Memories.ai has introduced Luci, a personal AI assistant designed to record, index, and retrieve daily information directly across a user’s ecosystem of devices. Operating natively on hardware powered by Snapdragon platforms, the system bridges PCs, smartphones, and smart glasses to maintain a continuous, private memory system without relying heavily on cloud storage.
Cross-Device Intelligence Powered by Snapdragon
Luci operates across a broad ecosystem of devices running on Snapdragon platforms, creating a unified and private personal memory network. According to Memories.ai founder Shawn Shen, the integration at the hardware level is what makes this capability possible. Snapdragon processors allow Luci to understand and remember information locally, utilizing the cloud primarily for contextual synchronization between devices rather than storing or processing raw personal data.
The technology expands on desktop-based observation tools demonstrated earlier, linking moments captured by smart glasses, notes taken on mobile devices, and documents opened on computers. This synchronization lets users retrieve information regardless of which device they are using.
On-Device Processing Architecture
The system relies on specialized on-device components to handle data efficiently. OmniCaptioner translates visual and auditory inputs captured by devices into concise, structured descriptions. Subsequently, OmniRetriever indexes these descriptions to build a searchable personal memory database.
By performing deep image analysis at the point of data acquisition, downstream processing requirements remain low. This design enables indexing, retrieval, and response generation to run directly on the Qualcomm Hexagon NPU using models scaled to hundreds of millions of parameters rather than billions. The continuous memory record requires only a few gigabytes of storage, and the memory index remains permanently on the device.
Privacy and the Future of Personal AI
Vinesh Sukumar, Vice President of Product Management at Qualcomm Technologies, noted that next-generation AI requires a focus on high performance, personal relevance, and privacy protection. Luci demonstrates how running AI workloads directly on local hardware allows users to maintain control over their personal data while benefiting from contextual assistance.
Worth a look