That Anthropic chart about AI taking over jobs? Don’t panic yet.

You’ve probably seen that chart floating around. The one from Anthropic that compares current AI job exposure (red) to theoretical capability (blue) across 22 occupation categories. At first glance, it looks like LLMs are about to swallow everything from arts and media to legal, business, and management. The blue area covers over 80% of tasks in most categories. Scary stuff.

But here’s the thing: that blue blob isn’t a prediction. It’s closer to a thought experiment with some serious caveats.

Anthropic’s “theoretical capability” numbers come from educated guesses about where LLMs might improve human productivity—not where they’ll replace humans entirely. The methodology relies on expert estimates of what current models could theoretically do if perfectly applied, ignoring real-world constraints like cost, reliability, latency, and the fact that most businesses aren’t going to let an LLM handle their legal contracts without a human in the loop.

I’ve been watching these capability-versus-deployment debates for years now, and this feels like déjà vu. Every major AI lab publishes something similar, and the press runs with the worst-case interpretation. The reality is messier. Even if an LLM can technically perform a task, that doesn’t mean it’ll be cost-effective, safe, or socially acceptable to deploy at scale.

Take the “Arts & Media” category, for example. Sure, an LLM can generate a passable blog post or a generic logo. But creative work isn’t just about output—it’s about context, taste, and iteration. The chart doesn’t capture that nuance. It’s measuring task completion in a vacuum, not real-world value.

The red “observed exposure” area is actually the more honest metric. It tracks what’s already happening in the job market, not what might happen in some idealized future. And that red area is a lot smaller.

Anthropic is a smart company doing interesting work, but this chart is a classic case of letting a flashy visualization outrun the underlying assumptions. The blue area represents potential, not destiny. And potential has a funny way of not translating into reality when you factor in regulation, corporate inertia, and the simple fact that most jobs are more than a checklist of tasks.

So next time someone shares that chart with a panicked caption, remind them that theoretical capability isn’t the same as practical deployment. The future of work is being shaped right now, and it looks a lot more like the red area than the blue one.