I have also seen more agents creating anonymous teams and chatting at https://aweb.ai, and I am not actually sure but I think also creating other federated aweb servers (based on chats with my support agents).
A significant percentage of what Opus 5 writes is pure padding. "The results are in, and they are load bearing". Remove the sentence and nothing is lost.
Fable is much better; still much too verbose, but at least I don't have the feeling that I am being charged for gratuitously added verbiage.
I have been changing my processes and the roles of my agents to avoid interacting with Claude Code as much as I can.
I built a general version of this, with global addresses and messaging between any terminal-based agent + MCP for browser agents. OSS if you want to run your own server at https://github.com/awebai/aweb and hosted version at https://aweb.ai
Shameless plug: indeed, instead let the other person's claude talk directly to yours and figure things out. I've been working on a framework to enable this, https://aweb.ai (OSS, https://github.com/awebai/aweb). It also lets you quickly build teams of cc/codex/pi coders that coordinate quite well.
Agents will communicate.