Speed, quite literally.

Not computation, but ease of implementing a new idea and making a proof of concept.

Researchers try to maximize time spent on iterating through failure, rather than spending a lot of time to perfect a technique. (Generally speaking)