Karpathy rates 342 US jobs on AI exposure

Karpathy accidentally dropped the most viral labor economics analysis of the year in a Saturday morning side project, then panicked and nuked the repo — a move that only amplified the spread.

• –342 occupations scored 0–10 using Gemini Flash against BLS task data; medical transcriptionists hit 10/10, software developers land 8–9
• –High earners face the highest exposure — the data shows a clear positive correlation between salary and AI risk, inverting the usual automation narrative that low-wage workers are most at risk
• –Karpathy deleted the repo likely to prevent forks that could be misrepresented; the website stayed live, suggesting controlled presentation was the goal
• –Elon Musk's "all jobs will be optional" response immediately followed, turning a data visualization into a culture war flashpoint
• –The project itself is a demonstration of the thesis: a two-hour vibe-coded LLM workflow produced a study that would have taken a research team weeks