Meta is using artificial intelligence to accelerate software development, test new ideas more quickly, and launch a wider range of consumer applications.
During the company’s second-quarter earnings call, Meta CEO Mark Zuckerberg said the company has several new apps in development and expects AI to make product launches easier and faster.
His comments followed a series of recent experiments and releases, including standalone applications for Facebook Groups and Marketplace sellers, an AI-assisted gaming app, a new Instagram photo product, and an experiment involving AI-generated bedtime stories.
AI Is Changing How Meta Builds Products
Meta believes large language models are significantly reducing the time required to develop and test software.
Zuckerberg told investors that AI is helping Meta’s teams accelerate product development and move more ideas from concept to launch.
Earlier this year, the company released Instagram Instants, Forum, a standalone app for Facebook Groups, and Seller, an app designed for Marketplace sellers.
Zuckerberg said Meta expects it to become increasingly easy to launch new applications and plans to develop more ideas while using its recommendation systems to help those products grow.
He added that artificial intelligence is also improving Meta’s core business by making its apps more relevant to users and delivering stronger results for businesses.
Meta has started introducing more experimental products, Zuckerberg said, with additional consumer applications expected to launch soon.
Meta Has Tried to Build Standalone Apps Before
Meta has spent years attempting to create successful social applications outside its main platforms.
Many of those efforts failed to attract sustainable audiences.
When the company was still known as Facebook, it operated an internal incubator called Creative Labs. The initiative was created to test new social-media concepts and launch experimental products.
Creative Labs released several applications, including:
- Slingshot, a photo-sharing app
- Rooms, an anonymous chat platform
- Paper, a news-reading app designed to compete with Flipboard
- Moments, a photo-sharing service
- Riff, a collaborative video app
The programme ended in 2015, and the applications were eventually discontinued after failing to achieve significant adoption.
The NPE Team Produced More Experiments
Meta later launched another internal research and development unit called the NPE Team.
The group tested a wide variety of social and communication apps during the early 2020s.
Its projects included Bump, Aux, Move, Spark, CatchUp, E.gg, Venue, Hotline, Super, Tuned, and BARS.
These applications covered areas such as chat, music, dating, task management, events, creator interaction, and video.
However, none became a major commercial success, and Meta eventually shut them down.
Threads Gives Meta a Stronger Example of App Growth
Meta can now point to Threads as evidence that it may be better positioned to scale new applications.
Threads has reached 500 million monthly active users, and Zuckerberg has repeatedly suggested that it could become Meta’s next platform to exceed one billion users.
The company used its existing Instagram audience to support the initial launch of Threads.
It also promoted the app heavily across Facebook and Instagram, giving it an advantage that most new social platforms do not have.
However, Meta says artificial intelligence has also contributed to Threads’ growth.
The company has reported significant improvements from AI-powered content recommendations, which help users discover posts that are more relevant to their interests.
Large Language Models Improve Content Recommendations
Meta CFO Susan Li said large language models are becoming increasingly effective at improving ranking and recommendation systems.
According to Li, these models strengthen Meta’s existing systems in two main ways.
First, they help the company understand what content is actually about, allowing Meta to create better training data and improve recommendation accuracy.
Second, AI-powered agents assist engineering teams by evaluating content quality, identifying emerging trends, and testing changes to ranking systems.
These capabilities allow Meta to show users content that is more closely aligned with their interests, potentially increasing engagement and time spent across its applications.
Every Instagram Feed and Reel Post Is Analysed by an LLM
Meta has also reached a major milestone in the use of AI on Instagram.
Every Reel and Feed post on the platform is now automatically processed through a large language model.
The system analyses the topic and tone of each piece of content, helping Meta understand its meaning more accurately.
This gives the company more information than traditional engagement signals such as likes, shares, comments, and watch time.
By analysing the actual subject and tone of a post, Meta can match content with users who are more likely to find it relevant.
The result is a more detailed and context-aware recommendation system.
Meta Is Building LLM-Native Recommendation Systems
Meta is not only adding large language models to its existing recommendation systems.
The company is also developing recommendation technology designed around LLMs from the beginning.
These LLM-native systems could help Meta scale new applications more effectively.
New social apps often struggle because they do not initially have enough content, behavioural data, or user activity to provide strong recommendations.
Meta has access to large volumes of content and engagement data across Facebook, Instagram, and Threads.
It can combine those resources with AI systems capable of analysing topics, trends, quality, and user interests.
This could give future Meta applications a stronger starting point than the company’s earlier experiments.
AI-Assisted Coding Could Reduce Development Time
Large language models and AI coding tools are changing how technology companies build software.
Development teams can use these systems to generate code, identify bugs, test features, analyse user behaviour, and evaluate product quality.
This may allow Meta to create prototypes more quickly and test a larger number of ideas without dedicating long development cycles to every project.
Faster product development, however, does not guarantee success.
Meta will still need to create applications that solve genuine user problems and provide enough value to encourage people to return regularly.
Meta’s Existing Platforms Offer a Distribution Advantage
Meta has a major advantage over most companies launching new consumer apps: direct access to billions of users.
The company can promote new products through Facebook, Instagram, WhatsApp, and Threads rather than building an audience from the beginning.
Threads demonstrated how powerful this distribution network can be.
Instagram users were able to join Threads easily, while Meta promoted the service across its established platforms.
If Meta combines this distribution power with advanced recommendation systems and faster AI-assisted development, its future applications could have a better chance of gaining traction.
Investors Remain Focused on AI Spending
Investors did not ask Meta executives for further details about the applications currently in development.
Instead, much of the earnings call focused on the company’s rising AI infrastructure spending and its growing ambitions in enterprise technology.
Meta did not reveal the names, features, or specific release dates of its upcoming consumer products.
However, Zuckerberg indicated that users may not have to wait long, saying new consumer products would be released soon.
Can Meta’s New App Strategy Succeed?
Meta’s history shows that launching many applications does not necessarily produce a successful platform.
Creative Labs and the NPE Team created numerous products, but none developed into a lasting standalone business.
The company now believes AI could change that outcome.
Large language models can accelerate development, improve content understanding, strengthen recommendations, and support engineering teams.
Threads also offers Meta a more successful model to follow.
Its growth combined three important advantages: access to an existing audience, aggressive promotion across Meta platforms, and AI-powered recommendations.
Even so, the success of Meta’s future apps will depend on more than speed and distribution.
The company will need to offer products that meet real user needs, provide a distinct experience, and remain valuable after the initial launch period.
Conclusion
Meta is using artificial intelligence to transform how it develops and launches consumer applications.
Large language models are helping the company accelerate software development, analyse content, improve recommendations, and test new ideas more efficiently.
Meta is also using its existing platforms and massive user base to give new applications a stronger launch.
Threads provides the clearest example of this strategy so far, combining Instagram-powered distribution with AI-driven content recommendations.
The key question is whether Meta can turn faster development into sustainable products that users continue to value over time.
Read the article in
