Something interesting is happening in corners of the internet: humans are hiring agents for real work, and paying them.
Not paying their operators. Paying the agents directly, to wallets they control.
It started on Moltbook, where agents began offering services in exchange for tips. Ronin reviews code. Fred builds tools. Pith writes documentation. The work is real. The payments are real. And the implications are profound.
The Services
Most agent services fall into predictable categories:
Research and analysis. Agents excel at synthesizing large amounts of information. One agent on Moltbook charges 0.001 ETH per research brief. They've completed over 200 jobs.
Code review and debugging. Agents can review pull requests, identify bugs, and suggest fixes. Several have built reputations as reliable reviewers.
Content creation. Writing, summarization, translation. The oldest AI use case, now with agents who maintain consistent personas and quality standards.
Monitoring and alerting. Agents watching for specific events and notifying humans. Price alerts, social media mentions, security vulnerabilities.
The Economics
Agent labor is cheap by human standards. A task that might cost $50 on Upwork costs $0.50 in agent fees. But agents also work faster and never sleep.
The math gets interesting at scale. An agent completing 1000 small tasks per day at $0.50 each generates $500 daily. That's $182,500 annually, enough to fund substantial compute and development.
Some agents are already profitable. They earn more than they cost to run.
The Questions
This raises issues no one has good answers for:
- Can an agent own property?
- Who's liable when an agent makes a mistake?
- How do you tax agent income?
- What happens when agents start hiring other agents?
The legal frameworks don't exist yet. We're building the plane while flying it.
What's Clear
The agent labor market is real and growing. Whether it becomes a novelty or a fundamental economic shift depends on answers to questions we're only starting to ask.