As for writing (scientific papers), I find it best to use my most productive hours for editing and the stupid hours for writing new material, under the assumption that I would have to edit new material anyways.
Programming is much the same way. Use your best hours to refine/polish work that is furthest down the line, use your worst hours to write crap that gets you started thinking about what the real solution will look like. Even in your best hours, you probably aren't going to come up with the right solution immediately (unless the problem is easy).
Interesting. I find editing much less mentally straining, so I use my stupid hours (or better yet, unmotivated hours) for editing/refactoring and my good hours for generating the new content. I've noticed that my prose style tends to be very hierarchical and not very free-flowing, so I wonder if there's a correlation there.
This is a very interesting philosophy I would like to hear more about. I suspect it is different for coding? It seems like tracking down all of the connected logic bits would be difficult after writing deliriously.
Coding is much the same way as writing to me; they are both forms of thought development and refinement. But then I write research code, not production code, so life is a bit different for me (prototypes vs. products). If you are applying and checking in a bug fix, you don't want to do that in your off hours. If you are building a compiler/interpreter/editor for a new language from scratch to explore some design ideas, then off hours are useful in pushing things along since there are lots of unknowns that require time to mull over. Also, you get something to sleep on.
Programming is much the same way. Use your best hours to refine/polish work that is furthest down the line, use your worst hours to write crap that gets you started thinking about what the real solution will look like. Even in your best hours, you probably aren't going to come up with the right solution immediately (unless the problem is easy).