Core Insights - The article discusses the launch of X's new recommendation algorithm, which has transitioned to an AI-driven model, moving away from manual rules and features [2][3][10]. Group 1: Algorithm Overview - The new X recommendation algorithm has been open-sourced, utilizing a Transformer architecture similar to that of xAI's Grok model [2]. - The algorithm removes most manual rules and features, relying instead on AI to learn user preferences based on past interactions [3][4]. - Two main systems drive content delivery: Thunder, which ensures users see updates from accounts they follow, and Phoenix, which predicts content users may like even if they do not follow the source [4][6]. Group 2: Content Evaluation - Phoenix evaluates content based on predicted user interactions, considering 15 potential behaviors, both positive and negative [6][7]. - The algorithm prioritizes content that is likely to generate engagement, rather than content that has already received high engagement [7][11]. - Factors such as comment engagement and user reading time are weighted heavily, with comments and author replies being significantly more valuable than likes [11][12]. Group 3: Implications for Creators - Creators are encouraged to adapt their content strategies to align with AI logic, moving away from traditional "traffic tricks" [10][13]. - The algorithm favors content that stimulates discussion, suggesting that engagement through comments is more beneficial than mere likes [28][39]. - The system's focus on emotional and controversial content may lead to a shift in creator behavior, potentially prioritizing sensationalism over thoughtful expression [13][14]. Group 4: Future Considerations - The algorithm will be updated every four weeks, providing creators with opportunities to adjust their strategies based on performance data [12]. - The shift towards AI-driven recommendations raises questions about the authenticity of content creation, as creators may feel pressured to cater to algorithmic preferences [13][39].

Core Viewpoint - GitHub has made Elon Musk's open-source recommendation algorithm system fully visible, which is primarily driven by AI models [1][2] Group 1: Algorithm Transparency and Community Reaction - The open-source announcement has generated significant excitement within the community, with many praising the transparency of the system [2] - Musk acknowledged the algorithm's shortcomings, stating it is "dumb" and requires substantial improvements, but emphasized the importance of transparency in the improvement process [4][5] - Musk has consistently criticized the previous platform's lack of openness and has followed through on his promise to publicly share Twitter's core recommendation algorithm since the acquisition [6][7] Group 2: Algorithm Mechanism - The recommendation system is built on a Transformer architecture similar to Grok-1, which learns from users' historical interactions (likes, replies, retweets) to recommend content [9] - The system begins by identifying the user and their recent activities, aiming to create a "real-time user profile" without pre-set assumptions [12][13] - Two types of user information are collected: Action Sequences (direct interest signals) and Features (long-term attributes) [14] Group 3: Content Filtering and Scoring - The algorithm filters through a vast amount of tweets to select a few thousand potentially relevant ones, using both familiar and external sources [16][17] - The system employs a Hydration module to complete candidate tweet information and a Filtering module to eliminate unwanted content [21][22] - The final scoring is done by a Phoenix ranking model, which predicts various user interactions and assigns scores based on weighted combinations of these predictions [25][26] Group 4: Key Features of the System - The system is purely data-driven, rejecting manual rules and allowing AI models to learn directly from raw user data [33] - It utilizes a candidate isolation mechanism to ensure independent scoring of each piece of content [34] - The algorithm predicts multiple user behaviors rather than providing a single recommendation score [36] - The modular design of the system supports rapid iteration and development [37] Group 5: Acknowledgment of Limitations - Despite the praise for transparency, the algorithm has been criticized for certain flaws, such as the lack of a time decay mechanism for "block" signals, which could negatively impact account recommendations [39][41] - Musk himself acknowledged the algorithm's deficiencies, indicating a commitment to ongoing improvements and updates every four weeks [42][44]

Core Viewpoint - Beijing Zhongchang Shiji Information Technology Co., Ltd. (referred to as "Shiji Information") plans to acquire a 13.50% minority stake in its subsidiary Shenzhen Sihon Software Co., Ltd. through a share issuance, despite Sihon Software experiencing declining revenues but increasing gross margins in recent years [1][5]. Group 1: Financial Performance - Sihon Software's technical service revenue has declined over the reporting periods (2023: 268 million yuan, 2024: 253 million yuan, 2025: 86 million yuan), while software product revenue has also decreased (2023: 90.57 million yuan, 2024: 72.36 million yuan, 2025: 16.50 million yuan) [2]. - Despite the revenue decline, the gross margin for payment technology services has increased significantly, reaching 60.07% in 2025, compared to 42.72% in the previous IPO application period [2]. Group 2: Customer Structure - The top five customers of Sihon Software contributed over 80% of its revenue during the reporting periods, with the first customer, Suixing Pay, contributing 44.88% of the revenue in early 2025 [3]. - The gross margin for Sihon Software's payment technology services is significantly higher than that of competitors like Lakala, indicating a sustainable business model [3]. Group 3: Cost Management - Sihon Software has seen a continuous decline in operating expenses, with sales expenses dropping from 35.96 million yuan to 7.67 million yuan, and R&D expenses decreasing from 53.24 million yuan to 13.30 million yuan [4]. - The company has successfully transitioned towards SaaS, with SaaS software revenue increasing from 15.06% in 2020 to 43.66% in early 2025, providing a new growth driver [4]. Group 4: Acquisition and Synergy - The acquisition will enhance Shiji Information's control over Sihon Software, which has served over 700,000 retail stores and demonstrated stable profitability [5]. - The transaction is subject to regulatory approval, and its completion is expected to positively impact the financial structure and business layout of the listed company [6].