Samantha Whitmore is an engineering leader and startup founder building cloud-based coding agents at Cursor, now part of SpaceX. A former engineering chief at Kensho and Maximus, co-founder of New Computer, and early LangChain contributor, she develops AI systems that retain context, adapt to their surroundings, and operate inside real working environments.
Whitmore studied physics and mathematics at Harvard and earned a master’s degree from the University of Cambridge. At Kensho, she built the company’s scoring engine and led engineering for financial-analysis products incorporating natural-language processing. She later became chief operating officer and head of engineering at Maximus, where a 2021 partnership with IMAX advanced AI-powered video enhancement and high-resolution upscaling.
Whitmore became LangChain’s first major contributor beyond founder Harrison Chase, helping develop memory for language-model applications. She subsequently co-founded New Computer with designer Jason Yuan and served as chief executive. Their product Dot combined long-term memory, voice interaction, and emotional context to create a personal AI guide; their later project Lorekeeper explored how an agent could understand relationships within a community and make useful introductions.
At Cursor, Whitmore co-authored the company’s work on cloud-agent development environments, including multi-repository workspaces, configuration as code, isolated secrets, network controls, audit logs, and rollbacks. Following Cursor’s acquisition by SpaceX, she said she would continue building with both teams.
- Memory as practical infrastructure. Useful agents need durable knowledge about prior interactions, relationships, and tasks; that context determines what they should reveal, withhold, or do next.
- Interfaces that respond to circumstance. Whitmore’s intelligent-interface demonstration combined camera-based pose detection, an LLM-managed world-state representation, and automatic switching between keyboard, text, and voice.
- Agents need complete working environments. Coding agents become more capable when they can inspect related repositories, install dependencies, run tests, and access approved services within enforceable security boundaries.
- Persistent, collaborative AI entities. Whitmore has championed agents that retain context across threads, operate in authenticated cloud environments, and coordinate with human teammates and other agents.
