Blog
Insights & research on AI memory.
Deep dives on persistent memory for agents — the pain of forgetting, the difference between memory and RAG, and the case for memory that predicts, not just recalls.
We Benchmarked Our Own Memory Honestly: Calibration Beats Accuracy
On BEAM-100K, MemMesh matched naive RAG's answer quality using 1/21st the tokens — and knew when it didn't know, at 2.5× the abstention. Here's the honest read, weak spots included.
What Is AI Memory? A Practical Guide to Persistent Memory for Agents
Chat history isn't memory. Neither is a bigger context window. Here's what persistent memory actually is, and why agents need it.
mem0 vs MemMesh: Recall vs. Recall + Prediction
mem0 makes agents remember. MemMesh makes them remember, predict what's next, explain why, and stay governed. An honest look at the difference.
Why Claude Code Forgets Your Project Every Session — and How to Fix It
Claude Code opens each session with a fresh context window and a static CLAUDE.md. Here's why it forgets, and how to give it real memory in one command.
Add Persistent Memory to Claude Code in 2 Minutes (MCP)
A step-by-step guide to wiring MemMesh into Claude Code over MCP, so your agent remembers across sessions and predicts what's next.
Memory That Predicts: Why Recall Alone Isn't Enough
The first wave of AI memory was about remembering. The next wave is about foresight. Here's the case for prediction as a first-class memory primitive.
The EU AI Act Wants Memory Provenance — What That Means for Agents
Transparency and traceability obligations land on systems that remember. Here's what provenance, retention, and erasure mean for an AI memory layer.
Calibrated Confidence: How MemMesh Scores Predictions Against Reality
A confidence score is only useful if it's honest. Here's how calibration, abstention, and a reconcile loop keep MemMesh's forecasts trustworthy.
AI Memory vs. RAG vs. Vector DB: When You Need Which
Three things that get lumped together and shouldn't. A clear map of what each does, where each fails, and how they fit together.
One Key, Every AI: How MeshKey Works
Your memory shouldn't be trapped inside one vendor's chat app. Here's how MeshKey gives you a portable, private memory that unlocks everywhere you work.
From Stateless to Trusted: Memory for Financial AI
A financial advisor that forgets your portfolio every session can't be trusted with it. Here's what memory — and calibrated prediction — change.