Where Do You Hide?
I was talking to a friend last week about where you put money when the edges keep dissolving.
We’d both been listening to the Dwarkesh interview with Dario Amodei — the one right after Anthropic’s $30 billion round at a $380 billion valuation. And the conversation kept circling the same question: if AI keeps compressing the value of specialized knowledge, what’s left? Where do you hide?
Not “what cool things will AI do.” The investor’s version. The one where you’re staring at your portfolio wondering which moats are real and which ones are made of sand.
Dario said something during the interview I haven’t been able to shake. He was describing how Claude Code happened. Anthropic’s coding tool. They built it for themselves. Used it internally, saw what it could do, then launched it externally. He even cut off competitors — including OpenAI — from using Claude for coding. That’s how valuable they thought it was.
I’ve seen this before.
Bezos didn’t set out to build AWS. Amazon needed massive computing infrastructure to run its own business. They solved their own problem, then realized every company on Earth had the same problem. AWS wasn’t a product idea. It was a side effect of solving their own hardest challenge.
Dario ran the same play. Anthropic needed better coding tools, so they built them, used them, refined them against their own work, and shipped them to everyone else. Solve your own problem, realize it’s everyone’s problem, sell the solution.
Dario has a neuroscience background. He came from biology, not software. And during the interview he talked about reaching Nobel-Prize-level AI by late 2026 or early 2027. He used the phrase “a country of geniuses in a datacenter.”
My friend and I had been talking about biotech. The brutal economics of it. Billion-dollar trials that take a decade. Ninety percent failure rates. The whole industry built on the assumption that specialized knowledge is scarce and expensive and slow to accumulate.
And then you hear Dario describe a continent of geniuses emerging inside a datacenter, and you think — what happens to the industries built around scarcity of expertise?
Clinical trials are going digital. The data layer is becoming machine-readable in ways it never was. I don’t know what Anthropic’s roadmap looks like. Dario didn’t announce a biotech product. But the guy ran the AWS playbook on coding, he comes from neuroscience, and he’s talking about Nobel-level AI within a year. I don’t think you need a press release to see where the line points.
That’s the exciting version of the story. The investor version is less fun.
The same force that might revolutionize biotech is also dissolving the edges that active investors have relied on for decades. The specialized knowledge that let a healthcare fund manager outperform — knowing the FDA process cold, reading a Phase III readout before the market digests it — that’s pattern recognition. And pattern recognition is exactly what these models are getting good at.
I wrote a while back that the biggest companies capture value because they own distribution. I still think that’s right. But there’s a version of the same argument that applies to investing itself. The “edge” that active managers sell — sector expertise, proprietary research, informational advantages — is built on specialized knowledge.
Specialized knowledge is exactly what’s getting cheaper.
The SPIVA data was already brutal before AI showed up. Over a 20-year period, roughly 90% of actively managed funds underperform their benchmark. Smart people. Real resources. Bloomberg terminals, analyst teams, decades of experience. And they can’t beat the index.
That was the world where informational edges were still real. Where knowing a sector deeply actually meant something.
Now the cost of expertise falls every quarter. The continent of geniuses isn’t just writing code. It’s reading earnings calls, parsing FDA filings, modeling drug trial outcomes, doing it at a scale and speed no team of analysts can match. The edges were already thin. They’re getting thinner.
Morgan Housel wrote about Bezos getting asked what’s going to change in the next ten years. Bezos said that’s the wrong question. The right question is what’s not going to change. People will always want low prices. Fast delivery. Vast selection. You can build a business around those things because they’re permanent.
I keep coming back to that.
What doesn’t change in investing? People will always want compounding. They’ll always want low fees. They’ll always want to not blow up. Time will always be more powerful than intelligence. Patience will always be rarer than skill.
Those are index fund virtues. Low cost, broad diversification, long time horizon, discipline through drawdowns. None of it depends on having an edge. None of it gets disrupted when AI makes expertise cheaper. Nobody’s ever going to wake up and say, “I wish my funds charged higher fees” or “I wish compounding stopped working.”
I think the honest version of where my head is at: I don’t know where the edges are anymore. And I’m not sure anyone else does either, including the people charging 2-and-20 for the privilege of claiming they do.
Over 20 years, almost nobody beats index funds. That was true before AI. Every force I can see — expertise getting cheaper, analysis getting faster, information spreading wider — makes the case for indexing stronger, not weaker.
My friend asked me where you hide when the moats dissolve. I think you stop trying to hide. You buy the whole market, keep your costs near zero, and let time do what it’s always done.
The answer didn’t change. The reasons just got more convincing.