Meta's AI-Powered App Launches Are Finally Paying Off

Meta's latest apps, built with the help of large language models (LLMs), are gaining traction, and the company has big plans for more launches.

Meta has been trying to produce new social apps for years, but it wasn't until the company started using LLMs that it saw real progress. During the second-quarter earnings call, Meta CEO Mark Zuckerberg said that the social giant has new apps in the works, following a recent spate of other launches that included an app for Marketplace sellers, one for Facebook Groups, a vibe-coded gaming app, a new photos app from Instagram, and an experiment involving AI bedtime stories.

Meta has spent years trying and failing to produce new, standalone social apps to complement its core platforms. Now, the company says that LLMs make it possible to ship software faster, allowing it to test new ideas at a quicker pace.

"I'm...excited about how AI is helping our teams speed up product development," Zuckerberg told investors on Wednesday's call. "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," he said.

One example of how AI is helping new apps scale is Meta's Threads app, which now has 500 million monthly active users. Zuckerberg likes to say Threads will one day become the company's next billion-user app. With Threads, Meta learned to heavily lean on its existing user base to help initially seed the app with people, then continued to heavily promote it across its existing platforms, including Facebook and Instagram. But LLMs are another key factor in Threads' growth, as the company said it sees "significant gains" from its AI-powered content recommendations.

"We are finding that LLMs are increasingly capable of delivering ranking and recommendations gains," Meta's CFO Susan Li told investors on the call. "First, they make our existing systems smarter by understanding what the content is actually about and generating better training data. Second, LLM-powered agents are also helping with engineering development by evaluating content quality, detecting trends, and testing ranking changes."

The company is also developing LLM-native recommendation systems, which could help it to better scale new apps as they arrive.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.