I've seen so many projects where the developers have taken this to heart way too much. The problem is they use it as an excuse to give up after stage 2 thinking stage 3 can come along when it actually becomes a problem.
The thing is that performance issues tend to creep up on you so slowly that you don't realise it until you have a very important client with tons of data screaming at you to fix it now. To fix it then you have a huge architecture issue because the whole system is so inefficient. So rather than rewrite the entire code base to actually be efficient you are forced to hack in some kind of caching solution. Then your problems multiply.
Make it fast from the start. Premature optimisation being the root of all evil is a lie!
I'm not sure you two actually disagree, but rather that you're thinking on different scales. Make it right, then make it fast is usually talking about specific snippets of code, rather than the whole program, so your whole program should be "made fast from the start" because you went through the process of making it right and then making it fast on the components themselves.
If you do that you'll write great, fast code but it'll take you longer to do it. That's great if you have unlimited resources, but try doing it in a startup where failing to ship in a month means your whole business fails.
Ok, it's a bit of an exaggeration and obviously depends a lot on what your definition of premature is. I guess the statement is actually a tautology, by definition premature means too soon. I just see people who think that any optimisation at all is bad, whereas in a lot of instances optimisation as you are writing the initial code is not premature. In my experience...
The thing is that performance issues tend to creep up on you so slowly that you don't realise it until you have a very important client with tons of data screaming at you to fix it now. To fix it then you have a huge architecture issue because the whole system is so inefficient. So rather than rewrite the entire code base to actually be efficient you are forced to hack in some kind of caching solution. Then your problems multiply.
Make it fast from the start. Premature optimisation being the root of all evil is a lie!