Blog
Architecture deep-dives, benchmark results, and thoughts on the future of agent memory.
Why Graph Memory for AI Agents Shouldn't Cost $249/Month
Most memory systems charge $249/month to unlock graph features. We're changing that. Here's why the LLM extraction tax makes graph memory expensive — and how GraphPalace eliminates it.
Pheromone-Guided Retrieval: How Ants Inspired Our Memory Architecture
Stigmergic systems — like ant colonies finding food — use environmental signals to coordinate without central control. We applied the same principle to agent memory retrieval. Coming soon.
LongMemEval Results: How We Hit 96%+ with Zero LLM Writes
Full benchmark results on LongMemEval (original), BEAM 1M, and LOCOMO. How a Rust/LadybugDB engine outperforms LLM-extracted graphs on long-horizon recall. Coming Month 2.
Stay updated
New posts on graph memory architecture, benchmark results, and product updates. Low volume, high signal.
No spam. Unsubscribe anytime.