What it is
Andrej Karpathy’s US Job Market Visualizer is an interactive map of US jobs built from Bureau of Labor Statistics data. It lets you explore hundreds of occupations by job growth, pay, education level, and a rough “Digital AI Exposure” score. Bigger boxes represent bigger job categories, so you can quickly see which roles employ a lot of people and may be more exposed to AI-driven change.
Official Source Link: Click Here
Why it is interesting
What makes this useful is not just the visualization. It is the way it frames AI exposure. The basic idea is simple: if a job happens mostly on a computer, AI is more likely to change it sooner. If a job depends on physical presence, hands-on work, or real-time human interaction, it tends to have more protection.
Important caveat
This tool is not meant to be treated like a formal report or a hard forecast. The site says clearly that it is a development tool for exploring job data visually, and the AI exposure layer is based on large language model scoring. It is best used as a thinking tool, not as a final answer.
What stands out
The biggest takeaway is that AI exposure and job decline are not the same thing. Some jobs look exposed and are already shrinking. Others also look highly exposed, but are still growing fast. That matters because it suggests AI is not only removing jobs. In many cases, it is changing how the work gets done.
Jobs already under pressure
The clearest risk zone is routine digital work. Roles like medical transcriptionists, desktop publishers, bookkeeping clerks, general office clerks, and customer service representatives show both high AI exposure and negative job outlooks in the dataset. These are the kinds of jobs where the work is structured, repetitive, and already done mainly through software, which makes them easier to automate or compress.
Jobs AI will affect without killing demand
This is the more important category for most readers. Jobs like software development, data science, operations research, and information security still show strong growth, even though the model scores them as highly exposed to AI. That suggests AI may not remove these roles, but it is very likely to change the workflow inside them.
What that likely means in practice
For a lot of knowledge work, the more realistic outcome is not instant replacement. It is workflow compression. Less time spent on first drafts, formatting, repetitive analysis, basic reporting, or routine coordination. More value shifts toward judgment, strategy, taste, decision-making, and client or team communication. The job still exists, but the expectation changes. This is an inference from the scoring approach and growth patterns, not a formal forecast from the site itself.
Jobs that look more resistant
The more protected roles are mostly physical, care-based, or field-based jobs. Things like wind turbine technicians, solar installers, home health aides, occupational therapists, and physical therapists still look much less exposed. These jobs may still use AI around the edges, but the core value of the work is harder to turn into a purely digital system.
One more pattern worth noticing
This is not just a low-wage disruption story. Many higher-paid, white-collar roles may feel AI pressure sooner because so much of the work is digital, repeatable, and done on a computer. The job may still exist, but the workflow changes and the value shifts away from routine execution toward judgment, strategy, and decision-making.
Core insight
The real lesson here is not “which jobs are doomed.” It is that digitally native work is likely to be reshaped first, and that reshaping may show up before the headline job losses do. If your work mostly happens on a laptop, your role may still grow, but the way you do that work is likely changing faster than many people expect.
Southeast Asia angle
This is based on US job data, so I would treat it as directional signal, not something to copy directly into Southeast Asia. Labor costs, industry mix, digital maturity, language environments, and business structures are different here. That is exactly why I find it useful as a starting point rather than an endpoint.
What is next
I will be doing a Southeast Asia version of this idea in the coming months. The goal is to build a more region-specific view of which roles, sectors, and workflows are likely to feel AI pressure first across this market.
What to do
Open the visualizer and check three things: your own role, the roles directly upstream from you, and the roles directly downstream from you. Then ask a better question than “Will AI replace this?” Ask which parts of the workflow are becoming cheaper, faster, or easier to automate, and where human value becomes more important.
If you care about practical AI adoption, this is worth a look. Not because it gives the final answer, but because it helps you ask better questions about where work is likely to change first.
Official Source Link: Click Here
“If it can be demonstrated, it can be systemized.”
Join our Telegram channel for updates and the free workflow pack:
https://t.me/TheWorkflowLab
Join the Telegram group (community):
https://t.me/+EfkbBrHCf4wOWY1