This article digs into a thought-provoking, data-heavy chart put together by Andrej Karpathy, who used to help run OpenAI and led Tesla’s AI division. He tapped into U.S. Bureau of Labor Statistics data, rating jobs from 0 to 10 for how exposed they are to automation, then stacked those scores up against income levels.
The chart, which disappeared pretty quickly after people started debating what it really meant, has kicked up fresh arguments. Folks are wondering—again—which jobs are actually threatened by AI and automation, and what all this means for regular workers and the economy at large.
What the chart reveals about automation exposure
Karpathy’s project gave each occupation a weighted “exposure” score, all based on official labor stats. The average exposure landed at 4.9 out of 10. But here’s the kicker: jobs making over $100,000 a year averaged 6.7, while jobs under $35,000 sat at just 3.4.
High-skill roles—think software developers, programmers, data scientists, mathematicians, financial analysts, paralegals, writers, editors, graphic designers, and market researchers—clustered close to 9 on the exposure scale. That’s become a talking point for anyone worried that AI is coming for the white-collar, well-paid crowd next.
Karpathy’s made it clear this was a two-hour side project, not a polished academic study. He pulled the chart after it got way more attention than he expected and started getting misread.
There’s no official methodology or deep analysis here, so it’s more of a conversation starter than a crystal ball. Still, it lines up with a lot of recent research hinting that educated, higher-earning workers might feel unique pressures as AI keeps advancing.
Limitations and potential misreadings
The chart’s got some quirks and blind spots. First off, “exposure” doesn’t mean you’ll lose your job tomorrow. It’s a mix of how easy it is to automate, how complex the tasks are, and whether AI tools are even available for the work.
Other things muddy the waters—like hiring trends, weird economic swings, or sudden changes in demand for tech skills. Even within the same job title, some people do work that’s way easier to automate than others, just depending on what their day-to-day looks like.
Implications for workers, employers, and policy
So, what does this all add up to? Some skilled, high-paying jobs might have more automation exposure, but that doesn’t mean robots are about to take over en masse. It does mean people need to get ready for AI to become a bigger part of their work lives, especially in technical and analytical fields.
Companies should think about upskilling, reworking jobs, and managing change thoughtfully, instead of just reaching for layoffs whenever automation pops up.
What can workers actually do to stay ahead?
- Upskill continuously: Try out new training in AI basics, data skills, or software that works alongside automation.
- Develop task-level adaptability: Focus on problem-solving, creativity, and people skills—stuff that’s still tough to automate.
- Engage with cross-functional roles: Step into roles that overlap with other areas, where AI can help you do more.
- Monitor labor-market signals: Keep an eye on what’s hot in your field and where AI is making waves, so you’re not caught off guard.
Industry responses and counterpoints
Industry voices have pushed back on some of the more alarmist takes. Citadel Securities, for example, suggested that AI-driven narratives need to be balanced with actual hiring activity and broader economic signals.
They pointed out an 11% year-over-year jump in Indeed postings for software engineers in early 2026. The company also mentioned that daily use of generative AI at work has stabilized.
Citadel highlighted other trends, like new business formation and local data-center construction. These developments, they argued, are fueling demand for technical workers and related infrastructure.
Some analysts are looking beyond just demand. They say that if automation really takes off, the rising cost of computing power could slow down how quickly machines replace people.
Basically, the economics of AI adoption hinge on the balance between what the tools can actually do, hardware costs, energy needs, and how much productivity organizations gain. There’s also this Anthropics-style distinction—what AI can technically do versus what it’s really being used for right now. That difference keeps coming up in these debates.
Here is the source article for this story: An OpenAI cofounder ‘vibe coded’ an analysis of the U.S. labor market’s exposure to AI, and the highest-paying jobs have the worst scores