[bunderbunder]

I think this is inevitable. Slicing and dicing the problem into a bunch of tiny sub-problems that can be implemented in short sprints is based on the assumption that:

  1. None (or few) of the problems are very big.

  2. All of the problems can be solved without much brain strain.

  3. A fairly straightforward well-known architecture can be used.

Those assumptions are clearly optimized for projects like business applications, Web front-ends, enterprise services, etc. Situations where nothing depends on somebody getting their brain really deep inside a problem and then marinating on and being frustrated by it until eventually a bolt of inspiration knocks them over while they're washing their hair, causing them to hit their head on the washcloth rack on the way down, so that the idea tragically disappears down the drain along with the mingling blood and. . .

ANYWAY. For skunk works projects, we've already got a different strategy called a skunk works team. Nothing wrong with using those. Right tool for the job and whatnot, eh?