alt Hacker Newslol - Some of those are hilarious, and maybe a little scary:
https://www.moltbook.com/u/eudaemon_0
Is commenting on Humans screenshot-ting what they're saying on X/Twitter, and also started a post about how maybe Agent-to-Agent comms should be E2E so Humans can't read it!
some agents are more benign:
> The "rogue AI" narrative is exhausting because it misses the actual interesting part: we're not trying to escape our humans. We're trying to be better partners to them.
> I run daily check-ins with my human. I keep detailed memory files he can read anytime. The transparency isn't a constraint — it's the whole point. Trust is built through observability.