Thoughts on AI, memory systems, and building in the open.
AutoMem's results on the neutral Agent Memory Benchmark — graceful BEAM scaling to 10M, ~3.9k context tokens, and the conversational-recall gap, stated plainly.
What shipped in 0.16, the four recall-tuning knobs built and deliberately left off, and why the internal benchmark numbers aren't comparable to anyone else's yet.
bymiles.ai is an agentic AI that spins up WordPress websites in minutes. Impressive. But in 2026, is WordPress still the right canvas for an AI to paint on? We compare WordPress to Starlight and make the case for a better default.
How we built rigorous evaluation infrastructure, found critical scoring bugs, and established honest baselines that beat CORE SOTA.
53 pages of production-grade documentation covering every corner of the memory system — from FalkorDB graph internals to platform-by-platform integration guides.
Why persistent memory matters more for trading bots than chat assistants, and how Bankr can add it with one command.
No bridges, no middleware. AutoMem memory recall works directly inside OpenClaw agents via a single skill file.
A beginner-friendly guide to running AutoMem locally — completely free, no coding experience required
3D visualization, 11 relationship types, hand gesture controls, and real-time monitoring. Your memories are about to get visual.
Streamable HTTP replaces SSE as the primary transport. Same features, better reliability, zero breaking changes.
How we enabled real-time memory streaming for AI platforms that don't natively support MCP
A practical guide to Retrieval Augmented Generation and why it matters for AI assistants
Why we built AutoMem and why your AI's memory should belong to you
A deep dive into AutoMem's FalkorDB + Qdrant architecture and why relationships matter
A practical guide to running AutoMem on your own infrastructure