I'm a bit surprised that conda is not mentioned anywhere (I'm not saying as best practice, but as an alternative or why envs are better)

Yup, especially when you are working with os-level dependencies like cuda, conda is the way to go. I think every tensorflow user can approve

…unless you're working on a project for arm64-based IoT devices (like Nvidia Jetson), where Conda is not available.

I use this with my m1 mac, and it works great.