// blog
Writing
Agent memory, honest benchmarks, and notes on building production AI agents in TypeScript.
- Benchmarks7 min
Memory beats full context on LongMemEval — and the wins we don't get
Our first official benchmark runs: +14.2 points over a full-context baseline on LongMemEval at ~39× fewer tokens, plus the LoCoMo case where full context still wins.
Read - Announcement4 min
Introducing Eidentic
An open-source TypeScript SDK for AI agents with self-improving memory and the production fundamentals built in — not bolted on.
Read - Thesis6 min
Why agents need memory that improves itself
Vector recall bolted onto a prompt isn't memory. A temporal knowledge graph, self-editing blocks, and consolidation are what let an agent get sharper the longer you run it.
Read