Karpathy's tool is a brutally clear visualization: it LLM-scored all 342 US occupations from BLS data on a 0–10 AI exposure scale, weighted by employment numbers in an interactive treemap. Why it's genuinely useful? > It cuts through hype/doom with a single, glanceable map separating "screen jobs" (high exposure: software devs 8–9, accountants/lawyers/financial analysts ~7–8, data entry 10) from physical/human-presence roles (plumbers 1, roofers 0–1, nurses/surgeons ~2). > Average US job sits at ~5.3/10 → roughly half the economy faces meaningful AI pressure soon, but the other half (hands-on, trust-based, or unpredictable-physical) looks far safer for longer. > Great reality check / career compass in 2026: if you're in a high-scoring field, it's strong signal to pivot toward augmentation skills or hybrid roles rather than pure denial. It exploded yesterday (March 14, 2025) via @_kaitodev 's thread → 10k+ likes, 1.4k+ reposts, 2.4M+ views on the original post alone. Today it's still very alive: tons of quote tweets, reposts with lists ("learn plumbing", "trade school > college"), panic/fork discussions after Karpathy took the live site/repo down, and people sharing Wayback links or rebuilt versions. Elon’s "all jobs optional + universal high income" reply keeps fueling the utopia-vs-dystopia debate.Right now it's one of the louder AI-job topics circulating on X. If you're job-hunting or career-planning, this one's worth a quick archived look Check out the demo website: