Andrej Karpathy published an interactive map of 342 professions' vulnerability to AI automation (based on US Bureau of Labor Statistics data). Each profession has a score from 0 to 10 (higher score = higher replacement risk). The economy-wide average is 4.9.
The main obvious trend: the more you work behind a screen, the higher the risk that AI will replace you. The more manual work — the lower.
Maximum risk (8–10): medical transcriptionists (10), programmers (8–9), data analysts (9), legal assistants, financial analysts.
Minimum risk (0–1): roofers, plumbers, electricians, construction workers, gardeners, nurses, and caregivers.
An interesting inversion: professions with $100K+ salaries scored an average of 6.7, while those under $35K scored only 3.4. The classic logic of "more education = more protection from automation" no longer holds.
Karpathy later deleted the repository from GitHub. Explained: "It was a two-hour Saturday vibe-code project." But he left the visualization site up, and users had already forked everything.
→ karpathy.ai/jobs