OpenAI is deploying a specialized training initiative to enhance ChatGPT's capabilities in niche professional sectors, including farming, ranching, nursing, and commercial aviation. Documents obtained by Business Insider reveal the scale and methodology of this effort, known internally as Project Stagecraft. The project involves hiring approximately 3,000 to 4,000 freelancers through Handshake AI, a data-labeling startup based in San Francisco. Contractors are compensated at rates of at least $50 per hour, with some expert roles paying up to $500 per hour, to create high-quality training data that reflects real-world workflows. The core objective of Project Stagecraft is to move AI capabilities beyond generalist functions like coding or tutoring and into specialized knowledge work. Unlike earlier data labeling tasks that focused on basic filtering or ranking, this initiative requires participants to possess specific industry expertise or advanced degrees. Contractors are instructed to develop professional personas, such as a nurse practitioner or a soil scientist, and generate detailed prompts that simulate complex job tasks. For example, a contractor might ask the AI to act as a financial manager for a theater, using a collective bargaining agreement to create a payroll spreadsheet for touring musicians. The goal is to ensure the model learns from genuine human expertise, allowing it to respond more accurately and ethically to professional requests. Handshake AI manages the project by employing a rigorous multi-step review process. Materials created by freelancers undergo two rounds of internal evaluation, including a final check by an industry expert to verify job-specific details. OpenAI conducts a third review to ensure the data meets its quality standards. According to the company's training manual, this process helps map economically relevant tasks and evaluates the model's ability to handle professional scenarios without simplifying the level of detail based on assumptions about what an AI might need. The project has expanded rapidly as the AI industry shifts toward requiring deeper domain knowledge. Handshake AI, originally a job platform for young professionals, has transitioned into a key player in the data-labeling sector, serving major tech companies including Meta. The spreadsheet of occupations obtained by reporters listed a diverse range of fields, from pharmacists and mechanical engineering technicians to fine artists and commercial pilots. Despite the high pay and technical precision required, the project has not been without controversy. Previous reports indicate that Handshake AI has withheld payments from some contractors, sometimes amounting to thousands of dollars, citing violations of platform rules. When questioned about these payment disputes, a Handshake representative declined to comment, and OpenAI did not respond to requests for clarification. While the initiative aims to produce a more capable and reliable AI, it has also raised concerns among workers. One contractor working on the project noted the irony of the situation, stating, We all were aware that we were basically training AI to replace us. As OpenAI continues to refine its models through such specialized data collection, the technology industry faces ongoing debates regarding the ethical implications of automating professions and the rights of the human workforce powering these systems.