I'm not sure I agree completely with his characterization of what happens. I've noticed that people, particularly in larger organizatons, will start shoehorning their "hacking work" into real projects if it is left out for too long. So, you end up with systems that are on bleeding-edge languages, tools, runtimes, and platforms because, frankly, the team was just dirt-sick of Scala/Java + Beans, even though that would've met the company's goals much more quickly and cheaply.

That said, I definitely support his statement that people should have some hacking / learning time budgeted, even if only to avoid the scenario I mentioned above.