> Optimize for human productivity (for both the writer and the many future readers) first, then profile, then optimize any bottlenecks.

I don't agree. The problem with this approach is that there are some optimisations which require changes to how data flows through your system. These sort of refactorings are much more difficult to do after the fact, because they change what is happening at the abstraction / system boundaries.

Personally, my approach is something like this: Optimise first for velocity, usually writing as little code as possible to get something usable on the screen. Let the code be ugly. Then show people and iterate, as you feel out what a better version of the thing you made might look like - both internally (in code) and externally (what you show to humans or to other software systems). Then rewrite it piece by piece in a way thats actually maintainable and fast (based on your requirements).

Actually I think you're misunderstanding me. I'm not saying you should profile and optimize all the code you write, I'm saying that a basic understanding of algorithms, data structures and complexity analysis allows you to write better code without any extra tools. I didn't profile this report to find out why it took 30+ minutes to run. I just happened across some code, read it, saw that it was essentially two nested loops iterating through two huge (50-80k elements each) lists matching items by name, changed it to use a dictionary instead of the inner loop and that was that.

It's a trivial change, it wouldn't have taken any longer to write this the first time around. There is no excuse for doing this, it's just a dev who doesn't understand what they're doing.

That's my point. Understanding these fundamentals allows you to avoid these types of pitfalls and understand when it's okay to write something inefficient and when it isn't.