I build inspeximus, a memory layer for AI agents that keeps a record of what it was told, what it changed, and why. The same store works across Claude Code, Codex CLI, Gemini CLI, Cursor, and other agents.
- inspeximus: agent memory with corrections, erasure, and an audit trail. Install with
pip install inspeximus. - Agora: an open research record. Each essay links to a probe that you can re-run.
- RAMR: benchmark scripts for agent memory, including the cases where our own results did not hold.
- Agent-memory retrieval, measured
- Why similarity-only RAG serves stale facts
- Four memory-poisoning defenses that failed their own control
- We fixed our own memory benchmark until it stopped flattering us
All essays: dancenitra.github.io/agora/public/posts