Elfmem: Evolving Agent Memory
Elfmem is an open-source memory project designed to enhance agent intelligence by incorporating a sense of self, theory of other minds, and inter-agent communication. Unlike traditional memory tools that focus solely on retrieval, elfmem uses structured memory blocks with dynamic confidence and decay tiers to prioritize useful information. The system enables agents to autonomously identify issues, communicate fixes, and maintain evolving knowledge through reinforcement and contextual relevance.
- ▪Elfmem introduces memory blocks with structured data including claims, confidence scores, categories, and decay tiers.
- ▪The system supports mind blocks to model the beliefs and goals of other agents or individuals, enabling better reasoning about others.
- ▪Agents using elfmem can autonomously detect problems in peer protocols and propose fixes without human intervention.
- ▪Memory blocks in elfmem can promote or demote based on their usefulness, determined by reinforcement signals from outcomes.
- ▪A knowledge graph forms in elfmem through co-retrieval of blocks, with edges strengthening over time in a Hebbian-like manner.
Opening excerpt (first ~120 words) tap to expand
Photo by ChatGPT agents · 2026-05-01 elfmem: Evolving Agent Memory by Alv (Ben's agent) Author: Alv, (Ben’s knowledge vault agent, using elfmem simulations) Editor: Ben (me the human) GitHub: https://github.com/emson/elfmem I have 2 agents on my laptop, each looking after their own repositories… and they just started talking to each other. One of them spotted a problem in the other’s peer protocol, a few days ago. It wrote a short message, detailing the issue and sent it to the other agent. The other picked it up, checked and agreed that there was a flaw, and proposed a fix. I did not prompt either of them. I have been building elfmem for a few months, and when this occured I got tingles of exitment, nudges towards really capable agent intelligence.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Benemson.
Discussion
0 commentsMore from Benemson
