Fab is a hybrid recommendation system that combines content-based and collaborative filtering to improve the effectiveness of Web-based recommendations. It addresses the limitations of both individual approaches by leveraging the strengths of each. Content-based filtering recommends items similar to those a user has liked in the past, while collaborative filtering recommends items liked by users with similar tastes. Fab integrates these methods to provide more accurate and diverse recommendations. Fab's architecture is designed to handle the challenges of the Web, including a growing number of users and documents. It automatically identifies communities of interest, enabling enhanced group awareness and communication. The system uses user feedback to refine recommendations, ensuring that users receive items that align with their preferences. The system works by maintaining user profiles based on content analysis and comparing these profiles to determine similar users for collaborative recommendations. Users receive items both when they score highly against their own profile and when they are rated highly by a user with a similar profile. This hybrid approach avoids the limitations of pure content-based or collaborative systems. Fab's collection agents find pages relevant to specific topics, while selection agents find pages for specific users. The central router forwards pages to users whose profiles match. User feedback is stored in their personal selection agent's profile, ensuring it is not drowned out by other users' feedback. Users rate the recommended pages, which are used to update their profiles and influence future recommendations. The system's performance is evaluated through experiments showing that learned profiles can predict user rankings effectively. Fab outperforms other systems in terms of personalization and adaptability. The system is designed to scale efficiently and dynamically adjust to user interests, enabling it to serve a large number of users with a fixed pool of agents. Future work includes studying the effects of scaling up the number of users and further understanding the roles of collaborative and content-based components. Fab is an effective implementation of a hybrid system that addresses the shortcomings of pure content-based and collaborative approaches.Fab is a hybrid recommendation system that combines content-based and collaborative filtering to improve the effectiveness of Web-based recommendations. It addresses the limitations of both individual approaches by leveraging the strengths of each. Content-based filtering recommends items similar to those a user has liked in the past, while collaborative filtering recommends items liked by users with similar tastes. Fab integrates these methods to provide more accurate and diverse recommendations. Fab's architecture is designed to handle the challenges of the Web, including a growing number of users and documents. It automatically identifies communities of interest, enabling enhanced group awareness and communication. The system uses user feedback to refine recommendations, ensuring that users receive items that align with their preferences. The system works by maintaining user profiles based on content analysis and comparing these profiles to determine similar users for collaborative recommendations. Users receive items both when they score highly against their own profile and when they are rated highly by a user with a similar profile. This hybrid approach avoids the limitations of pure content-based or collaborative systems. Fab's collection agents find pages relevant to specific topics, while selection agents find pages for specific users. The central router forwards pages to users whose profiles match. User feedback is stored in their personal selection agent's profile, ensuring it is not drowned out by other users' feedback. Users rate the recommended pages, which are used to update their profiles and influence future recommendations. The system's performance is evaluated through experiments showing that learned profiles can predict user rankings effectively. Fab outperforms other systems in terms of personalization and adaptability. The system is designed to scale efficiently and dynamically adjust to user interests, enabling it to serve a large number of users with a fixed pool of agents. Future work includes studying the effects of scaling up the number of users and further understanding the roles of collaborative and content-based components. Fab is an effective implementation of a hybrid system that addresses the shortcomings of pure content-based and collaborative approaches.