Meta uses LLMs to ship apps faster, more launches coming

Meta is leaning on large language models to slash app development time and fuel a new wave of standalone products, CEO Mark Zuckerberg said on the company's Q2 earnings call. Citing recent launches—Instagram Instants, the Forum groups app, Seller for Marketplace, a vibe-coded gaming app, and an AI bedtime stories experiment—Zuckerberg said AI lets teams ship software faster and test more ideas. "I expect it to become a lot easier to ship new apps," he said, promising more consumer products are "releasing soon."

This strategy contrasts with Meta's long history of failed standalone apps, from Creative Labs' Slingshot and Rooms to NPE Team's Bump and Tuned. The difference now is AI's measurable impact: CFO Susan Li said every Instagram Reel and Feed post is automatically processed through an LLM for topic and tone analysis, driving significant ranking gains. Threads, Meta's one successful new app, has hit 500 million monthly active users, buoyed by LLM-powered content recommendations. Meta is also developing LLM-native recommendation systems to help future apps scale quickly, though investors on the call focused more on AI spending and enterprise ambitions than the upcoming product pipeline.