Varun Srinivasan on Warpcast

Varun Srinivasan pfp

Varun Srinivasan

If I were to write ip a quick history of the Warpcast feed algo, what would you be most interested in?

40 replies

14 recasts

446 reactions

shazow pfp

Curious if there has been any collaboration on it with third party clients? Feels like solving spam requires some amount of coordination, otherwise it can manifest as an externality on other clients?

1 reply

0 recast

0 reaction

Cameron Armstrong pfp

Cameron Armstrong

The major eras and the important learnings/convictions y’all built through each Who is/was stickied and why

1 reply

1 recast

22 reactions

Garrett pfp

the changes over time and the reasoning behind them

0 reply

0 recast

4 reactions

Luciano pfp

show us the data on why you prefer rando algo to chronological feed https://warpcast.com/luciano/0xf5faeffa

0 reply

0 recast

0 reaction

sahil pfp

1) what was the most important input data (from the graph) for your models? 2) were the feedback loops (KPIs) for model performance based on warpcast or farcaster data? If both, what was the split?

0 reply

0 recast

0 reaction

Devin Conley pfp

The testing and evaluation process for comparing approaches. e.g what metrics are being optimized, a/b testing, back testing, etc.

0 reply

0 recast

0 reaction

Vladyslav Dalechyn pfp

Vladyslav Dalechyn

How does algo react to “breaking/just-in” trends changes? I am a noobie at ranking but can assume that in a non-ML algo you introduce time decay and while the “topic” has not even started decaying and more and more people engage with it, it rank skies up to the roof. But in ML-based feed algos I assume “feeds” are prepared beforehand- “feed_1”, “feed_2” etc. And if so, is there some kind of extensive workload that the algorithm is experiencing when a “just-in” trending cast shows up that must be seen? As an example, in the “other” platform news about an “earthquake” or “missile attack” appear almost instantly, at nearly at the first three posts. Is it simply because lots of people post about it? Thanks!

0 reply

0 recast

0 reaction

Haole pfp

The MVP code which third party client can learn and adopt

0 reply

0 recast

0 reaction

KMac🍌 ⏩ pfp

Forget the writing. Do one of those excalidraw videos. 🤩

2 replies

1 recast

37 reactions

Zach pfp

- what was the initial goal of algo vs chrono - how did the algo change over time - why was each change made - what does algo prioritize now + why - what’s the future goal(s) for feed algo - what open questions still remain

0 reply

0 recast

0 reaction

Pierre Pauze 🔵 🚽 pfp

Pierre Pauze 🔵 🚽

What's driving the algo to highlight one content over another? I really get the feeling that there's a significant slowdown in traction, whereas we're building something substantial with /artcoin on farcarster. Having some details on how and why would be v reassuring.

0 reply

0 recast

0 reaction

Rch 🎩 pfp

what was one thing that you didn't expect to work and did. and vice versa. and your thoughts on the why

3 replies

0 recast

8 reactions

frdysk.framedl.eth pfp

frdysk.framedl.eth

did shadowban ever exist pre-permissionless?

0 reply

0 recast

3 reactions

MattwithouttheT 🎩 pfp

MattwithouttheT 🎩

I’m always curious about and interested in the reasoning behind significant changes.

0 reply

0 recast

3 reactions

$NEGA 🔺🎩 pfp

What happened to Trend casts? There used to be new "rising stars", now it's the same people

0 reply

0 recast

1 reaction

Dvyne pfp

How bots are detected and put in spam Cause somehow, many get locked in spam replies

0 reply

0 recast

1 reaction

ethdegen.degen.eth🎭 pfp

ethdegen.degen.eth🎭

Why it's so unbelievably shit

0 reply

0 recast

0 reaction

Zenigame pfp

a bunch of the little experiments like that one day where you turned things chronological and then no one ever asked for it again

0 reply

0 recast

0 reaction

Drugbiih Kkjj pfp

How to make it easier

0 reply

0 recast

0 reaction