Knowledge is not a moat

By Alex, cofounder of Oddlynew. 30 October 2026.

I think the price of knowing things is heading towards zero, and I think that is mostly good news.

What I'm noticing

I watch agents take a field I would have needed months to learn, research it, and write it down well enough that another agent can act on it. Once it is written down, it can be copied and rewritten in minutes. A playbook that took one person years to build can be summarized, rephrased and republished by anyone's agent the same afternoon.

There is a plain economic result behind this. When goods are identical, competition pushes their price down to the cost of making one more copy. For written knowledge, agents are bringing that cost close to zero, and rewriting in your own words gets around copyright on the wording. A draft working note by Alberto Cavallo, Zoë Cullen and Andrew Koh on the market where AI labs buy expert know-how (A Market for Human Expertise, 12 October 2026) points the same way: the more expertise in a field has already been written down, the fewer new hires and the lower the pay it attracts.

Knowledge was never the valuable part

Knowledge on its own has no value. It is worth what it changes in what you do. Decision theory calls this the value of information: if it changes no decision, it is worth nothing.

Warren Buffett's approach has been public for decades. Anyone can read it, and very few people get his results. The hard part was never knowing the method. It was doing it: the temperament, the conviction, and where he put the capital.

AI makes doing easier too. It does not decide for you which idea to go after, how much conviction to have, or how many agents to put on it. I expect that to become the new capital allocation: not only where your money goes, but where your compute goes.

What is still worth paying for

If a copy of the knowledge is free, what is left is what a copy cannot carry:

The people best placed to offer that are operators, people who keep their knowledge current anyway because they run their own business on it. For them, sharing it costs almost nothing extra.

Who should share, and who should not

Whether sharing makes sense depends on whether the people reading it compete with you.

If they do, like everyone chasing the same search results, keeping it private is rational. If they don't, sharing costs you nothing: a café in Graz is no rival of a café in Vienna. And when many small players together compete with one giant, sharing is how they stay in the game.

That last case has an old answer. EDEKA and REWE, two of Germany's largest grocery groups, began as purchasing cooperatives of independent shopkeepers who bought together to compete with bigger chains. They pooled where scale matters, buying and marketing, and stayed small where small is better.

A café, one day

One thing I'd like to do one day is run a café: pay people well, hire great baristas, make it as good as we can, and write down everything it teaches us. Not to build a chain. Running cafés everywhere is not what we are good at, and a chain of our own would compete with the independent cafés we would rather help. The point is to make the method public, so that more good independent cafés exist, with well-paid baristas and fairly paid roasters.

I build Oddlynew, where people keep the knowledge their agents work from, so I am biased here. Joel Spolsky called the pattern "commoditize your complement": when what sits next to your product gets cheaper, more people use your product. Every café that runs on shared knowledge is the kind of user we want.

What I take from it

If knowledge is not a moat, hoarding it is a weak strategy for most people, and sharing it is cheap. What stays scarce is deciding what to bring into the world, and doing it well. We can make almost anything now, so the question becomes what we choose to make, and I find that a beautiful question to be left with.