Architecture before implementation
The expensive decisions are made before the first line of code.
Code is cheap to write and expensive to unwind. The choices that actually determine whether a system scales — where state lives, where boundaries fall, what each component is allowed to assume — are made at the diagram, not the keyboard.
I spend disproportionate time deciding what not to build and where the seams go. A clear architecture makes most implementation questions answer themselves, and makes the hard ones visible early, while they are still cheap to change.