Breaking News: Meta Reveals the Engineering Feat Behind Friend Bubbles
Facebook Reels has quietly launched Friend Bubbles, a feature that surfaces Reels watched and reacted to by a user's friends. Despite its simple appearance, the feature required a massive engineering overhaul to scale to billions of users.
In an exclusive episode of the Meta Tech Podcast, software engineers Subasree and Joseph detailed the complex machine learning model and cross-platform hurdles that made Friend Bubbles possible.
“What seems like a straightforward social cue actually demanded us to rethink our entire recommendation pipeline,” said Subasree, a senior engineer on the Facebook Reels team. “We had to balance relevance with privacy while maintaining real-time performance for billions of daily views.”
Background
Meta has been investing heavily in social discovery—surfacing content based on friends' activity rather than pure algorithmic recommendation. Friend Bubbles is the latest iteration of this strategy, which began with Stories and later expanded to Reels.
Early tests showed that users were more likely to engage with Reels if they saw a friend had already reacted. However, engineering that feature at scale proved challenging, particularly on iOS where notification permissions and background refresh limits differ from Android.
“We discovered that iOS and Android users behave very differently when it comes to passively sharing activity,” Joseph explained. “That forced us to build separate signal-processing layers for each platform.”
The Surprising Insight That Unlocked Everything
The team’s breakthrough came from analyzing reaction timing. Researchers found that showing a friend’s “like” within 24 hours of the Reel’s posting significantly boosted click-through rates, while older reactions had little effect.
This insight led to a lightweight real-time indexing system that prioritizes fresh friend interactions without overwhelming the main recommendation engine. The result: Friend Bubbles now processes trillions of social signals daily with latency under 50 milliseconds.
“We essentially built a secondary graph that maps social proximity to content freshness,” said Subasree. “It’s like a radar that only pings when your friends are actively watching something new.”
What This Means
For users, Friend Bubbles makes Reels feel more like a shared experience—similar to watching TV with friends who comment in real-time. For Meta, the feature strengthens the social graph, which is critical for ad targeting and ecosystem stickiness.
Industry analysts see this as Meta’s answer to TikTok’s virality. By amplifying friend-based discovery, Facebook aims to increase time spent on Reels and reduce churn to competitors.
“This is a strategic move to differentiate Reels from TikTok’s algorithm-driven feed,” noted a social media analyst at Gartner. “Friend Bubbles adds a layer of trust and serendipity that pure machine learning can’t replicate.”
Technical Challenges Overcome
Scaling Friend Bubbles required solving three major challenges: privacy (ensuring users can’t be tracked without consent), speed (surface bubbles instantly), and quality (avoiding spam or irrelevant friend activity).
The engineering team deployed a tiered architecture: a lightweight cache for hot data (recent friend reactions) and a separate sparse index for longer-tail signals. On Android, they leveraged foreground services to keep the bubble data fresh; on iOS, they used push-based updates to comply with Apple’s background limits.
How to Access Friend Bubbles
Friend Bubbles is rolling out globally on both iOS and Android. Users will see circular profile pictures of friends on Reels thumbnails in the Explore or Home feed. Tapping a bubble reveals that friend’s reaction (like, love, or comment).
Meta confirmed the feature respects existing privacy controls: users can hide their activity from specific friends or turn off the feature entirely via Settings > Reels > Friend Bubbles.
Listen to the Full Technical Deep Dive
For a deeper look, listen to the Meta Tech Podcast episode featuring Subasree and Joseph. Available on:
- Spotify
- Apple Podcasts
- Pocket Casts
Send feedback via Instagram, Threads, or X.
About Meta Tech Podcast
Produced by Meta, the podcast dives into engineering from low-level frameworks to end-user features. For career opportunities, visit the Meta Careers page.