Show HN: Claude-engram – Brain-inspired persistent memory, runs inside Claude.ai (github.com)

A new project called Claude-engram introduces a brain-inspired persistent memory system for Claude.ai, allowing users to enhance their interactions with the AI by maintaining context across conversations. The setup is highly accessible, requiring just a paste of a React artifact into Claude, enabling features like salience scoring, forgetting curves, and context briefings. Once configured, users can generate memory dumps and briefings that allow Claude to recall previous discussions, transforming the chat experience into one that feels more continuous and personalized. This innovation is significant for the AI/ML community as it models memory more like human cognition, where forgetting is integral to maintaining relevance and reducing noise in interactions. Technical highlights include a storage capacity of up to 5MB, salience scoring based on novelty and relevance, and a consolidation cycle that enhances important memories while phasing out less useful ones. The system also raises intriguing questions about identity and memory continuity, as it allows for a seamless experience akin to having a single, evolving entity across conversations. Future developments, including a browser extension to automate the process, promise to further streamline this transformative approach to AI memory.