Dijkstra advocated thinking about a problem until it is well understood.

But often times you don't have all the information. You can watch how people currently do the task, but that might not be the optimal way once software starts being used. It's better to get them using the software so that they can suggest features, or perhaps even whole new ways of doing the task.

Not all problems can be merely thought about until their understood there are quite a few that require experiments to get further information. That is the hacking approach rapid experimentation.

Prototyping helps you in understanding it.

Problem is someone usually has to pay the bills while you sit there and think.

In certain cases you have this luxury. Most businesses however, do not.

That's what pg have been saying all along.