I was reading Ryan Levesque's newsletter and a chart he referenced definitely caught my attention. 

He'd seen it from an Anthropic research paper on AI's impact on the labor market, Anthropic being the company behind Claude, the AI model I now use daily. The chart maps two things side by side for every major occupational category: what AI can theoretically do today and what it's actually being used for right now. 

Peter Walker, head of insights at Carta, helpfully redrew it as a bar chart because, as he put it, most radar charts should just be bar charts. He's not wrong.

The gap between those two numbers, depending on your field, is somewhere between jarring and quietly destabilizing. Computer and math, business and finance, management, legal - all sitting with theoretical capability well above 80%. Healthcare practitioners at 58%. And then, notably, healthcare support dropped to 38%.

That last comparison is worth sitting with for a moment. The more a healthcare role involves physical presence, hands-on care, sustained interaction with a patient - the less exposed it is, meaning the less AI is able to “replace” the role. The information and diagnostic side of medicine is more vulnerable than the bedside side. 

As a doctor, I find that clarifying rather than surprising. Information can be processed and output by AI. What happens in a room between two people, over time, cannot.

A small but necessary caveat before we go further: the paper was written by Anthropic about their own AI. Whether they've inflated what their model can theoretically do is a fair question. Anthropic reads it as "look how much room to grow." A skeptic reads it as "look how much it still can't do." I suspect the truth sits somewhere between those two, and I suspect Anthropic isn't far off. 

The wave is on the horizon. It hasn't hit shore yet. But the distance is closing.

For a long time, having information was a form of power. Knowing things other people didn't know was a competitive edge - in medicine, in law, in business, in almost every field. The internet removed that edge slowly, then quickly. Suddenly information was everywhere, and the edge shifted from having it to knowing what to do with it. 

AI doesn't just continue that trend. It tips it over entirely.

"It takes all the running you can do, to keep in the same place” says the Red Queen, from Lewis Carroll’s Alice in Wonderland. This theory became known as the Red Queen Hypothesis in evolutionary biology - describing a dynamic where species have to keep adapting just to maintain their position relative to their environment. 

Not to get ahead. Just to stay in place. 

I keep coming back to this not because I think the answer is to run harder or faster, but because it names something true about what adaptation actually requires. The landscape is shifting. Pretending it isn't, or resenting it, doesn't help anyone, especially not someone building a business or raising a daughter.

My daughter is growing up in a reality where almost anything can be generated - words, images, video. AI will increasingly be the thing handing her information, analysis and even creative work. What I want to teach her is to not fear that and also to critically evaluate everything she sees. 

The same thing applies to any business navigating this moment: the tool is not the thinking. We learned to filter social media (sort of) - we figured out hooks, filming strategies, production that’s designed to make someone seem credible. We developed an instinct, however imperfect, for what was performed versus what was real. 

We haven't developed that same instinct for AI yet. 

It feels more authoritative, complete, like something that has already done the thinking for you. That pause before accepting, the moment of deciding whether something is signal or noise, is what needs protecting.

Which brings me back to the chart. Look at where AI reaches furthest: fields built on data processing and where information equates value. That's also where the most commoditization is already happening. 

Look at where it can't reach as far - the physical, relational, human. The thing that sits outside AI's reach, in every field, isn't more information. It's judgment. The ability to look at a specific situation and know which pattern applies, why this one is different from the last one that looked similar, what the person in front of you actually needs as opposed to what they're asking for.

When I work with someone on their brand, the most valuable thing I bring isn't information. It's pattern recognition - finding the throughline in someone's story that they've been too close to see, connecting things they've told me that they don't yet realize are connected. You could train an AI to interview someone and surface their stories. But the piecing together, the sense of what fits and what doesn't, the ability to say "this is actually what you're about" - that's not a data retrieval thing. It's a finding-patterns-and-making-sense thing.

The chart sheds light on something many people have sensed for a while - tasks a machine can do faster and more accurately get automated first. That has always existed. Now we can see it more clearly and we get to decide how to create around it.