Meta says AI will finally break its app curse, but the company has two graveyards full of evidence to the contrary

Meta says AI will finally break its app curse, but the company has two graveyards full of evidence to the contrary

Meta's Q2 earnings call featured a familiar promise: AI is making it dramatically easier to ship standalone apps, and more are on the way. Mark Zuckerberg pointed to a recent wave of launches, including a Marketplace sellers app called Seller, a standalone Facebook Groups app called Forum, a vibe-coded gaming app, a new Instagram photos app, and an experiment involving AI bedtime stories 1.

"I expect it to become a lot easier to ship new apps," Zuckerberg told investors, adding that Meta plans to build out more ideas and use its recommendation systems to scale them 1.

Meta has made this promise before. Twice. And both times, the apps ended up in the same place: shut down.

The first graveyard: Creative Labs

When the company was still called Facebook, it ran an internal incubator called Creative Labs to test new social concepts. The group produced Slingshot, a photo-sharing app; Rooms, an anonymous chat app; Paper, a Flipboard competitor; Moments, a photo-sharing app; and Riff, a collaborative video app 1.

Creative Labs was wound down in 2015 after the company struggled to find an audience for any of these experiments 1.

The second graveyard: NPE Team

In the early 2020s, Meta tried again. The NPE Team, an internal R&D group, launched more than a dozen experimental apps over several years: the chat app Bump, social music app Aux, task app Move, dating app Spark, calling app CatchUp, zine maker E.gg, events app Venue, creator Q&A app Hotline, a Cameo competitor called Super, couples app Tuned, and music app BARS 1.

None became a breakout success. They were shut down one by one 2.

What has actually changed

The infrastructure argument has real substance. CFO Susan Li told investors that earlier this year, Meta reached a milestone where every Instagram Reel and Feed post is now automatically processed through a large language model and analyzed for topic and tone to improve recommendations 1. The company is also developing what it calls LLM-native recommendation systems, which it says could help new apps scale faster than previous efforts 1.

There is independent evidence that the recommendation engine is delivering. The Verge reported on the same earnings call that Zuckerberg said time spent on Instagram has increased by double digits year-over-year, driven by improvements to the AI-powered recommendations algorithm 3. If LLMs are genuinely making recommendations smarter at scale, a new app plugged into that engine could find and retain users faster than Slingshot or Rooms ever could.

Meta can also point to one genuine success: Threads, which Meta says has 500 million monthly active users 1. But Threads succeeded primarily because Meta leveraged its existing user base to seed the app and promoted it relentlessly across Facebook and Instagram 1. AI-powered recommendations may have accelerated its growth. They did not create the demand.

Why the apps actually died

The two graveyards tell a story that has nothing to do with build speed. Look at what these products competed with. Slingshot duplicated the photo-sharing and ephemeral messaging that Instagram already offered. Paper was a redesigned interface for the same Facebook feed users already scrolled every day. Rooms asked people to be anonymous on platforms whose advertising business depends on real identities. These apps shipped. They functioned. What killed them was strategic incoherence: Meta kept building standalone products that overlapped with features inside its own flagship apps, then starved them of the distribution that might have let them survive independently.

The NPE Team apps had the same structural problem. Each one targeted a narrow use case that Instagram, WhatsApp, or Facebook could absorb more efficiently. Eventually, they did. The standalone apps died because the mothership consumed them.

The question AI does not answer

Faster development is a real advantage. Smarter recommendation systems are a real advantage. But Meta's app failures were never a build problem. They were a strategy problem. Cheaper engineering does not answer the question that killed Creative Labs and NPE Team: why would users adopt a standalone app when the feature it offers can be folded into Instagram or Facebook?

Meta has never been short of engineering talent. It has been short of a coherent theory for why a standalone app should exist independently of the platforms it would cannibalize. Until that changes, the graveyard will keep growing, whether the apps are coded by humans or assembled by LLMs.