Breaking — Meta's new Friend Bubbles feature for Facebook Reels may appear straightforward—showing users which Reels their friends have watched and reacted to—but the engineering behind it required solving complex scalability problems across billions of users, two engineers from the Facebook Reels team revealed. The feature, which launched recently, is now being highlighted as a case study in balancing user-facing simplicity with backend complexity.
What Friend Bubbles Does
Friend Bubbles visually aggregates Reels that a user's friends have engaged with, creating a shared social discovery layer. On the Meta Tech Podcast, software engineers Subasree and Joseph detailed the process of building the feature, from evolving the machine learning model to handling platform-specific behaviors between iOS and Android.
"On its face, it's just bubbles showing what friends liked," said Subasree. "But to make that work for every user, we had to rethink our ranking models and data pipelines from scratch."
The Hidden Complexity
The engineers explained that the biggest hurdle was scaling collaborative signals—turning sparse friend interactions into meaningful recommendations for billions of people. Joseph noted, "We discovered that the key wasn't just in what friends watched, but in the subtle timing of their reactions. That insight changed our entire approach."
Another unexpected challenge was the difference in user behavior across platforms. "iOS users tend to view Reels in shorter bursts, while Android users often watch longer sequences. The model had to adapt to both without sacrificing personalization," Subasree added.
Background
Friend Bubbles is part of Meta's broader push to make Reels more social, competing with TikTok's algorithm-driven discovery. The feature leverages a machine learning model that originally ranked Reels by individual interest, but had to be overhauled to account for social graph data. The engineering team spent months iterating before finding the right balance.
The episode of the Meta Tech Podcast, which explores these challenges, is available on Spotify, Apple Podcasts, and Pocket Casts. Meta positions the feature as a step toward “social discovery” that scales without losing relevance.
What This Means for Users and Meta
For users, Friend Bubbles promises more relevant Reels curated by their social circle, potentially increasing engagement and time spent on the platform. For Meta, the success of this feature could inform how it builds other social AI tools, like shared feed elements in Facebook or *Instagram*.
Engineering-wise, the project demonstrates that even the simplest-seeming features require massive infrastructure when applied at Meta's scale. "We learned that friend-based discovery isn't a small add-on; it's a whole new data problem," said Joseph. "But once you solve it, the product feels effortless."
Listen and Learn More
The full podcast episode is available now. Meta invites feedback on Instagram, Threads, and X. Career opportunities related to this work can be found on the Meta Careers page.