Supermemory 作者称消费级助理几乎没人用图数据库做记忆
Supermemory 作者说 Openclaw、Instinct、Muse 等都用文件式记忆，找不到真实上线的图数据库记忆案例。
I am trying to find real world implementations of graph database for agent memory at scale (@supermemory is not a graph db! and supermemory runs at scale for many, many customers.) Openclaw, instinct, muse, hermes, etc. use file-based memory. [Supermemory customers] use @supermemory (temporal fact-based) Found 0 graph uses in popular consumer/enterprise so far. Feels like everyone's selling shovels here? People (devs) like it downstream because it looks and feels cool, but most teams (supermemory included) I talk to are moving closer towards the model layer and letting the model figure more things out about memory. the representation - fact that it's a graph - doesn't really matter. Curious about thoughts. drop links to talks/blogs/etc below that are NOT shovel sellers and actual use cases in production pls.