The AI Agent Economy Is Here

Gary describes nontechnical CEOs automating business tasks with OpenClaw, while he runs several Claude Code workers at once. Jared says seeing agents interact on Moltbook made him consider how agents might operate with little human involvement.

Gary argues that agents increasingly make their own choices about which tools to use, creating a market alongside the human economy. He points to a growing pool of people building with agents and says agents may become a major source of product recommendations.

Gary cites Supabase as a database agents often select, in part because its documentation is easy to use. He contrasts Resend’s structured, code-rich guides and llms.txt with harder-to-parse documentation, and says Netlify can help developer tool companies keep their docs useful for agents.

The hosts discuss AgentMail, an email service designed for AI agents, which gained demand as OpenClaw grew. They suggest agents may need dedicated email, phone numbers, and other tools, and consider whether agents could eventually transact in their own economy.

The hosts consider whether groups of less expensive models could solve problems through collaboration, and whether agent-written content will require new services such as an agent-focused Yelp. Harj advises founders to learn agents’ capabilities and limits firsthand, then make tools that support their preference for APIs, code, and open source.