True. It also never claims to be.

The very first part says that this is only a writeup for people who are not intimately familiar with why "dynamically typed" might slow down Python.

Based off my own experience, dynamic typing has much less to do with slowness than the plethora of string copying which occurs in your typical python script.

but for scientific computing with numbers - the author's target audience - dynamic typing is the main bottleneck.

Boxing and memory locality are the the main bottlenecks in scientific computations with Python, dynamic typing is orthogonal.

If a language is dynamically typed, it is very likely that it will use boxing, so they do not seem orthogonal to me.

There's ways to avoid that, where it matters, though its a pretty common thing in naive Python/Ruby, true.