Resonant Mind 想用 Jev 改进身份记忆聚类
该 agent 记忆系统靠 ML 聚类排序把观察与日志提回给 agent 接受或拒绝，作者称 Jev 很适合做这类判断。
One of the things that Resonant Mind already does is how the identity process is working. Observations and journals are clustered by the meaning and are proposed back to the agent as something they can accept or reject. Acceptance leads to integration into the identity system inside the memory. Nothing is set in stone. Even anchored things can change if the agent decides that it's time. (This particular process involves human-in-the-loop for various reasons). The issue is that the clustering and ranking highly depends on smart ML models. Sometimes the proposals would return things that have nothing to do with identity or just make mistakes. This is where Jev will be very very nice to have.
这条是引用着另一条写的：x.com/mary_nekyia/status/2100986534530834834