hi its scri馃叡锔弔on my non professional account where i tweet nonsense 馃槑馃檳

Joined July 2020
I realized there was not enough EDM in Shop It I can't wait to see what players do with this. My demos are always 100x lamer than UGC
1
20
1,147
Replying to @ericzona_
bro is staying ACTIVE
1
1
206
michael jackson lives in my embeddings
4
190
gravity can't contain my drip
1
166
Replying to @RbxXAXA
I haven鈥檛 had time to update it much in the past 2 weeks Last night I finally added some light monetization and a pretty big gameplay improvements. Still 0 home recs though, even after being live for 5-6 weeks and 4k visits Genuinely insane behavior I think
2
126
I tried something very bold (and stupid) - I quietly launched an alpha build of my game with no ads. I paid the $100k R to get my game approved for u16 immediately. I haven't announced the launch anywhere either. I wanted to see what would happen if you didn't try to pay 2 win.
5
29
6,199
this is now live and automatically taking down likebait! Clips need 100 likes before the model reviews them. Planning on decreaseing this to 75 tonight So far so good, 100% of the clips it's muted on the algorithm are low quality likebait
2
68
We just had our best month yet over at @robloxClipIt ! We broke our previous record of 366m monthly views - now at 452m monthly. Super excited to launch more algorithm improvements and help even more creators grow.
16
854
each of these points is a clip the model was *not* trained on blue = high quality red = low quality (like-bait) if I randomly select the low-quality clips (red points) this is what I see (the last clip has 63k likes holy algorithm gaming)
1
2
222
day 2 of sloperating up some algorithm improvements for Clip It We now have ~10x more labeled data and more model inputs - when the model flags something as non-likebait it's usually right (96.1%) - when it does see actual like-bait, it correctly identifies it 55% of the time
2
14
617
we train a few different types of models/classifiers then pick the best performer. We have a pretty small labeled dataset so far but we're already seeing some promising results. Excited to see how this looks when we 10x & 100x the data, as well as feed it more interesting data
1
6
213
starting to have *WAY* too much fun geeking out over clip it data We have real people label clips 0-1 by "quality" - the goal is to catch low quality clips as well as ultra high quality clips in the Clip It algorithm Then we train some small models to predict this quality value
1
12
520
can Roblox PLEASE stop cramming more random features into the analytics pages? Why do I have to scroll like two pages worth of these random tabs when I only use like 5 of them. Every day it gets more confusing.
3
7
1,237
bro is a majestic goofball
4
432
chat, meet Calvin鈥檚 new sister 馃専Callie馃専
1
1
17
2,050