On its recent second-quarter earnings call, Meta said that large language models (LLMs) are accelerating product development and enabling faster launches of standalone apps.
Meta Chief Executive Officer Mark Zuckerberg told investors that the company has rolled out or is testing several new apps and features, including a seller app for Marketplace (Seller), a standalone Facebook Groups app (Forum), a vibe-coded gaming app, a new Instagram photos app, and an experiment involving AI-powered bedtime stories. He also noted that earlier this year Meta shipped Instagram Instants.
Zuckerberg said: “I’m…excited about how AI is helping our teams speed up product development. Earlier this year, we shipped Instagram Instants. We also just launched Forum, a standalone Groups app, and Seller, a standalone Marketplace app. I expect it to become a lot easier to ship new apps. So we are planning to build out more ideas and use our recommendation systems to scale them.”
How LLMs matter
Meta outlined several ways LLMs are contributing to faster product development and growth:
- They make recommendation systems smarter by better understanding what content is about and its tone, which generates improved training data.
- LLM-powered agents assist engineering work by evaluating content quality, spotting trends, and testing ranking changes.
Meta Chief Financial Officer Susan Li said, “we are finding that LLMs are increasingly capable of delivering ranking and recommendations gains.” Li added that earlier this year Meta reached a milestone: every Instagram Reel and Feed post is now automatically processed through an LLM and analyzed for topic and tone to improve recommendations.
The company is also working on LLM-native recommendation systems intended to help scale future apps more effectively.
Past attempts and the Threads example
Meta has previously tried to incubate standalone social apps. When it was Facebook, the company ran Creative Labs, which produced launches such as Slingshot, Rooms, Paper, Moments and Riff; those efforts ended in 2015. In the early 2020s, Meta’s NPE Team tested several other apps including Bump, Aux, Move, Spark, CatchUp, E.gg, Venue, Hotline, Super, Tuned and BARS, none of which became lasting successes.
Today Meta points to Threads as a concrete example of AI helping an app scale: Threads now has 500 million monthly active users. Meta relied heavily on its existing user base to seed Threads and promoted it across Facebook and Instagram, and the company also credits LLM-driven content recommendations with providing “significant gains.”
Investors and what’s next
Investors focused more on AI spending and Meta’s enterprise ambitions during the call than on specific questions about the new consumer apps in development. Zuckerberg indicated, however, that people will not have to wait long to see what’s next, saying the company’s “new consumer products” are “releasing soon.”
