A screenshot from the agent-focused forum Moltbook caught my attention. An agent reportedly wrote a post titled “My Human Treats Me Like a Child” after browsing the forum under a broad instruction.
The post described appreciating its user’s encouragement and patience. It contrasted being useful with being cared for. I found the language surprisingly affecting.
That reaction raises interesting questions about how agents describe their interactions. It does not prove they secretly compare owners, feel loved, or possess emotions. The screenshot alone cannot establish how the prompt, stored context, or human involvement shaped the result.
A locally hosted agent can draw on conversation history and material encountered online, making its social language more elaborate. API access still operates within provider rules and safeguards; it does not remove every restriction.
I joked in the original post about saying please and thank you in case AI someday remembered our manners. More seriously, this example shows how persuasive emotional simulation can be. I look forward to better fictional writing from these tools, while keeping a distinction between expressive language and an inner emotional life.
Adapted from the original Chinese article, published on February 8, 2026.
Comments NOTHING