The traditional gig economy—once defined by driving cars or delivering groceries—is undergoing a high-stakes transformation. As the job market faces a period of cooling and uncertainty, a new generation of white-collar professionals is turning to an unexpected source of income: train artificial intelligence (AI) how to perform their own jobs via Mercor.

Mercor is paying experts $250/hour to train AI to do your work

At the center of this shift is Mercor, a Silicon Valley startup valued at $10 billion. The company has become a massive hub for specialists across almost every field imaginable. Their roster doesn’t just include traditional coders. The active recruiting list includes astronomers, poets, comedians, and even investment bankers. These experts spend their hours reviewing, critiquing, and refining the outputs of large language models for major players like OpenAI and Anthropic.

For a lot of people, the appeal is just money. In today’s job market, traditional full-time jobs seem to be getting harder to find. On the other hand, the hourly rates for AI training can be surprisingly high. A dermatologist could make as much as $250 an hour helping a healthcare AI improve its diagnostic suggestions. Poets, on the other hand, can make $150 an hour teaching models how to express emotions and write in a way that makes sense. There are a lot of examples like these one out here.

However, the hiring process is a far cry from a standard corporate application. Prospective trainers often find themselves being interviewed by AI proctors rather than human recruiters. Once hired, the level of oversight is intense . To prevent contractors from using AI to grade other AI, workers must often use time-tracking software that monitors their progress and ensures the “human touch” remains authentic. It sounds like the “evolution” of the classic supervision tool that some remote worker jobs use, right?

The professional dilemma: Training your AI job replacement

The sweet paycheck is undoubtedly a welcome relief for those between jobs. However, the work comes with a unique psychological weight. There is a persistent irony in the air: by making these models more accurate, workers are effectively building the tech that could eventually automate their original professions. In other words, these workers are training their AI replacements for their jobs.

Many contractors view the situation with a mix of pragmatism and unease. Some see AI as an inevitable wave that cannot be stopped. So, if they don’t help refine it, someone else will. Others find that the work has unexpected benefits, such as sharpening their own writing or technical skills through the act of critiquing the machine’s attempts.

There isn’t always a good relationship between these “human teachers for AI” and the big tech companies. Some people who have applied are worried about complicated contracts that seem to give the company a lot of control over their intellectual property.

Companies say that these terms only apply to work done on a certain project. Still, the conflict shows that there is an imbalance of power in this new job market. When a candidate tries to talk about these terms, an automated assistant will often politely but firmly say “no.” Because of this, they only have one choice: accept the terms or lose the job.

This ironic situation has become a vital lifeline for professionals who are trying to make sense of the modern workforce, whether it is a temporary bridge or a permanent new part of the economy.