I tend to agree, but you don't have to move at the same speed as the framework. If you have a version that is working well for your needs, stick with it for a year. Then decide when and where to upgrade. You can let the dust settle on new features, let other people learn the harsh lessons and come up with best practices, and then you follow in their footsteps, stably upgrading your codebase to exactly the point that makes sense.

Or, if you prefer a business cliche, prefer leading edge over bleeding edge.