DeltaMemory: A Persistent Memory Layer for AI Agents That Learns Over Time
By
Shreejit Sen
Hard to chew. Probably not worth the jaw work.
Summary
DeltaMemory is a new AI memory layer designed to solve the problem of AI agents forgetting information between sessions. Unlike vector databases or RAG systems, it extracts facts, builds knowledge graphs, and learns over time. The product is built in Rust, claims to be #1 on LoCoMo, 2x faster than alternatives like Mem0, and 97% cheaper at scale.
Key quotes
· 3 pulledAI agents are getting smarter, but they still forget everything between sessions.
Not a vector DB, not RAG, a real memory layer that extracts facts, builds a knowledge graph, and actually learns over time.
Rust native. #1 on LoCoMo. 2x faster than Mem0 and other alternatives. 97% cheaper at scale.
You might also wanna read
YourMemory: Open-source persistent memory layer for AI agents using Ebbinghaus forgetting curve
YourMemory is an open-source persistent memory layer for AI agents that implements Ebbinghaus forgetting curve decay to mimic human memory.
github.com·1mo ago
δ-mem: A Compact Online Memory Mechanism for Efficient Long-Context LLM Processing
The article presents δ-mem, a lightweight memory mechanism for large language models that augments frozen full-attention backbones with a co
arxiv.org·16d ago
DiffMem: Git-Based Differential Memory System for AI Agents
DiffMem is a proof-of-concept project that implements a git-based memory storage system for AI agents and conversational systems. It uses Ma
github.com·9mo ago
Robot Memory System Enables AI Robots to Learn from Past Experiences
Robot Memory (robotmem) is a persistent memory system for AI robots that enables robots to learn from past experiences. The system stores ep
github.com·2mo ago
Hippo-memory: Biologically-Inspired Memory System for AI Agents with Decay and Consolidation Features
Hippo-memory is a biologically-inspired memory system for AI agents that addresses the problem of AI agents forgetting everything between se
github.com·1mo ago
Letta Code: A Memory-First Coding Agent for Persistent Learning Across Sessions
Letta Code is a memory-first coding agent designed for long-lived, persistent AI agents that learn and improve across multiple sessions rath
letta.com·5mo ago