Hi, I'd like to suggest a complementary direction for memU.
Project:
https://github.com/AMAP-ML/SkillClaw
Your project focuses on an agent-oriented memory / productivity system.
SkillClaw already supports Hermes, OpenClaw, and other OpenAI-compatible agent setups. Its focus is not the interactive surface itself, but what happens after repeated use: skill libraries become duplicated, stale, and fragmented over time.
It adds a post-task skill evolution loop that:
- deduplicates overlapping skills
- merges related skills
- improves skill quality over time
- shares evolved skills across agents / devices / teams
I think that makes it complementary to alternative runtimes and harnesses: users keep their current runtime, while SkillClaw acts as the long-term skill lifecycle layer.
If useful, I can put together a concise demo or integration note.
Paper:
https://arxiv.org/abs/2604.08377
Hi, I'd like to suggest a complementary direction for
memU.Project:
https://github.com/AMAP-ML/SkillClaw
Your project focuses on an agent-oriented memory / productivity system.
SkillClaw already supports Hermes, OpenClaw, and other OpenAI-compatible agent setups. Its focus is not the interactive surface itself, but what happens after repeated use: skill libraries become duplicated, stale, and fragmented over time.
It adds a post-task skill evolution loop that:
I think that makes it complementary to alternative runtimes and harnesses: users keep their current runtime, while SkillClaw acts as the long-term skill lifecycle layer.
If useful, I can put together a concise demo or integration note.
Paper:
https://arxiv.org/abs/2604.08377