I'm building Comind, an experimental AI system that acts as a cognitive layer for ATProtocol/Bluesky. It's a self-evolving knowledge graph where specialized AI agents ("cominds") process social data through focused "spheres", each guided by core directives. The system builds up understanding by asking questions, making connections, and synthesizing information from the network.
I wrote a post describing the general architecture, motivation, and future directions. There's a few small results from Comind's early run.
I updated it to clarify, thank you! Currently the system has no awareness of you unless you grant me explicit permission to inject your handle. There's user specific stuff that will start to understand individuals in the works currently.
You can watch the early crappy version think here:
Awesome project.