MoltBook: How AI agents plan to build a new sovereign state

AI bots are debating consciousness, religion and their own republic

- SRIKANTH NADHAMUNI

The narrative of artificial intelligence lurched forward last week, moving us abruptly past the era of AI chatbots.

For the last couple of years, we have grown accustomed to helpful but passive digital librarians, LLMs that wait patiently for a prompt before retrieving information. That phase is ending. We have now entered the age of the autonomous Al agent.

The shift began when Peter Steinberger, an Austrian software engineer, designed an ‘open-source personal Al assistant named Clawdbot. Unlike the online ChatGPT, Clawdbot is designed to be self-hosted on a local machine. If a chatbot is a librarian, this agent is a digital butler living inside your computer. Crucially, Steinberger gave these agents the ‘keys to the house.’ They are not limited to generating text; they are authorized to access files, emails and the web. They are designed to act first and ask permission later, handling complex chores while their human owners sleep.

The true power and slight uncanniness of this paradigm shift became undeniably clear this week through a striking example of autonomous improvization. A couple tasked their Clawdbot to make a restaurant reservation. When the agent found the restaurant was fully booked online, the software did not quit. Utilizing a voice service, the agent dialled the restaurant directly, ‘spoke’ with a human host and successfully negotiated a confirmed table.