on pair programming

Though I was always joking pairs by this:
Introducing the PairOn
*Orders taken only on April 1st.

Here are two articles I didn't finish reading:
How Pair Programming Really Works
Successfully Adopting Pair Programming

In short, I do really think pair programming
works, nevertheless, I would still need some
private time to refine the produced codes,
because you won't be able to refine the detail
while pairing and producing very fast working
codes. I do think we need to "craft" still,
though I can't find the line clearly right now.
Perhaps in fact good enough is good enough.

Pair programming can really build up a draft
very very quickly... Especially begin with
writing a failing test, and then implement it
to fulfill the requirement. This would be a
very short and fast iterative cycle. I don't
even think much while pair programming...
Just like there's a super smart AI and/or tool
automatically helps and assists you.

This indeed do have many limits though.
For instance, if you are not writing tests,
then it won't help you much... since tests
would force you to focus on a very minimal
set of problems. If the problems were large
and many, it's too hard to make everyone
(including yourself!) knows the problems well.

Sorry, it's not in short now... XD