Research Sessions use a dedicated data model instead of hidden Channels or shared Items. They reuse connector, extraction, AI, security, and localization services, but their evidence, syntheses, and conversations remain separate because recurring feed concepts such as read state, votes, digests, and channel membership do not apply to one-time research.