I think there are two parts to programming: making a mental model of a problem that can be translated to computer instructions, and doing the translation. There's other ways to express that model than writing code; writing docs, talks, PR reviews. I wish we measured the modeling instead of the tasks