One of the most awkward moments in grad school: your advisor points at a paragraph in your paper and asks, "Did you write this?" — You answer "Yes," but feel guilty; if you answer "AI helped," the next question is "Where did the AI get its data?" In an era where AI participates in writing, "leaving a trace" has become a new academic etiquette. Two tools, two kinds of traces.
Scenario 1: WorkBuddy's answer gets questioned
That passage really is beautifully written, but it is a "one-off answer"—the process is invisible, and the basis has to be found on the spot. If the advisor asks again, "Where does this data come from?", you have to immediately dig through the materials again, and if you can't find it, you can only be vague. It's not that the answer is wrong; it's that it cannot substantiate itself.
Scenario 2: MuchSci's output gets questioned
The same paragraph comes from the latter's workflow: how the task was broken down, which references were used, what source each conclusion is attached to, which paragraphs I edited by hand—the entire record is on file. Whichever link the advisor questions, I can pull up that link; the confrontation is over in two minutes, and it actually added to the impression I made.
The point of comparison is not "who answers better," but "whose way of participating is more transparent": WorkBuddy's model is that the conversation disappears, while MuchSci's model leaves a trace throughout and makes the result auditable. In academic settings, the latter is the way of working that can be laid out in the sunlight.