulad30gpp4mhxue69uhkummn9ekx7mqpz3mhxue69uhhyetvv9ujuerpd46hxtnfduq36amnwvaz7tmwdaehgu3dwp6kytnhv4kxcmmjv3jhytnwv46qq5gspu is right that we should have a pretty low tolerance for false positives for deleted content

I just didn't expect false positives after ~200 items

NIP-32 for "providing context" ala community notes

That's true, I'm only storing deletes by the current user right now though, so a set isn't too heavy. I'll probably end up revisiting at some point

Just tried using bloom filters to store deletes, but unfortunately, false positives seem to be pretty frequent. Guess I'll have to store a set of ids.