agent swarm 用 Jev 做快速反应层

Safi @safishamsii原推发布于 2026.10.10

在 X 上打开原推

This is how agent swarms should actually work. Each agent gets its own Graphify brain. Persistent memory, direct communication, zero manual context sharing. No more messenger boy. The agents just remember and keep building. Layer in jev from @typesafeai + or any other System 1 models for the fast reactive layer and the whole swarm compounds.

这条是引用着另一条写的：x.com/daospora_net/status/2108574569036496997