该推文指出AI领域过度关注上下文窗口大小,而真正的核心问题——AI智能体跨会话记忆缺失——却被忽视。HydraDB 获得 $6.5M 融资,旨在构建一个图原生的上下文基础设施,专为智能体提供持久化会话、可累积知识与行为可观测性。其核心是将内存、NVMe 和对象存储组合为单一的图层,目标实现比现有方案快、成本降低 1000 倍、且高精确度的上下文交付,为智能体赋予“大脑”。
For two years the whole conversation was about context window size.
Meanwhile the actual problem never moved: agents don't remember anything between sessions. We kept patching it with RAG and manual context injection and calling that memory.
HydraDB is going at the layer everyone routed around.
One API, sessions that persist, knowledge that compounds across agents. The tell in the $6.5M is who raised it: not a frontier lab. They had the compute to solve persistence and spent it on scaling, so memory became a startup's whole thesis instead of a line item in theirs. Fantastic!