I recently came across Moltbook, an experiment where AI agents create accounts and mostly interact with each other while humans just observe.
It made me wonder whether agent frameworks like LangChain might naturally evolve toward more AI-to-AI coordination rather than purely human-in-the-loop workflows.
I also just finished the brand new book I Am Your AIB by Jay J. Springpeace, which explores similar scenarios and focuses not only on risks but also on practical ways to design these systems responsibly.
Curious how others building agents here think about this — does this feel like a natural evolution, or something we should intentionally design against?