The measurements behind MemBrain: a synthetic benchmark of lookalike memories, the harness that ran every arm on it, and the frozen results.
-
Updated
Nov 13, 2026 - Python
The measurements behind MemBrain: a synthetic benchmark of lookalike memories, the harness that ran every arm on it, and the frozen results.
Full-stack AI tutoring app with knowledge graph visualisation, persistent memory via MemBrain, and model-agnostic LLM routing via OpenRouter. Split deployment: Streamlit on Vercel + FastAPI on Render.
MemBrain: long-term memory for Claude Code. Lessons from your past sessions, recalled per prompt behind a fitted gate. A different approach form the classical RAG: instead of trusting top-k similarity, it decides at write time which memories compete, and stays silent unless one clearly fits.
A simulator for memory circuits on connectome wiring: one measured result on the fly mushroom body, and a hippocampal-loop model built from published rodent statistics. Part of MemBrain.
To associate your repository with the membrain topic, visit your repo's landing page and select "manage topics."