It was the arbitrariness of academic work that got to me. Real world work tends to be clearly directed at a goal, and if you acheive the goal it doesn't tend to matter much how you did it. Academic assessments (poorly designed ones, which happens to be most of them in my experience) expect you to do things exactly in a prescribed manner. Essentially 50% of what you're doing isn't actually useful learning, it's domain-specific learning where the domain is "what my professor happens to care about".