[ajuc]

Brainstorming isn't for problem-solving, especially when problems are small and well defined. Communication overhead destroys the efficiency.

I've participated in several team programming competitions. The way every team worked was - everybody read all tasks, quickly decide who works on what, and then we work solo on one problem each in parallel, when somebody finishes (s)he can help others who are stuck or take on another tasks. Talking about the problem all the way was way too slow and didn't much helped.

But problem-solving isn't the only creative thinking people do. When creating a story for table-top RPGs brainstorming works great.

[ouid]

I think the opitmal number of people working on a problem is somewhere between 1 and 2. The amount of communication required obviously scales very poorly, but in principle, having to explain your process to someone else functions well. I prefer pair programming to solo work, usually. I catch more bugs, and am forced to explain things to myself as I explain them to my partner.

[ajuc]

Pair programming works well for tasks that take a few days. For tasks that take hours it's not worth it IMHO.

[ouid]

I agree, and the premise of going in on a programming contest with a team is presumably that you are all capable of solving some problem on your own. I was merely arguing that there is value to be had in numbers greater than 1, but not numbers greater than 2, in my experience.