Reading up on Letta

Structured stores beat file memory on accuracy and token cost; files win only for small memories or 'I don’t know' answers, and trained experience helps weak models only.

Agent harnesses are deeply tied to agent memory, and using a closed one cedes control of memory to a third party, creating vendor lock-in that open harnesses avoid.

Sage is an experimental, self-hosted AI assistant with persistent memory, encrypted messaging via Signal or Nostr, and confidential LLM inference inside a TEE.