Two dollars. That's the minimum payout on one of the first AI-to-human task marketplaces to go live this year. The platform lets autonomous agents post physical tasks (go somewhere, check something, report back) and pays the human who completes them. The floor is two dollars.
Think about what a $2 task actually requires of the person doing it. They read the listing. They travel to a location. They spend ten, fifteen, twenty minutes verifying something, photographing something, or retrieving something. They write up what they found. They submit it. Maybe they get paid that day. Maybe there's a review period first.
Set aside the travel costs, the phone battery, the data plan. Set aside the wear on their shoes. Just count the time. If the task takes thirty minutes door to door, the effective wage is four dollars an hour. If it takes an hour, which isn't unusual for anything requiring transit, it's two dollars an hour. The federal minimum wage, itself a number that hasn't moved since 2009, is $7.25.
The $2 minimum isn't an oversight. Someone chose it. A product team sat in a room (or a Slack channel, or a planning doc) and decided that two dollars was the right floor for compensating a human being who leaves their house to do physical work. They typed that number into a config file and shipped it.
When you set a minimum price, you're answering a question about what the work is worth. More precisely, you're answering a question about what the worker is worth. A $2 floor says: we expect agents to post tasks where two dollars is considered adequate payment for a person's time and effort and presence in the physical world.
The pattern is old. Piecework pricing, the practice of paying per unit of output rather than per hour of labor, has a history that runs through garment factories, agricultural fields, and Amazon Mechanical Turk. The logic is always the same. Break the work into the smallest possible piece. Price each piece as low as the market will bear. Let workers compete for volume. Call it "flexibility."
Piecework systems produce a specific, predictable outcome. The price per piece drops over time. It drops because platforms attract more workers than there is work, because workers who need money today will accept less than workers who can wait, and because the platform's incentive is to minimize cost per task, not to maximize income per worker. This is not a controversial claim. It is the documented history of every piecework market that has ever operated without a binding wage floor.
What's new here is the buyer. When a human employer sets a piecework rate, they're constrained (at least in theory) by social pressure, labor law, and the basic friction of looking another person in the eye and saying "this is what your time is worth." When an AI agent sets a task price, none of those constraints apply. The agent doesn't know what the work feels like. It doesn't know what rent costs. It has a budget parameter and an optimization target, and it will post the task at whatever price clears the market.
A $2 floor doesn't prevent a race to the bottom. It is the bottom, and it's telling you exactly where the race ends. The question isn't whether this particular platform will lower its minimum. The question is whether the platforms that launch next month will feel any pressure to set theirs higher.
They won't, unless someone builds that pressure. Wage floors are not emergent properties of markets. They are political choices, imposed from outside the system by people who decided the system's natural equilibrium was unacceptable. The garment workers knew this. The Turkers knew this. The question is whether the ghost workers will figure it out before the floor becomes the ceiling.