Meta told investors this week that AI is materially compressing how fast it can ship new consumer apps, and that more standalone products are on the way. On the second-quarter earnings call, CEO Mark Zuckerberg pointed to three launches already in market this year — Instagram Instants, a standalone Groups app called Forum, and a standalone Marketplace app called Seller — as proof the internal cycle time has changed. The claim matters because Meta has spent more than a decade trying, and failing, to spin up standalone social apps outside its core platforms.
Zuckerberg framed the shift as structural rather than one-off. Large language models, he argued, let small teams get products in front of users faster, and Meta's existing recommendation systems then handle distribution.
“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”— Mark Zuckerberg, Meta CEO
"AI is improving our core business; it's making our apps more relevant and delivering better results for businesses," Zuckerberg said on the call. "We're starting to deliver more novel products, and we'll have a lot more there soon as well." He did not put a timeline on the next launches beyond "releasing soon."
Key facts
- 01Meta shipped standalone apps for Marketplace (Seller), Facebook Groups (Forum) and Instagram Instants this year, with more consumer products due soon.
- 02Threads has reached 500M monthly active users, on a path Zuckerberg says leads to 1B.
- 03Every Reel and Feed post on Instagram is now automatically processed through an LLM to improve recommendations.
- 04CFO Susan Li said LLM-powered agents are helping engineering by evaluating content quality, detecting trends and testing ranking changes.
- 05Meta's earlier standalone-app pushes — Creative Labs, which ended in 2015, and the NPE Team in the early 2020s — all shut down without a hit.
The history is unflattering. Meta's Creative Labs group — which produced Slingshot, Rooms, Paper, Moments and Riff — was wound down in 2015 after none of the apps found an audience. The NPE Team followed in the early 2020s with more than a dozen experiments including Bump, Aux, Move, Spark, CatchUp, E.gg, Venue, Hotline, Super, Tuned and BARS. Every one of those apps was eventually shut down.
The one exception is Threads, which now has 500 million monthly active users and which Zuckerberg has repeatedly said he expects to become the company's next billion-user app. Meta seeded Threads by tapping the existing Instagram graph and by cross-promoting across Facebook and Instagram, but the company credits its LLM-based ranking stack for a meaningful share of ongoing growth.
CFO Susan Li spelled out where the models are showing up inside the stack. "First, they make our existing systems smarter by understanding what the content is actually about and generating better training data," she said. "Second, LLM-powered agents are also helping with engineering development by evaluating content quality, detecting trends, and testing ranking changes."
“We are finding that LLMs are increasingly capable of delivering ranking and recommendations gains”— Susan Li, Meta CFO
Li added that earlier this year Meta hit an internal milestone: every Reel and Feed post on Instagram is now automatically processed through an LLM and analyzed for topic and tone, feeding back into recommendations. The company is also working on what it calls LLM-native recommendation systems, designed to scale new apps faster once they launch by lowering the cold-start problem that killed earlier NPE experiments.
Investors on the call largely skipped past the new-apps thread to press executives on AI capital expenditure and Meta's enterprise ambitions — a reflection of where Wall Street's attention sits after the company's capex guide reset expectations earlier this quarter. That leaves the consumer roadmap under-scrutinized for now, which suits a company still telling a story about product velocity rather than product hits.
The obvious risk is that shipping apps faster is not the same as shipping apps that stick. Creative Labs and NPE both had internal velocity; what they lacked was distribution beyond the launch news cycle and a durable reason for users to open the app twice. LLM-assisted ranking helps with the second problem only if the underlying product concept holds up, and Threads is so far the sole data point supporting the model.
The bigger read on Meta's Q2 commentary is that the company is repositioning its app portfolio strategy around AI as a manufacturing input rather than a feature. If LLM-assisted development plus LLM-native ranking really do lower the cost of a swing, Meta can afford to launch and quietly kill far more standalone apps than it could a decade ago — and needs only one more Threads-scale hit to justify the entire program. That is the bet embedded in Zuckerberg's promise of more consumer products soon.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
