I think the point is that it is risky to design before you fully understand the problem you're solving. From my interpretation of the slides, they are often working on tasks they have little experience with. Instead of spending time thinking about potential design issues (which may or may not exist), build something that works and the patterns which the feature requires will emerge, enabling an elegant design afterwards.
Having personally been bitten by doing each of these out of order, they make a lot of sense.