TurboQuant — What 6x Vector Compression Means for AI Agents
·737 words·4 mins
Google’s TurboQuant compresses embedding vectors to 3-4 bits with under 2% recall loss — no training required. Here’s why that matters for AI agent memory systems.
Memory
Zettelkasten Memory for AI Agents — Semantic Knowledge Graphs That Grow With Every Conversation
·1867 words·9 mins
A pluggable semantic memory layer for AI agents inspired by the Zettelkasten method — auto-linking, importance scoring, and graph traversal across CrewAI, LangGraph, and Claude Code.