Is Altman quietly building an "AI version of X"? ChatGPT social function exposed

According to multiple insiders, OpenAI is testing a social information flow function similar to X (formerly Twitter), and the prototype has been launched. Users can post content in ChatGPT just like posting. Most of these contents revolve around how to use ChatGPT to solve practical problems or create pictures. The posting behavior is jokingly called a "yeet".

This is not groundless. OpenAI has already made a prototype based on image generation function, with an interface close to that of a social platform and an interactive experience of streaming content. CEO Sam Altman has even privately demonstrated it to outsiders and solicited feedback. It is not yet clear whether this social function will be released as a new product or integrated into ChatGPT, but it is certain that it is no longer just a chatbot.

Interestingly, this wave of social experiments happened to be at the time when GPT-4o’s image generation function was becoming popular. Just a few days after its launch, the social platform was flooded with Ghibli-style AI images, and the number of new users exceeded 1 million within an hour. Even Altman joked helplessly: “Stop generating new images. We engineers need to rest.”

Not long after (just last week) , OpenAI launched another heavyweight feature: ChatGPT can remember user conversation records . In simple terms, this AI can remember what you said, how you like to answer questions, and even form a certain "personalized expression style".

These two seemingly unrelated updates actually point in the same direction: AI content × social communication.

On the surface, this is just an expansion of product functions, but from a strategic perspective, OpenAI is evolving from a "tool company" to a "platform product", and social interaction is the most critical step.

There are actually three games behind this:

The first level is the desire for user retention.

One of the biggest bottlenecks of conversational AI is that it is “use it and leave”, which makes it difficult to form stickiness. However, information flow products are masters of retention - X, Douyin, Instagram are all like this. Connecting social content with AI applications can not only lower the threshold and inspire inspiration, but also form a flywheel effect of “watch → try → share → watch”.

Figure: GPT-4o is used to generate Ghibli-style pictures, which are popular on the Internet and still popular today

In the past, GPT had done many functions without much success, but the result was a picture generation that directly set off a social network explosion. This experiment may have made OpenAI realize that AI is not just about making good tools, but also about making people "see, want to forward, and spread". And this kind of communication power is the core competitiveness of social platforms such as X and TikTok.

Altman has seen the explosive power brought by "pictures + social networking".

The second layer, “memory function” is the infrastructure of AI social personality

Once AI can remember you, imitate you, and represent you, the content you post on social platforms can also be generated by AI, and even have a unified style. In the future, we may see that everyone's ChatGPT is like a digital avatar, continuously producing recognizable content in the information flow.

The third level is the battle for platform-level data.

Having a social scene means that you can obtain a steady stream of "real user behavior data" - this is a valuable resource for training the next generation of AI models. Currently, Meta uses Instagram and Facebook data to support the Llama model, and Musk strongly binds X and xAI. If OpenAI cannot obtain similar data inflows, it may be at a disadvantage in the future competition for large models.

Figure: Musk’s xAI’s Grok and X are deeply integrated and can search for the latest tweets.

Social platforms are not only information fields, but also data mines that AI models rely on for survival. If you don’t have one, you must create one.

This experiment of "AI social platformization" is not just a competition among major companies, but is more likely to reconstruct the way content is produced and consumed.

For ordinary users, the addition of information flow may completely change the way we use ChatGPT - it is no longer a "single-person conversation tool", but an open platform where you can get inspiration, imitate others' usage, and actively share;

For content creators, this may be the next AI-native traffic portal, and early deployment is expected to establish influence;

For industry practitioners, the signal behind this is very clear: OpenAI is trying to build a closed loop of "data-content-user", and the step towards platformization has been taken, which deserves continued attention.

When AI is not just a tool but a part of the platform, how should we redefine "content" and "social"? Would you be willing to brush content, post, and build your own AI personality in ChatGPT?