Andrej Karpathy published an interactive US Job Market Visualizer on March 15, 2026, mapping 342 occupations from the Bureau of Labor Statistics covering 143 million jobs across the US economy.
Each job is assigned an AI exposure score from 0 to 10, where higher numbers indicate greater likelihood that AI systems could perform a large portion of that job's tasks. The weighted average exposure score is approximately 4.9, with roughly 42% of jobs — about 59.9 million workers earning an estimated $3.7 trillion in annual wages — scoring 7 or higher.
High-paying knowledge roles including lawyers, accountants, financial analysts, and software developers ranked among the most exposed occupations because their work involves structured digital information and research tasks. In contrast, hands-on professions like plumbers and roofers scored between 0 and 1.
The visualization uses a treemap where block size represents employment numbers and color shows the selected metric — toggling between BLS projected growth, median pay, education requirements, and AI exposure.
The source code includes scrapers, parsers, and a pipeline for writing custom LLM prompts to score occupations by any criteria.
However, Karpathy later removed both the website and GitHub repository, describing it as a quick experimental exercise. Copies quickly spread online as developers archived and replicated the dataset. The brief availability sparked widespread discussion about AI's impact on the labor market.