Indeed this is not for any environment and projects. So take it with a grain of salt. That said, I think this piece has a core truth to it which is more general. Software architectures shouldn't be considered as something fixed as soon as they are planned, they need to be validated through use and to be prepared to evolve over time as needed.

Need to know if two shapes overlap? Good explanation of an elegant algorithm to do it.

The ordering used for matrix multiplications definitely matters.