Charles LiuInstagram

Hopefully IG doesnt take this down 😩 #reverseengineering #swe #cs #tech #ai

1:12English(auto-generated)5 segments217 words, 1 min read

Search inside any video

SavedThat transcribes your saved videos and lets you search across all of them instantly. Save this video and find any moment.

TL;DR

This video explores the AI models behind Instagram's algorithm, detailing how they predict user behavior and optimize engagement.

Instagram algorithm analysisAI models in Instagramuser behavior predictionengagement optimizationreel interaction metricscarousel slide algorithmsInstagram app featuressocial media AI technology

Chapters

  1. 0:00Introduction to Instagram's Algorithm
    01
  2. 0:10AI Models Overview
    02
  3. 0:20Predicting App Usage
    03
  4. 0:30Reel Interaction Predictions
    04
  5. 0:40Carousel Slide Selection
    05
  6. 0:50Niche Use Cases of AI
    06
  7. 1:00Conclusion and Next Steps
    07

Transcript

0:00

I reverse engineered Instagram to see what makes its algorithm so addicting. It starts with the 87 different AI models I found in its code base. For instance, this model takes 44 possible inputs, including your battery level, your last 7 days of activity, and the day of the week, and predicts when you'll open the app next. Another model, named L-I-G-E-G-R, takes your reel history and uses that to predict

0:23

how likely you'll complete the reel you're watching, how likely you'll continue watching reels, and how likely you will comment, like, or share that reel. They also have another model that decides which carousel slide to show you first, your likelihood that you will interact or tap on a story, and even a budgeting system that uses what they call common currency

0:41

to see which ads are eligible for viewing. They also have models that serve really niche use cases, just like this one that tracks activity over the past 200 times you open the reels tab just to predict if you'll open it up again and if it should preload that tab for you.

1:01

So yeah, apparently this is what 16 years of retention optimization on an app looks like. There's a lot more to cover, so follow and I will post a part 2.

1:07

There's a lot more to cover, so follow and I will post a part 2.

Are you the creator or rights holder of this video? Request removal of this transcript.

Related Transcripts

Never lose a moment again

Save videos from YouTube, Instagram, and TikTok. Search everything that was said, and jump to the second.

Start Saving Videos - Free Trial