nicholas 🧨
@nicholas
TikTok proved reco algos + A/B testing new content with small audiences is the winning formula years ago. Reverse Chrono feed sure why not but it is not necessary to have a healthy feed
2 replies
0 recast
1 reaction
Kiran
@neuroswish
yes but a couple of nuances: -this works for tiktok bc the nature of full-screen video means you get implicit +/- signal from users instantly (did they watch the whole thing? did they skip quickly?) - much harder to gauge these signals from tweet-style posts bc of multiple tweets on-screen simultaneously. users *must* like/comment/rt to indicate any sort of preference or signal imo, chrono feed is def an archaic artifact of early social products that had to use your social graph as a proxy to determine your interests. tiktok-style consumption is much more effective but in order for text-based products to leverage it, the UI must change substantially from what it is now.
1 reply
0 recast
0 reaction
nicholas 🧨
@nicholas
100% although its dogshit rn, x is going to show that you can do the same thing with tweets. there's enough indicators from touchscreen + time spent + text analysis to make it work. not to mention the ai tech has gotten so much better since tiktok/toutiao. big problem is fc lacks content to make it work. but is it really a surprise, they've done almost no marketing. attention factory tells amazing story about bytedance paying for preinstalls in china to get douyin off the ground.
0 reply
0 recast
0 reaction
