Lao Ma wants to open source the X platform algorithm, which is a major change in social media platforms. Lao Ma tweeted early this morning that he will open source the new X algorithm (including all the code used to decide on natural and advertising posts recommended to users) within 7 days. It will be repeated every four weeks and accompanied by detailed developer instructions to help you understand the specific changes. ”Worth noting 1. Why is it said to be a major change in social media platforms? For a long time, the recommendation mechanism of social platforms has been regarded as an "invisible core asset", and the recommendation mechanism is basically a black box model. Everyone has explored the algorithm logic of the platform through practice. The recommendation algorithm of social platforms plays an extremely critical but almost unaudited role: it determines what information is amplified, what viewpoints are ignored, what emotions are stimulated, and what discussions are drowned out. Previously, Lao Ma publicly admitted that there were significant vulnerabilities in the "Recommended for You" system and promised to fix them. Through open-source algorithms, X attempts to transform the long-standing "algorithmic black box" problem into an engineering problem that can be verified, discussed, and improved externally, using technical transparency to hedge against user trust loss. 2. AI based fully AI driven recommendation engine algorithm The algorithm that X is about to release is not based on the heuristic rules of the old Twitter era, but is designed from scratch by the xAI team. The system runs on over 20000 GPUs in the Colossus data center and is deeply involved in real-time online inference, not just for offline training or grading assistance. This means that X has directly handed over the decision-making power of 'what content is worth seeing' to the large model system centered around Grok. Completing system switching without stopping service is described by Lao Ma as "replacing engines during aircraft flight". Behind this metaphor, it implies that the model must perform self calibration in real traffic, real feedback, and continuous data distribution. This is no longer AI in the laboratory, but a human preference system for continuous online learning. 3. So what is the significance of industry and users? 1) Enhance transparency by publicly disclosing algorithm code, allowing users, developers, and society to examine the operation of recommendation systems, avoid hidden biases or manipulation, and ensure fairer content distribution. This helps reveal how algorithms affect information flow, such as prioritizing which posts to display 2) Promoting community engagement and innovation: Open source allows global developers to contribute improvement suggestions, fix bugs, or create derivative versions, driving the evolution of recommendation algorithms. Similar to past initiatives of open-source Twitter algorithms, this can inspire more innovative applications. 3) Promoting industry standards: As a measure of large platforms, this may encourage other companies (such as Meta or TikTok) to follow suit and form more open industry norms, which is conducive to the healthy development of the entire digital ecosystem 4. What is the core logic of X's new algorithm? Looking at the content posted by Lao Ma, his goal for X's new algorithm is: "Even for new users without any fans, as long as the content is excellent enough, it should be seen by many people This means that the number of fans will no longer be the main threshold for content distribution, and recommendation systems will shift from being "network driven" to being "content and model evaluation driven". Grok's recommendation logic will "break the inherent bias of traffic allocation" and no longer limit the exposure of high-quality content based on account size or historical weight. By directly analyzing the quality of the content itself (such as information integrity, innovation, and relevance to user interests), rather than relying on past account data, Grok is expected to give high-quality content from small and medium-sized accounts equal exposure opportunities as top accounts. For creators, this is not simply a positive development, but a change in risk structure: the moat of top accounts is weakened, the threshold for newcomers to start cold is lowered, success relies less on the past and more on whether current content is judged by models as "worth spreading".