Working with Emmett has helped challenge me to apply some higher order (i.e. meta-level) thinking to my research career. For example, one major weakness of mine is reading and listening critically. I'm used to treating textbooks and lectures as authoritative and focusing on understanding the concepts presented, whereas interacting with research requires not only understanding the content at face value, but also reasoning about a whole host of other issues, like...
It's been an uncomfortable mindset shift to realize that no, just because a research paper was published in a reputable conference doesn't mean it can't have flaws. I've noticed that certain professors and peers are quite skilled at skimming a paper and then voicing criticism, or more generally developing opinions. Maybe this adversarial mindset is also more common in systems research, where baseline comparisons and evaluating against (dunking on) previous systems is the norm.
As for my own workflow, it's similarly not sufficient to just think about how to accomplish my current task. Emmett called me out last year for just collecting and showing data without understanding it. Does the data match what we would expect from napkin math/theoretical bounds? Are there outliers in the data, what caused them, and can they be eliminated? Is this the best way to present the data in a way that anyone can understand? Are there additional experiments we can run to explain some feature of the data?
I am reminded of the principles of intentional practice. It's not effective to repeat the same motions mindlessly; learning requires constant reflection and readjustment.