Vellum

Vellum is a personal artificial intelligence assistant designed to operate as a dedicated layer for an individual’s daily life. Unlike general-purpose chatbots that rely on session-based interactions, Vellum maintains a persistent memory system that continuously stores and recalls user-specific information such as goals, preferences, schedules, and personal context. This memory layer works similarly to human recollection, enabling the assistant to understand nuanced requests without requiring repetitive instructions. Users can set up the assistant in under thirty seconds by choosing a name, linking their accounts, and sharing their objectives. Once configured, Vellum runs around the clock on the Vellum Platform, eliminating the need for local hardware like a virtual private server or a Mac Mini. The assistant is accessible from multiple touchpoints including desktop (Mac and web), mobile (iOS), messaging platforms (Slack and Telegram), and telephone calls, allowing users to interact with it wherever they are. Vellum learns from user feedback and adjusts its behavior over time; the more a person provides instructions or corrections, the more precisely the assistant aligns with their expectations. It can perform a wide range of tasks such as managing calendars, setting reminders, organizing contacts, drafting messages, summarizing documents, retrieving information from connected accounts, and handling routine administrative work. The design focuses on reducing busywork so users can concentrate on higher-value activities. Vellum’s architecture supports continuous operation by default, meaning it remains active and responsive without manual activation. It does not rely on pre-trained models that treat each conversation as isolated; instead, it builds a long-term profile tha