[susijdjdjxa]

I think this argument fails because “firm-space” is sparse. There are many companies, but there are way more possible companies given all the degrees of freedom. Also, the number of “competition events” is small. Companies don’t have the opportunity to run gradient descent to optimize themselves. A bunch of companies can make the same bad decision for the same bad reason, and they will each succeed or fail because of a bunch of unrelated factors.