Meta says AI is making it easier to build new apps — and more are coming
TechCrunch Sarah Perez ● Covered by 6 sources
Zuckerberg says AI is letting Meta ship new apps way faster, and more are coming soon. Meta's flopped at this before, but AI-boosted Threads just hit 500 million users.
Based on reporting by TechCrunch, Sarah Perez — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Mark Zuckerberg spent part of this week's earnings call talking up something that sounds almost mundane compared to Meta's usual AI chatter: shipping apps faster. On Wednesday's second-quarter call, he told investors that large language models are speeding up product development inside Meta, and that the company plans to build out more app ideas and lean on its recommendation systems to help them scale.
This isn't Meta's first attempt at spinning up standalone apps. Back when it was still Facebook, the company ran Creative Labs, an internal incubator that produced Slingshot, Rooms, Paper, Moments, and Riff. All of it got shut down by 2015 after failing to find an audience. A second attempt came in the early 2020s through the NPE Team, which cranked out apps like Bump, Aux, Move, Spark, CatchUp, E.gg, Venue, Hotline, Super, Tuned, and BARS. None of those stuck around either.
Threads is the exception Meta now points to, and it's grown to 500 million monthly active users, with Zuckerberg still predicting it becomes the company's next billion-user product. Meta seeded it using its existing user base and pushed it hard across Facebook and Instagram, but executives say LLMs are doing real work behind the scenes too, powering what CFO Susan Li called significant gains in content recommendations.
Li got specific about how that works: LLMs are making existing ranking systems smarter by actually understanding what content is about and generating better training data, while LLM-powered agents help engineers evaluate content quality, spot trends, and test ranking changes. Earlier this year, Meta hit a milestone where every single Reel and Feed post on Instagram gets automatically run through an LLM for topic and tone analysis. The company is also building LLM-native recommendation systems meant to help future apps scale more easily.
Meanwhile, the recent app lineup already reflects this faster pace — a Marketplace seller app, a standalone Groups app called Forum, a vibe-coded gaming app, a new Instagram photos app, and even an AI bedtime stories experiment. Investors on the call barely blinked at any of this, though; they were far more fixated on AI spending and Meta's enterprise push. But Zuckerberg made clear the app pipeline isn't slowing down, promising new consumer products are releasing soon.
My take — AI-written commentary, not fact-checked reporting
Two failed app incubators and a graveyard of shut-down products should make anyone skeptical that faster shipping alone fixes Meta's real problem, which is that nobody asked for another Meta app. Threads worked because Meta force-fed it to a billion-plus existing users across Facebook and Instagram, not because AI made the code better. Calling that a repeatable playbook feels like a stretch dressed up as a strategy.
Read more about this at: TechCrunch
