For example when i first read SICP years ago, i was working on Protocol State Machines and was stuck (at work) on how to design a verification framework for the same which our clients needed to satisfy themselves that our implementation worked correctly as the official specifications laid out. SICP gave me enough knowledge to understand the problem and the enthusiasm to design a simple DSL to solve the same. Thus by reading it and without doing a single exercise (nor learning Scheme well) i had a huge ROI on my Time invested in reading that book.
So have confidence in yourself, and read anything and everything that you find interesting without beating up yourself over not doing it "properly". The point is to get exposed to new concepts/ideas and gain new perspectives/insights on already known knowledge.