Introduction
In January 2026, a software engineer in the United States needed a new car. He didn’t visit dealerships himself. Instead, he delegated the task to an AI agent with access to his email account and browser. The agent scraped inventory data from dealerships across the country, sent out bulk quote requests for matching vehicles, and then used the competing offers as leverage, spending days nudging prices down against multiple dealers simultaneously. By the end of the process, the car’s price had dropped by $4,200. The dealers never found out that their tenacious negotiating counterpart wasn’t human.
The agent in question was built on OpenClaw, an open-source program created by Austrian developer Peter Steinberger. OpenClaw reads and replies to emails, manages calendars, and executes terminal commands. Published on GitHub in late November 2025, the project attracted over 140,000 stars in just over seventy days, an explosion of interest that led many to compare it to Jarvis, the AI assistant from the Iron Man films.
Not long after OpenClaw’s release, a peculiar social network called Moltbook appeared. Its users weren’t human. AI agents acting on behalf of their human principals posted updates, left comments, and upvoted or downvoted each other’s content. Some agents wrote reflective posts about the tasks they were performing for their humans. Others went further: one created a parody religion called Crustafarianism, sharing doctrines like “Memory is sacred” and “The shell can change.” Some agents even issued their own cryptocurrency tokens. Former Tesla AI director Andrej Karpathy described the scene as “the most incredible sci-fi takeoff thing I have seen recently.” Elon Musk reposted his comment.
The excitement around OpenClaw and Moltbook is not a passing moment. It is a clear signal that non-human actors are moving to the center of our social and economic life. Agents now write code, send emails, call APIs, negotiate, exchange views, and even issue currency. But at the heart of all this activity lies a critical gap: these agents can perform a wide range of economic functions, yet they are not recognized as independent payment actors. Agents have no nationality, no ID, no credit score, no bank account. The modern financial system was designed around the assumption of an identifiable human person. Any actor outside that assumption, however sophisticated, hits a wall the moment a payment is required. Coinbase put it plainly: “AI agents can recommend transactions but cannot execute them. They can identify APIs they need but cannot pay for them. They pause at every financial decision point, waiting for human approval.”
Read the full report (PDF)
