Rakesh Kumar Rout
@rakeshimmy
283 Following
47 Followers
9 replies
17 recasts
90 reactions
15 replies
6 recasts
81 reactions
2 replies
1 recast
4 reactions
5 replies
6 recasts
141 reactions
36 replies
33 recasts
417 reactions
honest personal take (disclaimer: i'm not an ML expert, i just study and use a lot of social apps):
i think its highly unimaginative and tired to tell new users on a social platform to "reply a lot" in order to get any engagement.
social platforms, for a long time, pinned discovery as a user problem. tiktok changed that. they didn't expect users to earn attention the hard way (hustling to self-promote and get follows).
instead tiktok took on discovery as a platform problem and decoupled distribution from follower count entirely: a content-first, graph-agnostic approach.
on tiktok, a user's responsibility is to make good content and the algo would do the rest based micro interactions and watch behavior. YouTube took a similar approach, but favored creators who mastered their (effort intensive) rules: SEO knowledge, thumbnails, content cadence, etc.
if we're hearing that discovery is a problem, the question should be: whose problem â the user's or the platform's?
take long-form text platform substack: the recommender engine places more of the burden on the platform than the user, and the platform bias is towards quality via a trust graph.
it recommends newsletters based on what you read, introduces editorial curation, and the "recommended by other writers" feature means that small / new writers can get regularly recommended to audiences that a specific writer's taste.
each platform makes a design choice about who should work to be seen. on substack, they believe that good writers should be lifted through network effects and curation.
does this scale easily? no, it scales slowly but is compounding.
so why is this so hard for short-form text-based platforms like twitter, threads, bluesky, or farcaster?
because short-form text is typically low-signal, high-noise. short posts on their own carry very little context or signal. it's easier to produce, but harder to evaluate. as a result, discovery for short-form text platform has relied heavily on follower graphs.
twitter initially solved this with hashtags (then failed to do it at scale with lists, fleets, and circle). substack partially solved this with its writer graph (trust graph) and categories (topic / semantic clustering).
and i think farcaster's open data, mini apps, and interoperability (zkTLS) can be leveraged for creative solutions here that aren't possible on other platforms. that excites me. 49 replies
40 recasts
442 reactions
151 replies
186 recasts
1037 reactions
48 replies
42 recasts
340 reactions
17 replies
8 recasts
70 reactions
40 replies
90 recasts
726 reactions
5 replies
13 recasts
86 reactions
0 reply
0 recast
0 reaction
TL;DR: The Spam Reaper is back with a vengeance aka 25% of my Good Followers got marked as Spam this week
⌠Follower numbers going crazy, down by a lot
⌠Rewards leaderboard scores are weird (probably points being removed because people are marked as spam)
⌠Overall engagement is down/strange
⌠Much confusion and people feel down on themselves and are taking it personally. Donât. Itâs not you.
ââThe Great Purge...Againââ
Your follower counts are going down because the spam algorithm is very aggressive this week. When someone gets labeled as spammy, aka a â0â label, the Farcaster client no longer counts them in your follower count. They are still following you and you can see it on the protocol.
ââ-Spam Label Official Updates Aren't Updatingââ-
The spam data set hasnât been updated in 11 days. We used to get an update every week on Thursday evenings. That cadence has been inconsistent since the beginning of May. Any frames that show your spam label are 11 days out of the date. The spam algorithm runs daily and people are re-labeled. But until the json file gets updated, we canât see your true status. We are just guessing. If you are in âshow more repliesâ inside of a channel, you might not have a spam label. The channel might be set to prioritize members replies. If you comment on a naked cast, aka one thatâs not in a channel and you are in âshow more repliesâ you probably have a spam label.
If you have a â0â spam label, you arenât eligible for the Rewards Leaderboard. Your engagement also doesnât count for others. So if your points have gone down, people either deleted their replies/undid a recast/like OR They got a spam label of "0" sometime this week and their engagement no longer counts so your points went down.
ââ-Engagement is Weirdââ-
Engagement is weird this week. The team is clearly tweaking the home feed algorithm and itâs starting to have effects. You canât fight the home feed, but you can change your default feed to âfollowingâ and use that more and you can pick a few of your favorite channels and scroll them and interact to train the home feed to show you some more interesting things.
With @tipn and @noiceapp in play, some folks may be changing how they engage and liking, recasting, and commenting less or being more careful with their engagement since money in now attached. Just a theory but this could be a factor in why engagement feels different.
I don't think it's one single thing, but a multitude of factors converging all at once.
ââ-Vibes are Down, Let's Fix Themââ-
Iâve seen many casts where folks are feeling down; feeling like they arenât doing well because the metrics they can see are down. Donât feel bad. Itâs not you. Itâs just how it is this week. You arenât being ignored or mass unfollowed. Itâs not personal, itâs just the algorithm. Donât be so hard on yourself. If you show up here often, engage with people authentically, and show us your whole self, it will work out. It might be slow. It might not be measurable today, but consistency is key.
Since engagement is down, go out of your way to engage this week.
â· Find one new person to follow
â· Scroll a few well curated channels on topics and you like and say hi to some new people
â· Respond to all authentic replies on your own casts
â· Tag a friend into a conversation they may not have seen
â· Use quote casts to elevate great content and share it with a new audience and add to the conversation, but be sure to give some engagement to the OC (original caster!)
â· Reply more if you have the time. You could make someoneâs whole week
ââ-Verifications are Comingââ-
Be sure to have at least $25 in major tokens in your Farcaster wallet and do all the verifications you are comfortable doing. I donât think the visibility changes have gone into place officially but they are coming.
To be more visible you will need:
1. A good spam label (Level 2)
2. Multiple verifications
3. Money in your wallet
4. Engagement*
*engagement goes both ways. Be generous with your likes, recasts, and comments and people will engage back. And if they don't, why are you following them on a social network? Your attention is valuable!
https://paragraph.com/@pichi/25percent-of-my-good-followers-are-gone 17 replies
19 recasts
80 reactions
50 replies
48 recasts
767 reactions
32 replies
10 recasts
161 reactions
14 replies
14 recasts
104 reactions
1 reply
1 recast
4 reactions
23 replies
18 recasts
295 reactions
45 replies
31 recasts
189 reactions
0 reply
0 recast
0 reaction
