Dan Romero
@dwr.eth
Virtually all of the people on Farcaster with large audiences: 1. Don't have large audiences elsewhere (Farcaster is their first time with a large audience) 2. Didn't know me or Varun and didn't know each other. The "Farcaster OG scene" is net new creation of Farcaster. 3. Were early and stuck around when everyone made fun of Farcaster (and people still do) 4. Continued to cast interesting stuff, casted in channels when we suggested people cast in channels 5. Warpcast onboarding asks all users for their interests 6. Based on the interests, we pair the account with an initial set of follows 7. There are 200+ accounts on the list, weighted toward accounts that get the most engagement / drive retention, but a large group of people benefit 8. On our list to improve, but the reality is outside of "I should be on the list!" haven't seen any concrete proposals to improve
50 replies
10 recasts
247 reactions
David Furlong
@df
Re 8. Here are some suggestions that will drive faster feedback loops and likely behavioural change A) if you haven't casted N times in the last week, you don't get auto followers B) Weight it based on (power badge weighted) engagement rate in last week - maybe to 50% C) Expand beyond 200 casters to include algorithmically any caster with a power badge
1 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
A) If Vitalik doesn't cast but is generally considered one of the most interesting accounts on the platform, why hurt the new user with an arbitrary user B) That's effectively where we are now? C) Again, give me a list of people that should be on the list. You're assumption is that there is an equally interesting set of users to pick from.
2 replies
0 recast
2 reactions
0xqeew
@0xqeew
can you provide a reference to the list? would be interesting to look it up
2 replies
0 recast
0 reaction
David Furlong
@df
A) I agree for Vitalik, but there aren't 200 Vitaliks - he's the exception not the rule. B) I think the feedback loop is still too slow, many inactive, non famous casters getting 10k+ followers a week, happy to send. I think it's worth an AB test C) This is a hard task because the list isn't public, but here's an example: https://warpcast.com/mats. Your framing includes an implicit assumption that I would challenge: that the only thing that matters is how interesting the caster is. The follow acts as a status reward to the caster producing interesting content, and only rewarding the top 1% of long term casters with disproportionate status seems like it's missing out on incentivizing the other 99% (there are tradeoffs). Secondly, less interesting casters may be more likely to respond & engage/DM, and thus having some status/follower diversity in the autofollow accounts may lead to other behaviours between users - I don't have the data here though and am speculating, this may be meaningless.
2 replies
0 recast
2 reactions
