Meta is framing AI as more than a feature inside its existing platforms. The company now says large language models are changing how quickly it can create, test and scale new standalone apps.
During this week's second-quarter earnings call, Meta CEO Mark Zuckerberg told investors that the company has more apps in development. The comments follow a recent run of launches, including 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.
AI is changing Meta's app-building rhythm
Meta has tried many times to build new social products outside its biggest platforms. The difference now, according to the company, is that large language models can help teams move faster from idea to launch.
Zuckerberg described that shift directly on the investor call: "I'm…excited about how AI is helping our teams speed up product development," he said. He pointed to Instagram Instants, Forum and Seller as recent examples of standalone products that have already shipped.
He also made clear that Meta does not see these launches as isolated experiments. "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," Zuckerberg said.
That combination matters. Shipping a new app quickly is only one part of the challenge. Meta is also leaning on its recommendation systems to help those apps find users, surface content and grow inside a broader network of Facebook, Instagram and other company products.
A long history of experiments that did not last
The push arrives after Meta spent years looking for standalone social apps that could complement its core services. Earlier efforts produced many launches, but none became the kind of enduring breakout product Meta wanted.
In its earlier days, when the company was known as Facebook, it ran an internal incubator called Creative Labs. That group tested new social concepts and released products including Slingshot, Rooms, Paper, Moments and Riff.
Those apps covered a wide range of ideas: photo sharing, anonymous chat, a Flipboard-style reading experience, collaborative video and other social formats. The effort came to an end in 2015, and the apps were eventually shuttered after struggling to find an audience.
Meta tried again in the early 2020s through an internal R&D group called NPE Team. Its experiments included Bump, Aux, Move, Spark, CatchUp, E.gg, Venue, Hotline, Super, Tuned, BARS and others.
Again, the pattern was the same. Meta launched many concepts, but none became a breakout success, and the apps were shut down.
Threads gives Meta a stronger example
Meta can now point to Threads as a more promising case. The app has reached 500 million monthly active users, and Zuckerberg has said Threads could become the company's next billion-user app.
Threads also shows how Meta's current playbook differs from some earlier standalone experiments. The company used its existing user base to seed the app with people, then promoted it across Facebook and Instagram.
AI is part of that growth story too. Meta says it has seen "significant gains" from AI-powered content recommendations on Threads. That suggests the company is not only using AI to build apps faster, but also to make those apps more useful once people arrive.
The approach creates a clearer path for future launches:
- Use AI and LLMs to speed up product development.
- Launch standalone apps tied to real activity on Meta's platforms.
- Use recommendation systems to help new products scale.
- Apply lessons from Threads to new consumer products.
LLMs are becoming part of the recommendation engine
Meta CFO Susan Li gave investors more detail on how LLMs are changing the company's ranking and recommendation work. "We are finding that LLMs are increasingly capable of delivering ranking and recommendations gains," she said.
Li said LLMs improve existing systems by helping them understand what content is about and by generating better training data. She also said LLM-powered agents are helping engineering teams evaluate content quality, detect trends and test ranking changes.
Meta has already reached one internal milestone in that work. Earlier this year, every Reel and Feed post on Instagram became automatically processed through an LLM and analyzed for topic and tone, which helps improve recommendations.
The company is also developing LLM-native recommendation systems. If those systems work as intended, they could give Meta a stronger way to scale new apps as they arrive, especially when those apps need relevant content and active user loops quickly.
More Meta apps are coming soon
Investors did not press executives for more detail about the specific apps Meta has in development. Their attention was focused more on AI spending and the company's growing enterprise ambitions.
Still, Zuckerberg signaled that the next wave may not be far away. He said the "new consumer products" were "releasing soon."
For Meta, the stakes are straightforward. The company has a long record of launching standalone social experiments that faded. Now it is arguing that AI changes the economics and speed of that process, while recommendation systems may give new apps a better chance to reach the right users.
The coming launches will test whether that belief holds beyond Threads. Meta has built new apps before. This time, it says AI is making the process faster, and that more ideas are already moving through the pipeline.