Unraveling X’s Algorithm: The Open-Source Magic Behind Grok’s Personalized “For You” Feed

Unraveling X’s Algorithm: The Open-Source Magic Behind Grok’s Personalized “For You” Feed



X Algorithm Transparency: Understanding the ‘For You’ Feed

X Algorithm Transparency: Understanding the ‘For You’ Feed

X has lifted the veil on its platform’s inner workings, specifically detailing how the ‘For You’ feed is curated for users globally. The company has made its algorithm code available on GitHub, a move intended to boost transparency.

In announcing its newly open-sourced algorithm, X has simplified the operational process into ten key stages.

The feed is developed through multiple phases, utilising the company’s proprietary Grok AI model, marking a transition from manually set guidelines to AI-driven solutions. Essentially, X’s mechanism sifts through over 100 million posts, filters them to around 1,500 highly pertinent entries, and showcases these on users’ feeds.

How Grok’s Open-Sourced System Curates the ‘For You’ Feed

The initial phase indicates that the system gathers users’ personal data, including the accounts they follow along with the posts they’ve previously liked, shared, or commented on. This approach helps in establishing a digital representation of a user’s interests.

The subsequent phase involves collecting posts and interactions from users’ network accounts, which are accessed via a rapid system named “Thunder.” This system also identifies content from outside the user’s network that aligns with their interests using AI analysis.

Enhancing Content Relevance

Next, X’s system enriches the filtered posts with additional metadata, such as whether a user subscribes to X Premium, post paywall details, video quality, and other relevant information. It also eliminates duplicate posts, outdated content, and entries featuring keywords that the user has muted. This process guarantees that the feed avoids showing content that the user has already bypassed or wishes to avoid.

Ranking Posts for Engagement

Grok plays a crucial role in the ranking methodology by allocating significance scores to each post. This advanced AI model predicts the likelihood of user interaction with a post, whether through likes, comments, shares, or clicks. Based on these assessments, each post is scored, highlighting high-quality content while diminishing the visibility of less relevant entries.

Finalising the ‘For You’ Feed

Ultimately, the ‘For You’ feed is compiled from the top-rated posts. A thorough final check confirms that no deleted posts, spam, or inappropriate content is included in the feed.



Exit mobile version