In general I'm against optimizations done during the coding process itself. Most of the time there's a straightforward solution which works quite well. Optimizing time-critical code is another question.
I'll agree with that, I see too many people breaking projects trying to reduce too many lines of code without rigorous testing to make sure it still does what it did originally.
General rule of thumb:
If code works, don't change it
If it could be changed for the better,
spend more time planning before just crunching,
Use the benefits your've found for your next project
currently working on 3 projects, 102k lines of code
