The convergence tax
Companies using the same models to ask the same questions will get the same answers. Sameness is a cost, and nobody books it.
same model, same prompt, same pitch deck
trying to answer →Why do AI products converge?What is a moat in an AI world?
Ask three well-run companies to use the same frontier model to answer “what should our product strategy be?” and you get three versions of one memo. Nobody did anything wrong. The inputs were the same, so the outputs regress to the same place.
I think of this as a tax. It doesn’t show up on any statement, but it is paid in margin: when execution converges, products become substitutes, and substitutes compete on price. The more of the thinking a company outsources to the shared model, the more of it becomes identical to everyone else’s.Brand designers noticed a smaller version of this in the 2010s, when so many startups arrived at the same clean sans-serif logo that it got a name: “blanding”.
The way out is to feed the model something the others can’t: your own data, your customers’ actual words, a view nobody else holds. Premium moves to the things the model cannot supply.
The Convergence Test is the attempt to measure this instead of asserting it. Until the numbers are in, this stays a note.