One of Farhan's premises for pair programming is:<p>> Many people assume that the bottleneck with coding is typing, but the bottleneck is actually at the thought level — it’s at deciding what to code<p>but in my experience coding, this is not true at all. Most of the time, it is very easy to figure out what needs to be built. What takes time is:<p>- integrating with broken libraries<p>- fixing said libraries<p>- discovering and fixing edge cases as you code<p>- writing extensive integration and unit tests<p>- making sure the code is well documented<p>- going through code review cycles, fixing nits and tests<p>- refactoring the code to make it easier to modify<p>- setting up rollout mechanisms to migrate from one library/database to the next<p>It's all mundane wiring that takes the most time, and I can't imagine how a second programmer peering over my shoulder while I do these things would make it go faster/better. But maybe this is just true for large companies, where requirements are made pretty clear (versus startups, where engineers contribute significantly to figuring out what needs to be built).