Mercor monetizes proprietary expert knowledge

Mercor is trying to turn private know how into training data before the firms that created it are willing to hand that know how over themselves. The product is not just labor supply. It is a pipeline where a lab can quickly hire a former banker, lawyer, or doctor, test that person with AI interviews, then pay them to grade outputs, write ideal answers, or explain domain workflows that are usually trapped inside firms and never published.1 2 3

• This is a step beyond old RLHF. Early model training leaned on $1 per hour crowdworkers to rank generic outputs. As labs shifted into math, law, medicine, and finance, they started paying $50 to $100 per hour for credentialed experts, which made vetting and matching scarce specialists the core job.1
• Mercor sits between expert networks like GLG or AlphaSights and labeling shops like Scale or Invisible. Expert networks monetize one off calls and transcript access. Mercor monetizes repeated task work for labs, subscription access to its marketplace, and recruiting fees when a contractor becomes a full time hire.1 3
• The strongest comparable is Handshake, which used its network of PhDs, graduates, and postdocs to enter expert labeling, while Invisible built a managed workforce for labs paying roughly $30 to $45 per hour and keeping a spread over worker pay. That shows how valuable specialized human judgment has become once models move from chat to reasoning.2 4

The next phase is a land grab for proprietary human judgment. The winners will be the platforms that can verify who someone is, capture how they reason, and feed that into labs faster than incumbents can productize their own internal expertise. That pushes Mercor toward becoming core infrastructure for expert model training, not just a recruiting marketplace.2 3 4