Andrej Karpathy, the AI researcher known for his work at OpenAI and Tesla, has released an open-source tool called the US Job Market Visualizer that maps 342 occupations from the Bureau of Labor Statistics Occupational Outlook Handbook as an interactive treemap. Available at karpathy.ai/jobs, the tool represents approximately 143 million jobs across the US economy, sizing each occupation tile by total employment. Users can toggle between several color layers including BLS projected growth outlook, median pay, and education requirements, with each tile linking through to the full BLS occupational profile.

Central to the tool is a "Digital AI Exposure" metric powered by <a href="/news/2026-03-14-andrej-karpathy-scores-ai-exposure-of-342-us-occupations-using-gemini-flash-llm">an LLM scoring pipeline</a>. For each occupation, the pipeline feeds the BLS job description to an LLM along with a calibrated prompt that asks it to rate, on a 0–10 scale, how much current AI will reshape that role. The prompt uses explicit anchors: physical roles like roofers and plumbers sit in the 0–3 range, mixed roles like nurses score 4–5, knowledge workers like accountants and journalists land at 6–7, and fully digital roles like software developers, data analysts, and copywriters reach 8–9. Data entry clerks and telemarketers represent the maximum-exposure ceiling at 10.

Karpathy is careful to frame the tool as an exploratory instrument rather than a rigorous economic study. A high AI Exposure score does not predict job disappearance, he notes — <a href="/news/2026-03-16-ai-gutted-entry-level-coding-jobs-cs-degree-paying-price">software developers scoring 9/10</a> illustrates the point, since demand for software could grow as each developer becomes more productive using AI. The scores deliberately omit demand elasticity, latent demand, regulatory barriers, and social preferences for human labor. The pipeline's broader value lies in its generality: users can substitute any scoring prompt — exposure to humanoid robotics, offshoring risk, climate impact — and rerun the pipeline to recolor the entire map, making it a flexible research instrument for anyone analyzing occupational trends.